Is it possible that there are no real speed figs?

[Capper Al]

Everyone knows the formulas being used:


Parallel Speed Chart adjustment
Track variant
Daily variant


But consider how the speed figures are adjusted when it comes to the daily variant. If speed makers are trying to peg a daily variant the first thing they say is something like this: These $20,000 Claimers should be running this distance at 1:05(example). They ran a 1:10 so they are 5 over. Then they subtract 5 for the daily variant. (It's not exactly this simple, but overall this is what they do.) So the daily variant would be adjusted with a minus 5 to equalize. Here's the catch, with all this equalizing they are negating the actual speed and going by what the class of horses should of ran. So is it speed or is it class? The speed has been factored out of it.

[cj]

Maybe that is how speed figures were made in 1950. I doubt many do it that way now. I know I don't.

[Flysofree]

Everybody and their brother has a speed figure.What I can't figure out is why or how the "run up distances" are valuable and or needed.

[cj]

Everybody and their brother has a speed figure.What I can't figure out is why or how the "run up distances" are valuable and or needed.

Sometimes they are, sometimes they aren't. Maybe this will help:

http://timeformusblog.com/2013/12/17/run-up-and-its-effect-on-final-time/

[Capper Al]

Maybe that is how speed figures were made in 1950. I doubt many do it that way now. I know I don't.

Beyer's book was in the seventies. Many publications today say they do it this was.

If not, and you can say without giving away trade secrets, what are the basics of today's speed formulas?

[Capper Al]

Sometimes they are, sometimes they aren't. Maybe this will help:

http://timeformusblog.com/2013/12/17/run-up-and-its-effect-on-final-time/

Thanks for the link.

Run-up times can be double or more for Euro races.

[cj]

Beyer's book was in the seventies. Many publications today say they do it this was.

If not, and you can say without giving away trade secrets, what are the basics of today's speed formulas?

I was exaggerating with the 1950s comment of course. I'm not going to give away much, but I don't use pars at all. It is basically impossible at most places any longer. There are so many different conditions you would never have a big enough sample size to have anything remotely reliable.

How would you get a par, for example, for a NW3X allowance race? Are you going to go back a decade to get 15 of them? Are you going to lump circuits together? Regions?

You will also see things like this at one track for 5k claimers:

Open
NW2L
NW3L
NW4L
NW16m
NW1y

What do you do for a par for 2yo MSW? Does it matter if it is June or September? 6f or 8.5f? I could go on and on.

I think by far the best way to make variants is with projections. You look at each and every horse that ran on a card. Instead of having between 1 and 12 data points (based on surface, inner/outer, etc.), you have anywhere from 10 to 100 or more. You'll have to ignore some of them based on distance, surface, trainer changes, etc. But it is still way better than just using a class par for each race.

[Elliott Sidewater]

To answer the OP's original question, no it's not possible that there are no real speed figs. Some are a lot better than others, and figure makers like CJ have to work very hard at maintaining a level of consistency that is a first line priority in this craft. I made my own speed figures for 14 years, and had to give it up in the mid 90's because it was simply too much work for a recreational player. Maybe CJ will chime in on this, but back then I made some of my best and most confident bets when my figures disagreed with the Beyer Speed Figures. How often did this happen? Often enough to matter.

[illinoisbred]

When I 1st began making figures in 1982, I went by the book...par times,time chart. It wasn't too long,perhaps a year,before it started to sink in that my figures were too compressed. There just wasn't that great a difference between say a 5,000 claimer and a 15,000.Also, by the book,young horses presented a huge problem. I was too willing to downgrade what were truly big performances to fit a parameter. Anyways, soon I began projecting figures and it freed my figures from the chains that bound them. Projection is by far and wide the very best method to employ. I only "do" 1 circuit,and for 33 years it was Illinois..now for the past year SoCal.

[CincyHorseplayer]

Reads more like an editorial on disenchantment with numbers Al? There are plenty of people who use no speed/pace figures. I know many who use trainer/pedigree information and that's it. Whatever works.

[Capper Al]

I was exaggerating with the 1950s comment of course. I'm not going to give away much, but I don't use pars at all. It is basically impossible at most places any longer. There are so many different conditions you would never have a big enough sample size to have anything remotely reliable.

How would you get a par, for example, for a NW3X allowance race? Are you going to go back a decade to get 15 of them? Are you going to lump circuits together? Regions?

You will also see things like this at one track for 5k claimers:

Open
NW2L
NW3L
NW4L
NW16m
NW1y

What do you do for a par for 2yo MSW? Does it matter if it is June or September? 6f or 8.5f? I could go on and on.

I think by far the best way to make variants is with projections. You look at each and every horse that ran on a card. Instead of having between 1 and 12 data points (based on surface, inner/outer, etc.), you have anywhere from 10 to 100 or more. You'll have to ignore some of them based on distance, surface, trainer changes, etc. But it is still way better than just using a class par for each race.

Are you saying it is more a horse to horse comparison than a horse to race rating (Class) comparison? How would one escape race rating even in a horse to horse comparison since the race rating is a composite of the horses? Not following.

[cj]

Are you saying it is more a horse to horse comparison than a horse to race rating (Class) comparison? How would one escape race rating even in a horse to horse comparison since the race rating is a composite of the horses? Not following.

You are comparing the times the horse runs to what the horse has run in the past. I look at each horse's last four races compared to what it ran today. So for each race card I can have hundreds of data points. I'm not comparing horse to horse, I'm comparing each horse to itself.

[Capper Al]

To answer the OP's original question, no it's not possible that there are no real speed figs. Some are a lot better than others, and figure makers like CJ have to work very hard at maintaining a level of consistency that is a first line priority in this craft. I made my own speed figures for 14 years, and had to give it up in the mid 90's because it was simply too much work for a recreational player. Maybe CJ will chime in on this, but back then I made some of my best and most confident bets when my figures disagreed with the Beyer Speed Figures. How often did this happen? Often enough to matter.

The discussion really is about the daily variant actually nullify the speed by adjusting to the class of the field? In the OP, there was an example of adjusting the daily variant to what a field of $20,000 claimers at a certain track are expected to run. By making this adjustment, aren't they nullifying speed?

I know it's a lot of work making your own Beyer style figures with parallel charts, track variants, and daily variants.

[Capper Al]

You are comparing the times the horse runs to what the horse has run in the past. I look at each horse's last four races compared to what it ran today. So for each race card I can have hundreds of data points. I'm not comparing horse to horse, I'm comparing each horse to itself.

That is a very interesting alternative approach. Thanks for sharing.

[Capper Al]

When I 1st began making figures in 1982, I went by the book...par times,time chart. It wasn't too long,perhaps a year,before it started to sink in that my figures were too compressed. There just wasn't that great a difference between say a 5,000 claimer and a 15,000.Also, by the book,young horses presented a huge problem. I was too willing to downgrade what were truly big performances to fit a parameter. Anyways, soon I began projecting figures and it freed my figures from the chains that bound them. Projection is by far and wide the very best method to employ. I only "do" 1 circuit,and for 33 years it was Illinois..now for the past year SoCal.

Without giving away any secrets if you can, what do you project on, horse to field?

[Capper Al]

Reads more like an editorial on disenchantment with numbers Al? There are plenty of people who use no speed/pace figures. I know many who use trainer/pedigree information and that's it. Whatever works.

Cincy,

I'm in awe of those who can do pedigree/trainer. These usually are real horse people, not us guys playing with numbers.

Ciao

[illinoisbred]

Without giving away any secrets if you can, what do you project on, horse to field?
Pretty much the same as CJ stated in post #12. I project the individual horse off his/her previous efforts.

[classhandicapper]

That is a very interesting alternative approach. Thanks for sharing.

You should read Beyer's books. He goes over creating projection variants.

The major problem with projection variants is that they are subjective. They are only as good as the figure maker is at interpreting the results of races. That's why you see multiple competent figure makers disagree so often about the same race. Also, subtle little biases in thinking can cause your figures to drift faster or slower over time unless you have some way of anchoring them (which others would argue causes still other problems) .

[cj]

You should read Beyer's books. He goes over creating projection variants.

The major problem with projection variants is that they are subjective. They are only as good as the figure maker is at interpreting the results of races. That's why you see multiple competent figure makers disagree so often about the same race. Also, subtle little biases in thinking can cause your figures to drift faster or slower over time unless you have some way of anchoring them (which others would argue causes still other problems) .

What he wrote about is a long, long way from what I do now.

[Stillriledup]

You are comparing the times the horse runs to what the horse has run in the past. I look at each horse's last four races compared to what it ran today. So for each race card I can have hundreds of data points. I'm not comparing horse to horse, I'm comparing each horse to itself.

I think that's what thorograph does too.

[thaskalos]

You are comparing the times the horse runs to what the horse has run in the past. I look at each horse's last four races compared to what it ran today. So for each race card I can have hundreds of data points. I'm not comparing horse to horse, I'm comparing each horse to itself.

Yes...but those four past races don't exist in a vacuum; their figures are determined, at least theoretically, by the four races that preceded THEM. And those four races were influenced by the four races that were run even EARLIER...and so on down the line. In order to compare the horse "against itself"...we have to form some sort of definition of the horse's "self", as it was BEFORE we could have four races at our disposal to base our assessment on.

How do we form a numerical opinion about a horse when it only has one or two prior starts...and we can't review its last four races in order to make up our mind?

[illinoisbred]

Yes...but those four past races don't exist in a vacuum; their figures are determined, at least theoretically, by the four races that preceded THEM. And those four races were influenced by the four races that were run even EARLIER...and so on down the line. In order to compare the horse "against itself"...we have to form some sort of definition of the horse's "self", as it was BEFORE we could have four races at our disposal to base our assessment on.

How do we form a numerical opinion about a horse when it only has one or two prior starts...and we can't review its last four races in order to make up our mind?
Just speaking for myself here... I don't make projections on very lightly raced young horses. I generally look at the older horses in the races before and after and strive to obtain the "best fit" off older established runners. There can be real havoc when a track runs 2,3 consecutive maiden races for 2,3 yr olds,particularly if you suspect the surface speed may have changed during the time interval.

[EMD4ME]

Cincy,

I'm in awe of those who can do pedigree/trainer. These usually are real horse people, not us guys playing with numbers.

Ciao

As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.

[illinoisbred]

As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.
That might be jumping to conclusions. I'm pretty sure most competent figuremakers also use much other data in arriving at wagering decisions. Pace and speed figures are a tool..a big tool...the biggest wrench in the toolbox for many.

[thaskalos]

Just speaking for myself here... I don't make projections on very lightly raced young horses. I generally look at the older horses in the races before and after and strive to obtain the "best fit" off older established runners. There can be real havoc when a track runs 2,3 consecutive maiden races for 2,3 yr olds,particularly if you suspect the surface speed may have changed during the time interval.

It doesn't have to be a young horse. It could be a foreign older horse, coming to this country for the first time. Or a shipper from a "better" circuit, which finds the local stock much more to his liking, and improves dramatically in his local races. Or an old claimer who suddenly awakens...once getting claimed by a "miracle-working" trainer.

There are times when the past about a horse must be discarded...and we must begin anew in our grading of the horse. How do we progress in those cases where we DON'T have four relevant past races to rely on for our calculations? What do we use for a bench-mark then?

[Flysofree]

Is there a simple explanation, that even a slow learner like me can understand WHY times include "run up distances" and are not simply timed from the gate?

If it's a complicated math formula than that lets me out..

[Capper Al]

As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.

I use trainer stats also, but couldn't play on only trainer/pedigree.

[EMD4ME]

That might be jumping to conclusions. I'm pretty sure most competent figuremakers also use much other data in arriving at wagering decisions. Pace and speed figures are a tool..a big tool...the biggest wrench in the toolbox for many.
Maybe I wasn't clear, my fault.

I wasn't saying that fig makers don't. I was replying to AL's comment about how certain players only use some tools and can't rely upon only a few tools.

[Capper Al]

You should read Beyer's books. He goes over creating projection variants.

The major problem with projection variants is that they are subjective. They are only as good as the figure maker is at interpreting the results of races. That's why you see multiple competent figure makers disagree so often about the same race. Also, subtle little biases in thinking can cause your figures to drift faster or slower over time unless you have some way of anchoring them (which others would argue causes still other problems) .

Can't say I remember his projection time. But might that still involve a daily variant?

[Capper Al]

Is there a simple explanation, that even a slow learner like me can understand WHY times include "run up distances" and are not simply timed from the gate?

If it's a complicated math formula than that lets me out..

I believe it's just how they setup their tracks. They all set them up differently.

[thaskalos]

As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.

Don't be so shocked. Most bettors out there are "specialists", whether they realize it or not. To some of us, myself included, looking at "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" is not only unnecessary...it could even be called UNDESIRABLE...because a horse that is validated by "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" often gets the short end of the price on the board.

[illinoisbred]

It doesn't have to be a young horse. It could be a foreign older horse, coming to this country for the first time. Or a shipper from a "better" circuit, which finds the local stock much more to his liking, and improves dramatically in his local races. Or an old claimer who suddenly awakens...once getting claimed by a "miracle-working" trainer.

There are times when the past about a horse must be discarded...and we must begin anew in our grading of the horse. How do we progress in those cases where we DON'T have four relevant past races to rely on for our calculations? What do we use for a bench-mark then?
Agree. Familarity with a circuit does help. I know both you and me know who the "usual suspects"...the "miracle-workers" are on the Chicago circuit. Foreign stock does pose a problem for me since moving my play to SoCal, but thus far, in a year, there haven't really been any that have come out as serious runners. They've pretty much been sandwiched in between the established runners..or behind them by finish position in most cases. But yes, there can be difficult situations and again,a maker just looks to get a "good fit" in and around the strangers.

[illinoisbred]

Maybe I wasn't clear, my fault.

I wasn't saying that fig makers don't. I was replying to AL's comment about how certain players only use some tools and can't rely upon only a few tools.
Understand...no problem.

[EMD4ME]

Don't be so shocked. Most bettors out there are "specialists", whether they realize it or not. To some of us, myself included, looking at "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" is not only unnecessary...it could even be called UNDESIRABLE...because a horse that is validated by "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" often gets the short end of the price on the board.

Maybe I wasn't clear here also.

I didn't say that a horse needs to be "validated" by everything. I meant that I like to look at everything.

Snake Oil Charlie was a hopeless 80/1 at Belmont this summer. Was 0/13. Near dead last finish in every start in his life. Dirt and Turf.

I looked at the Dam's breeding. All Turf ROUTE. Threw him into pics in Snake Oil Charlie's FIRST TURF ROUTE. It was a crappy field. By looking at every single horse, inside and out (all replays, trainers tendencies, breeding, etc.) I knew the whole field sucked. Snake Oil Charlie had a chance, a small chance to show improvement.

He did and paid near $200 to win.

That's what I'm talking about. :)

[dasch]

I don't agree with a daily variant because things change from race to race(pace,run-up,trip,wind,etc), to me this is just guesstimating at best.

I also don't agree with projecting because you will still use some opinion on coming to what number you end up giving even if marginal.

What I do is REMOVE pace and trip(wideness) from a horse's performance leaving me with what I call the "core" number. A horse's "speed" or "performance" rating can vary by many factors but my core numbers don't change much. This of course requires accurate info(lengths beaten etc) but there is almost zero opinion on my part involved in making the number.

In the last week Little Curlin at 47-1 on Sunday and Generosidade at 71-1 the previous Sunday were both TOP win contenders of mine based on their core numbers. NOT close so you had to include because of their odds but TOP win core numbers. Some people will see this as a redboard but im just giving an example of if you can figure out a way to do things differently there STILL is tremendous value out there and I am not giving anything away just by telling you WHAT I do as the most important part and what it has taken me years to figure out is the HOW.

[EMD4ME]

I don't agree with a daily variant because things change from race to race(pace,run-up,trip,wind,etc), to me this is just guesstimating at best.

I also don't agree with projecting because you will still use some opinion on coming to what number you end up giving even if marginal.

What I do is REMOVE pace and trip(wideness) from a horse's performance leaving me with what I call the "core" number. A horse's "speed" or "performance" rating can vary by many factors but my core numbers don't change much. This of course requires accurate info(lengths beaten etc) but there is almost zero opinion on my part involved in making the number.

In the last week Little Curlin at 47-1 on Sunday and Generosidade at 71-1 the previous Sunday were both TOP win contenders of mine based on their core numbers. NOT close so you had to include because of their odds but TOP win core numbers. Some people will see this as a redboard but im just giving an example of if you can figure out a way to do things differently there STILL is tremendous value out there and I am not giving anything away just by telling you WHAT I do as the most important part and what it has taken me years to figure out is the HOW.

Can you expand on those thoughts please?

In writing vs in person, things can get construed. Just want to see what you mean.

Thank you

[thaskalos]

Maybe I wasn't clear here also.

I didn't say that a horse needs to be "validated" by everything. I meant that I like to look at everything.

Snake Oil Charlie was a hopeless 80/1 at Belmont this summer. Was 0/13. Near dead last finish in every start in his life. Dirt and Turf.

I looked at the Dam's breeding. All Turf ROUTE. Threw him into pics in Snake Oil Charlie's FIRST TURF ROUTE. It was a crappy field. By looking at every single horse, inside and out (all replays, trainers tendencies, breeding, etc.) I knew the whole field sucked. Snake Oil Charlie had a chance, a small chance to show improvement.

He did and paid near $200 to win.

That's what I'm talking about. :)

I understood what you said, EMD...but yours is a confusing method of play, IMO...for me at least. What do I do when my breeding stats conflict with the discoveries of my pace/speed methods? Is a horse like Snake Oil Charlie, who finishes dead-last repeatedly when running in turf sprints, suddenly elevated to win-contender status when switching to a turf ROUTE...simply because the breeding stats suggest a fondness for the turf routes? How often do we see turf routes won by horses who were running atrociously in turf sprints?

"The whole field sucked"...you say. Well...not to argue with you, but I don't categorize whole fields in terms like that. To me, whole fields don't "suck", not are they "GREAT". Whether it's the Breeders Cup Classic, or a lowly $5,000 claimer...it presents its own challenges, and features its own "stars". I am happy for you for cashing big on Snake Oil Charlie...but I would never have made that bet myself.

[dasch]

Can you expand on those thoughts please?

In writing vs in person, things can get construed. Just want to see what you mean.

Thank you

I remove pace(plus or minus) and then remove/adjust for the wideness of the trip. To do this accurately I watch every replay and make my own trip/wideness notes and adjustments. What I am left with is an EXACT number. NO opinion, no projection, no variant added in.

This probably didn't help clear things up much as the important info lies in the method but I wont go into any details about that nor would I ask or expect anybody to provide details on the work they have spent years working on.

[EMD4ME]

I understood what you said, EMD...but yours is a confusing method of play, IMO...for me at least. What do I do when my breeding stats conflict with the discoveries of my pace/speed methods? Is a horse like Snake Oil Charlie, who finishes dead-last repeatedly when running in turf sprints, suddenly elevated to win-contender status when switching to a turf ROUTE...simply because the breeding stats suggest a fondness for the turf routes? How often do we see turf routes won by horses who were running atrociously in turf sprints?

"The whole field sucked"...you say. Well...not to argue with you, but I don't categorize whole fields in terms like that. To me, whole fields don't "suck", not are they "GREAT". Whether it's the Breeders Cup Classic, or a lowly $5,000 claimer...it presents its own challenges, and features its own "stars". I am happy for you for cashing big on Snake Oil Charlie...but I would never have made that bet myself.

I have 2 categories of races (in terms of trip notes). Was it a performance field or a non performance field. For me, when today's field has 1 or 2 or more horses who performed (via poor trip or solid trip) in their last race, then I throw a horse like Snake Oil Charlie out as he would need to improve by leaps and bounds to beat a solid contender. If it's a non performance race, where every contender has been poor or non impressive under optimum circumstances in recent races, then I look for anything that can improve.

The aforementioned assumes similar distance, similar circumstances, no layoff horses etc.

By the way, we don't argue :) I love to interact and learn from eachother. It's a discussion.

[EMD4ME]

I remove pace(plus or minus) and then remove/adjust for the wideness of the trip. To do this accurately I watch every replay and make my own trip/wideness notes and adjustments. What I am left with is an EXACT number. NO opinion, no projection, no variant added in.

This probably didn't help clear things up much as the important info lies in the method but I wont go into any details about that nor would I ask or expect anybody to provide details on the work they have spent years working on.

Thanks dasch. I thought I knew where you were coming from. Now I know. Great post.

I don't expect you to divulge more. Was just wondering if where you were coming from is where I thought you were coming from.

[Capper Al]

As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.

Like I said, I am in awe of those who can only use trainer/pedigree. I use almost every factor.

[Nitro]

I’m sometimes curiouslly intrigued with threads like this only because many years ago I used to do the same thing. Although CJ is correct to some degree when he stated: I'm not going to give away much, but I don't use pars at all. It is basically impossible at most places any longerThere are so many different conditions you would never have a big enough sample size to have anything remotely reliable.But I would think that certain things like knowing the local horse population at your favorite track really haven’t changed all that much. So looking at conditions within a specific population can reduce the sample size significantly. So I completely agree with:Agree. Familarity with a circuit does help.
I found that the one thing that really firmed up those PAR values for adjusting my figures was to average the PAR values for all the races at same distance (and track surface). My figures were so accurate that I was able to predict the running times of races the day after the PARS were created. (assuming there weren’t any changes to the track due to the weather). But as much fun as that all was I found a much more powerful method for predicting the future.


As a guy who looks at everything and I mean eeeeevvvveeeerrrryyyyytttthhhhhiiiinnnnggggg (pace figs, speed figs, ALL replays, sire stats, dam stats, dam's dam's stats. dam's sires stats, trainer intent, trainer %, track biases, j/tr tedencies, charts, competition of each horse in each race for all runner's last 5 races, betting patterns, etc.) I am SHOCKED that people out there actually only look at a couple of things and not everything.Well you're really going to be SHOCKED to find out that there are some like myself who only look at ONE FACTOR. And I don’t see it as part of “everything” listed in your comment! As far as I’m concerned you’ve left out the most significant aspect of playing the horses in real time. You can view and subjectively analyze all of the past history on these horses and yet the most critical factor is missing from the equation: Current Condition on Race Day.

Cincy,
I'm in awe of those who can do pedigree/trainer. These usually are real horse people, not us guys playing with numbers.
CiaoThe REAL horse people know a bit more then that! Just curious Al. If you’re in awe about the “Insiders” how would feel about some Outsiders who are in touch with Insider intentions in real time?
.
.
.

[ultracapper]

You are comparing the times the horse runs to what the horse has run in the past. I look at each horse's last four races compared to what it ran today. So for each race card I can have hundreds of data points. I'm not comparing horse to horse, I'm comparing each horse to itself.

Isn't that what Beyer does? He finds a number of horses that ran as expected, to start the process, and then goes from there?

Edit: I should have gotten deeper into the discussion before asking a ground-bottom question.

[Elliott Sidewater]

The discussion really is about the daily variant actually nullify the speed by adjusting to the class of the field? In the OP, there was an example of adjusting the daily variant to what a field of $20,000 claimers at a certain track are expected to run. By making this adjustment, aren't they nullifying speed?

I know it's a lot of work making your own Beyer style figures with parallel charts, track variants, and daily variants.
No, the daily variant does not nullify the speed. There's a big difference between adjustment and nullification. The variant is used to adjust the speed. You really should read Beyer's 1st book Picking Winners, it explains the process in detail and the basics have not changed since the 70's.

[steveb]

Pretty much the same as CJ stated in post #12. I project the individual horse off his/her previous efforts.

to do that then you would need a pace factor which then means its not really a speed figure, because in most instances what any individual horse will do depends on the race pace, that is, what the leading horse/s are doing.

speed figures rightfully should have huge variations in my opinion that reflects the different ways races are run, the different makeup of the race, and so on.
i don't and never have believed in projections.

then you could you make additional figures that might give adjustments to the individual runners depending on how they went or were suited, not to the race as a whole

[illinoisbred]

to do that then you would need a pace factor which then means its not really a speed figure, because in most instances what any individual horse will do depends on the race pace, that is, what the leading horse/s are doing.

speed figures rightfully should have huge variations in my opinion that reflects the different ways races are run, the different makeup of the race, and so on.
i don't and never have believed in projections.

then you could you make additional figures that might give adjustments to the individual runners depending on how they went or were suited, not to the race as a whole
Correct...and I do incorporate the pace factor after the fact,when the pace is known. I didn't go into any deeper detail in my earlier posts than necessary. But yes, final figures fluctuate depending on the pace that is set.

[cj]

Yes...but those four past races don't exist in a vacuum; their figures are determined, at least theoretically, by the four races that preceded THEM. And those four races were influenced by the four races that were run even EARLIER...and so on down the line. In order to compare the horse "against itself"...we have to form some sort of definition of the horse's "self", as it was BEFORE we could have four races at our disposal to base our assessment on.

How do we form a numerical opinion about a horse when it only has one or two prior starts...and we can't review its last four races in order to make up our mind?

Like I said...I won't give away everything I do. I tried all races, last 10, last 5, etc and 4 was the number I liked best. That said, not all the races are weighted equally and some are tossed as unreliable. All the factors you mention are things considered in my weighting of how likely a figure is to be reliable in determining the speed of today's track.

Even then all races don't always fit into a neat puzzle which is why we have ? next to some races.

[Nitro]

Correct...and I do incorporate the pace factor after the fact,when the pace is known. I didn't go into any deeper detail in my earlier posts than necessary. But yes, final figures fluctuate depending on the pace that is set.Looking at the raw Pace of a race is pretty much a “Catch-22” when it comes to adjusting speed figures. That’s because not only is it impacted by the running styles of the entries in a race, their running styles will ALWAYS be influenced by the track conditions (which includes Wind velocity). I used to deal with this by establishing PAR values not only for the Final time, but every call in the race. When making the final adjustment the inner calls would definitely be considered.

[VigorsTheGrey]

Are BEYER figures adjusted to track variance before published in DRF?

For that matter, what exactly does Andy include in his figure?

[VigorsTheGrey]

Like I said...I won't give away everything I do. I tried all races, last 10, last 5, etc and 4 was the number I liked best. That said, not all the races are weighted equally and some are tossed as unreliable. All the factors you mention are things considered in my weighting of how likely a figure is to be reliable in determining the speed of today's track.

Even then all races don't always fit into a neat puzzle which is why we have ? next to some races.

How reliable is it to use Beyer figures for races that are not run at today's distance/ surface/ venue? Should I only use Beyer listed for ones that match today's conditions?

[VigorsTheGrey]

Don't be so shocked. Most bettors out there are "specialists", whether they realize it or not. To some of us, myself included, looking at "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" is not only unnecessary...it could even be called UNDESIRABLE...because a horse that is validated by "eeeeevvvveeeerrryyyyytttthhhhhiiiinnnnggggg" often gets the short end of the price on the board.

How many hours before betting a race does your handicapping begin?

I know guys that have horses to watch lists going, their own virtual stable, then just wait for these horses to show up...trip handicappers you know....

How do their methods differ from the guy who just opens the form up and starts 'capping?

I used to think that each race was/ is a puzzle that could be figured out...but I'm starting to use other approaches like being up on horses coming out of Keeneland auctions, understanding form cycles, expecting the unexpected...using my intuition regarding trainers intentions, etc...

[VigorsTheGrey]

Is there a simple explanation, that even a slow learner like me can understand WHY times include "run up distances" and are not simply timed from the gate?

If it's a complicated math formula than that lets me out..

What I understand is that the run-up distance is at least partially a track maintenance issue....and I could be very wrong about all of this and I am only speculating here....for the grass anyway, the crew doesn't want to keep putting the gate in exactly the same place all they time cause it makes ruts or compacts differently if they did...But then again I heard some track have published run-up distances so I don't know about that....Also, when the dogs are up and the rail is outwardly placed this must affect the run up distance....Gulfstream I heard has a lot of variance in this regard and heard that their speed figures are very unreliable for turf races by massive amounts...

[Cratos]

No, the daily variant does not nullify the speed. There's a big difference between adjustment and nullification. The variant is used to adjust the speed. You really should read Beyer's 1st book Picking Winners, it explains the process in detail and the basics have not changed since the 70's.
I not sure what you mean by nullify speed, but speed is retarded, reduced, or stopped (nullified) by any interruption in its path and this is verified by Newton's 1st Law of Motion which states in part "a body in motion continues to move at a constant velocity (speed) unless acted upon by an external force."

In horseracing the horse (body in motion) is acted upon by 4 primary external forces which are surface resistance force, wind resistance force, air resistance force, and side force in the turn.

You suggested that the poster should read Beyer's "Picking Winners" which I have read but I didn't see anything in it remotely near Newton's Law; theoretical or applied.

[Cratos]

Everybody and their brother has a speed figure.What I can't figure out is why or how the "run up distances" are valuable and or needed.
The run-up have a functional value by the racetrack as being part of the race and it is because the racetrack apparently want "running race starts."

To do that there has to be a "standing start"; hence the run-up.

However we don't live in a perfect world and some races don't have run-ups; and run-ups have different distances because of lack of available track area and different race distances.

Also with some elementary calculus of physics you can apply Newton's 2nd Law and calculate the magnitude of the horse's acceleration in the run-up.

[Elliott Sidewater]

I not sure what you mean by nullify speed, but speed is retarded, reduced, or stopped (nullified) by any interruption in its path and this is verified by Newton's 1st Law of Motion which states in part "a body in motion continues to move at a constant velocity (speed) unless acted upon by an external force."

In horseracing the horse (body in motion) is acted upon by 4 primary external forces which are surface resistance force, wind resistance force, air resistance force, and side force in the turn.

You suggested that the poster should read Beyer's "Picking Winners" which I have read but I didn't see anything in it remotely near Newton's Law; theoretical or applied.
He used the word nullify. I was just responding to his OP. There was no intention to discuss or debate Newtonian physics. This was about speed figures, the numerical components being raw speed and variant. Period. How many books would Beyer have sold under the title "The effects of friction and air resistance on race horses; a treatise"?

[Cratos]

He used the word nullify. I was just responding to his OP. There was no intention to discuss or debate Newtonian physics. This was about speed figures, the numerical components being raw speed and variant. Period. How many books would Beyer have sold under the title "The effects of friction and air resistance on race horses; a treatise"?
I don't want to turn this into an argument, but an understanding of basic physics is essential in the understanding of "speed" in horseracing.

I couldn't care less about how many books could have been sold by Beyer, but I do believe that if a basic Newtonian explanation of speed had been given in his book as opposed to anecdotal rhetoric, sales of the book would have probably been better.

This is a fundamental problem in horseracing; explaining the "what" and not the "why."

[jasperson]

I was exaggerating with the 1950s comment of course. I'm not going to give away much, but I don't use pars at all. It is basically impossible at most places any longer. There are so many different conditions you would never have a big enough sample size to have anything remotely reliable.

How would you get a par, for example, for a NW3X allowance race? Are you going to go back a decade to get 15 of them? Are you going to lump circuits together? Regions?

You will also see things like this at one track for 5k claimers:

Open
NW2L
NW3L
NW4L
NW16m
NW1y

What do you do for a par for 2yo MSW? Does it matter if it is June or September? 6f or 8.5f? I could go on and on.

I think by far the best way to make variants is with projections. You look at each and every horse that ran on a card. Instead of having between 1 and 12 data points (based on surface, inner/outer, etc.), you have anywhere from 10 to 100 or more. You'll have to ignore some of them based on distance, surface, trainer changes, etc. But it is still way better than just using a class par for each race.
I agree with you C J and you have not thrown in state breds. This more than ever have convinced me that keeping speed figures and pace figures as separate factors is the way to go. Run up distance and wind has more effect on 1f time than the time for the finish. On 1 mile tracks the distance is a factor because part of 4f pace is on the turn in the shorter sprints. I hate Gulfstream because almost half the races are on turf and that dilutes the sample size for computing the dirt track variant. This is also true at Aq and Bel in the summer.

[cj]

How reliable is it to use Beyer figures for races that are not run at today's distance/ surface/ venue? Should I only use Beyer listed for ones that match today's conditions?

I haven't used Beyer figures for a few decades so you are asking the wrong guy.

[cj]

Also with some elementary calculus of physics you can apply Newton's 2nd Law and calculate the magnitude of the horse's acceleration in the run-up.

Sure, if you know the actual run up. What is posted in the charts (and data files, same data) often does not match reality.

[cj]

He used the word nullify. I was just responding to his OP. There was no intention to discuss or debate Newtonian physics. This was about speed figures, the numerical components being raw speed and variant. Period. How many books would Beyer have sold under the title "The effects of friction and air resistance on race horses; a treatise"?

Cratos can never resist getting into speed figure threads and doing the old "you're way is wrong" post. He never really tells you what he does, though.

[cj]

I don't want to turn this into an argument, but an understanding of basic physics is essential in the understanding of "speed" in horseracing.

I couldn't care less about how many books could have been sold by Beyer, but I do believe that if a basic Newtonian explanation of speed had been given in his book as opposed to anecdotal rhetoric, sales of the book would have probably been better.

This is a fundamental problem in horseracing; explaining the "what" and not the "why."

Yes, I'm sure people would have flocked to book stores to read that type explanation of speed figures. :lol: :lol: :lol: :lol: :lol:

[Cratos]

Cratos can never resist getting into speed figure threads and doing the old "you're way is wrong" post. He never really tells you what he does, though.
Yes I do; I use the principles of science and math to the extent that they are applicable and all of those principles are public knowledge.

[cj]

Yes I do; I use the principles of science and math to the extent that they are applicable and all of those principles are public knowledge.

:sleeping:

[Tom]

Here it is, CJ......hot off the presses.
All our questions answered!

[cj]

Cratos every time a speed figure thread he doesn't like pops up.

http://i3.kym-cdn.com/entries/icons/facebook/000/017/966/jordan-crying.jpg

[PaceAdvantage]

Are BEYER figures adjusted to track variance before published in DRF?

For that matter, what exactly does Andy include in his figure?This thread was going along quite nicely until the sound of screeching brakes appeared right about here...just sayin' what everyone else is thinking... :lol:

[cj]

This thread was going along quite nicely until the sound of screeching brakes appeared right about here...just sayin' what everyone else is thinking... :lol:

Yes, pretty much what always happens in threads about speed figures.

[Cratos]

Yes, I'm sure people would have flocked to book stores to read that type explanation of speed figures. :lol: :lol: :lol: :lol: :lol:
As usual you missed the thesis of my post.

It is understanding through the application of science and math.

What I find disingenuous about your retort is that you started another thread about the awareness and making known to the new owner/trainer the various drugs given to a claimed horse.

From what I read, the thread and the supporting info within the thread puts me in the affirmative.

However when the same is said about "speedfigures" you have a different response.

[cj]

As usual you missed the thesis of my post.

It is understanding through the application of science and math.

What I find disingenuous about your retort is that you started another thread about the awareness and making known to the new owner/trainer the various drugs given to a claimed horse.

From what I read, the thread and the supporting info within the thread puts me in the affirmative.

However when the same is said about "speedfigures" you have a different response.

I have no idea how you tied these two things together or what you are trying to say. I'm probably just a poor reader. I'll move on.

[classhandicapper]

What he wrote about is a long, long way from what I do now.

What everyone else is doing is long way from what I do now, which is to ignore figures altogether. ;) But if he wants to get the gist of what projection variants are about, those Beyer books are a good start.

[ReplayRandall]

The OP's question, "Is it possible that there are no real speed figs? is yes, because most are either obsolete or overexposed to the point of NO edge, in the context of profitability. IMO, all that's left for finding a profitable edge are PACE figures, select workout reports, tote-board analysis and replays...

[cj]

What everyone else is doing is long way from what I do now, which is to ignore figures altogether. ;) But if he wants to get the gist of what projection variants are about, those Beyer books are a good start.

For a guy that ignores them you sure like jumping into every speed figure thread.

[NorCalGreg]

The OP's question, "Is it possible that there are no real speed figs? is yes, because most are either obsolete or overexposed to the point of NO edge, in the context of profitability. IMO, all that's left for finding a profitable edge are PACE figures, select workout reports, tote-board analysis and replays...


He didn't ask if speed figures were relevant, profitable, or any such thing--he asked if--since speed ratings are based on pars--and pars are based on class-levels..then are speed ratings just a function of CLASS?

He could have said ....DO PACE FIGURES REALLY EXIST? Pace figs are speed figures of sections of the race, in whole. So pace is a function of speed-- speed is a function of class--so, IS HANDICAPPING ULTIMATELY BASED ON CLASS?
That question is about as useful and relevant as anything in this thread.

[ReplayRandall]

He didn't ask if speed figures were relevant, profitable, or any such thing--he asked if--since speed ratings are based on pars--and pars are based on class-levels..then are speed ratings just a function of CLASS?

He could have said ....DO PACE FIGURES REALLY EXIST? Pace figs are speed figures of sections of the race, in whole. So pace is a function of speed-- speed is a function of class--so, IS HANDICAPPING ULTIMATELY BASED ON CLASS?
That question is about as useful and relevant as anything in this thread.

Does a horse know what "class" he's in?.... :lol:

[thaskalos]

Does a horse know what "class" he's in?.... :lol:
No...but the TRAINER does. And if the trainer thinks that his horse is outclassed, then the horse ain't winning the race...no matter what the HORSE thinks. :)

"Class" matters...even if it exists only in the trainer's mind.

[classhandicapper]

For a guy that ignores them you sure like jumping into every speed figure thread.

I understand them well enough to know all the problems that have caused me so much intellectual grief over the years. I'm trying to save people a few decades and some hair. ;)

[ReplayRandall]

No...but the TRAINER does. And if the trainer thinks that his horse is outclassed, then the horse ain't winning the race...no matter what the HORSE thinks. :)

"Class" matters...even if it exists only in the trainer's mind.

Ahh, the mystery of class, the unsolvable question......especially with the chaos of today's class conditions.

[classhandicapper]

Ahh, the mystery of class, the unsolvable question......especially with the chaos of today's class conditions.

This how the typical discussion on class goes.

1. Someone uses a definition of class they read in some handicapping book written 50-60 years ago.

2. Everyone agrees that class does not exist (based on that definition).

The problem with that discussion is that assumes the definition from 50-60 years ago was the right one. IMO that is not the case. It also assumes that looking at the quality of competition a horse has been competing against while earning his figures or compared to other contenders does not matter at all. That is demonstrably false.

Some kid is scoring 25 points per 36 minutes competing in the NCAA. Will he be able to do that in his first year in the NBA?

Golden State lost to San Antonio by a few points and Knicks beat The Lakers. Who is the class, the Knicks because they won or Golden State even though they lost?

Two teams had the same record and point differential against the same competition in the regular season. They are about to meet in the playoffs with the pressure ramped up. One has playoff tested players that have won multiple championships and the other has several key players that have never been to the playoffs.

At even money who do you want?

Class and speed are just different paths to the same answer. You look at both, try to get better at both, and go with whatever gets you the best results in different circumstances, including both.

[castaway01]

This how the typical discussion on class goes.

1. Someone uses a definition of class they read in some handicapping book written 50-60 years ago.

2. Everyone agrees that class does not exist (based on that definition).



Funny you say that, because this thread was supposed to be about speed figures. And what actually happened was:

1. Someone uses a definition of how to make figures they read in some handicapping book written 40 years ago.

2. Everyone agrees that speed figures are inaccurate/overvalued/useless (based on that definition).

So your theories about class are as "accurate" as everyone else's theories about ALL speed figures based on what they thought they might have read in a book 40 years ago.

[ReplayRandall]

This how the typical discussion on class goes.

1. Someone uses a definition of class they read in some handicapping book written 50-60 years ago.

2. Everyone agrees that class does not exist (based on that definition).

The problem with that discussion is that assumes the definition from 50-60 years ago was the right one. IMO that is not the case. It also assumes that looking at the quality of competition a horse has been competing against while earning his figures or compared to other contenders does not matter at all. That is demonstrably false.

Some kid is scoring 25 points per 36 minutes competing in the NCAA. Will he be able to do that in his first year in the NBA?

Golden State lost to San Antonio by a few points and Knicks beat The Lakers. Who is the class, the Knicks because they won or Golden State even though they lost?

Two teams had the same record and point differential against the same competition in the regular season. They are about to meet in the playoffs with the pressure ramped up. One has playoff tested players that have won multiple championships and the other has several key players that have never been to the playoffs.

At even money who do you want?

Class and speed are just different paths to the same answer. You look at both, try to get better at both, and go with whatever gets you the best results in different circumstances, including both.

Good post, even though I tossed class and speed figures long ago.....

[classhandicapper]

Funny you say that, because this thread was supposed to be about speed figures. And what actually happened was:

1. Someone uses a definition of how to make figures they read in some handicapping book written 40 years ago.

2. Everyone agrees that speed figures are inaccurate/overvalued/useless (based on that definition).

So your theories about class are as "accurate" as everyone else's theories about ALL speed figures based on what they thought they might have read in a book 40 years ago.

:lol:

Good one.

The understanding of figure making is evolving over time. CJ has done some excellent work, Beyer is still tinkering and improving his numbers etc... I think the OP should read the literature out there, try to get up to speed, and then go from there.

It's all good info.

I am not using figures much these days because that's what everyone else is doing. I am searching for higher ROI alternatives.

[Tom]

I find the use of performance evaluation - Quinn's condition book - and speed/pace figure mesh very nicely. Subjective and objective at the same time. Of course, you have to update the conditions a bit from the book, but really, not a lot - the spirit of the book is still functional.

[EMD4ME]

I’m sometimes curiouslly intrigued with threads like this only because many years ago I used to do the same thing. Although CJ is correct to some degree when he stated: But I would think that certain things like knowing the local horse population at your favorite track really haven’t changed all that much. So looking at conditions within a specific population can reduce the sample size significantly. So I completely agree with:
I found that the one thing that really firmed up those PAR values for adjusting my figures was to average the PAR values for all the races at same distance (and track surface). My figures were so accurate that I was able to predict the running times of races the day after the PARS were created. (assuming there weren’t any changes to the track due to the weather). But as much fun as that all was I found a much more powerful method for predicting the future.


Well you're really going to be SHOCKED to find out that there are some like myself who only look at ONE FACTOR. And I don’t see it as part of “everything” listed in your comment! As far as I’m concerned you’ve left out the most significant aspect of playing the horses in real time. You can view and subjectively analyze all of the past history on these horses and yet the most critical factor is missing from the equation: Current Condition on Race Day.

The REAL horse people know a bit more then that! Just curious Al. If you’re in awe about the “Insiders” how would feel about some Outsiders who are in touch with Insider intentions in real time?
.
.
.

Hey Nitro, look at my original quote please. It ended with "etc.". Of course I look at current form.

I watch the horses walk from the stable to the paddock. I watch them warm up. And I look at the board for hot money.

It's all part of "etc"....

[castaway01]

:lol:

Good one.

The understanding of figure making is evolving over time. CJ has done some excellent work, Beyer is still tinkering and improving his numbers etc... I think the OP should read the literature out there, try to get up to speed, and then go from there.

It's all good info.

I am not using figures much these days because that's what everyone else is doing. I am searching for higher ROI alternatives.

I knew what you meant. It's tough when some in the thread are at the point of trying to understand how figures are made, a bunch of people are confused about how they're made, and CJ has advanced to a whole other level of how they're made.

You're better off doing whatever it is you do if it works for you.

[ultracapper]

The day the numbers start running around on a piece of paper and you can bet on the one that gets there first ...........................

[cj]

No...but the TRAINER does. And if the trainer thinks that his horse is outclassed, then the horse ain't winning the race...no matter what the HORSE thinks. :)

"Class" matters...even if it exists only in the trainer's mind.

Horses win races all the time that trainers don't think the horse will win. I've seen it first hand many, many times.

[cj]

The day the numbers start running around on a piece of paper and you can bet on the one that gets there first ...........................


Speed figures are certainly not going to make you a winner alone. But without them, most people will bet on a lot of slow, hopeless horses.

[Capper Al]

An example of what was meant in the OP:

Two races on the same card with the same weather and track surface. Both races, A and B, should of been run on the Paralle speedl chart for their class as 95. Race A's winner runs at 100. Race B's winner runs at 90. If only these two races (for example sake) made up the card the adjustment would be zero, +5 and -5. This is derived by the average of the difference from the race speed minus the expected speed for the class. So the speed of the class is central to the calculation of the speed figure. What's being measured is the actual speed contrasted against the expected class speed. But what if the expected class by race type is wrong? Then the speed can't be correct. And determining class is a more debatable case than speed itself. We just don't look at class as the dependant variable when discussing speed.

[EMD4ME]

An example of what was meant in the OP:

Two races on the same card with the same weather and track surface. Both races, A and B, should of been run on the Paralle speedl chart for their class as 95. Race A's winner runs at 100. Race B's winner runs at 90. If only these two races (for example sake) made up the card the adjustment would be zero, +5 and -5. This is derived by the average of the difference from the race speed minus the expected speed for the class. So the speed of the class is central to the calculation of the speed figure. What's being measured is the actual speed contrasted against the expected class speed. But what if the expected class by race type is wrong? Then the speed can't be correct. And determining class is a more debatable case than speed itself. We just don't look at class as the dependant variable when discussing speed.

What were the paces of each race?

Was it 22 45 109 and
231/5 46 2/5 110

What were the recent efforts of the horses that comprised each field? Any juiced up runners in the 1st field? Did the 1st field have a super duel which set up a perfect tripper for a peak effort?

Many variables when dissecting how a card ran and if a variant should be adjusted/split etc.

[ReplayRandall]

Speed figures are certainly not going to make you a winner alone. But without them, most people will bet on a lot of slow, hopeless horses.

Let's not get carried away, CJ. Your pace figures will do just fine......

[Capper Al]

What were the paces of each race?

Was it 22 45 109 and
231/5 46 2/5 110

What were the recent efforts of the horses that comprised each field? Any juiced up runners in the 1st field? Did the 1st field have a super duel which set up a perfect tripper for a peak effort?

Many variables when dissecting how a card ran and if a variant should be adjusted/split etc.

Pace does affect speed, but do you use a chart to transform pace to speed figures like is commonly done by race type, surface, and distance?

[EMD4ME]

Pace does affect speed, but do you use a chart to transform pace to speed figures like is commonly done by race type, surface, and distance?

With all due respect, not answering a question when my question is not answered. Despite the fact that I was simply making a point.

[steveb]

Class and speed are just different paths to the same answer. You look at both, try to get better at both, and go with whatever gets you the best results in different circumstances, including both.

that would depend wholly on who is doing it, and how.
to me class and speed are the same thing, so they are figured at the same time, via the same path.
class figures speed and speed figures class.
of course class can refer to more than one thing, like the class of the race or the horse's inherent class.

the problem with looking at things your way, is that you only know your way, or ways that are in the public domain.
then there are the other ways.
i am sure there are people doing it in ways nobody has thought of before, just the same as my way was known only to me for most of my racing life.
my way though has probably lost its value, as it is no longer known to only me.

it is the same where i read people saying speed is no longer profitable, but of course there is speed, and then there is speed.
speed(or in my way speed & class) will cover a broad spectrum, and while some ways may be overbet others vastly different, will still be going strongly.

[steveb]

No...but the TRAINER does. And if the trainer thinks that his horse is outclassed, then the horse ain't winning the race...no matter what the HORSE thinks. :)

"Class" matters...even if it exists only in the trainer's mind.

while i understand(i think!) where you are coming from, my experience is that most trainers don't have a clue about class.
over here is australia, i have lost count of the times trainers have sought me out asking advice about where they should enter their horse/s or where i think their horse would be most suited.
then they might want a critique of their efforts, invariably after the horse has won its last start, and they need indifferent opinions, not clouded by their own bias or lack of understanding of racing.
knowing horses is vastly different to knowing racing, and most of them only know the horse part.

[woodbinepmi]

Nothing like a good old posting about speed figures to get everybody's feathers in a ruffle.

[steveb]

Nothing like a good old posting about speed figures to get everybody's feathers in a ruffle.


i can't see any ruffled feathers?
all i see are conflicting opinions.
not to mention that there is no way everybody thinks of speed in the same way.
thus you can have people canning it, but are totally ignorant of what it is to other people.
i guess the majority will only know what they have read, and form their opinions from that.
i doubt that anybody would be making a good go of it, from published methods, thus they don't use it.
who wants to know what everybody else knows?

[Capper Al]

With all due respect, not answering a question when my question is not answered. Despite the fact that I was simply making a point.

NP

[Capper Al]

So since determining class first as the dependent variable when calculating speed and adjusting the daily variant as the expected performance of a class of horses(field) is a routine application, the question follows: Is what we commonly call speed actually a horse's relative class performance in a race?

[djm1959]

ive never picked a loser!!! the horses or humans involved make the mistakes not me,,, if only i was that smart!!

[barn32]

...And I look at the board for hot money. So do the fluctuations of a horses odds allow you to predict outcome?

[cj]

So since determining class first as the dependent variable when calculating speed and adjusting the daily variant as the expected performance of a class of horses(field) is a routine application, the question follows: Is what we commonly call speed actually a horse's relative class performance in a race?

Like I said, I don't even consider class in my figures. It is never a consideration. Very few do it the way you seem to think is still common practice any longer. It is hardly routine.

[Capper Al]

Like I said, I don't even consider class in my figures. It is never a consideration. Very few do it the way you seem to think is still common practice any longer. It is hardly routine.

So are you saying that when you compare a horse to itself that you are using a feet per second standard rather than a class or race type standard? A standard of some sort is needed to measure. Did Beyer also change his ways?

[CincyHorseplayer]

The way they are made might be different but CJ still makes speed figures. He has an overall number and points out that the 2 are different. Beyer says he makes speed figures that are really performance figures.

I use a lot of varying factors in my handicapping but still profile tracks via figures and ratings derived from figures. I could look back on stacks of printouts from 2015 but I know just speed by itself is around 83% on the top 2 field ratings for a sample of at least several thousand races. I still hear that it's overbet or a trend of the masses but I think that is 1990's talk. And I think all the handicapping factors are cyclical. They go into and out of use and relevance. Many are great but have few applications on a daily basis. If we were analyzing horses playing chess I think it would be safe to say we could throw out how fast they make moves. But these horses run and that can be measured and fast wins so it's kind of hard for me to grasp this factor as a tossout. Even though my focal point is betting against numbers regularly.

[raybo]

So are you saying that when you compare a horse to itself that you are using a feet per second standard rather than a class or race type standard? A standard of some sort is needed to measure. Did Beyer also change his ways?

Class is so ambiguous, it is rare that 2 individuals even have the same definition of class. So, with that said, why would anyone calculate anything, using "class" as the "standard". My experience tells me that there is no "standard class".

How fast horses run, and how long they can run fast, seems to me to be much more of a "standard" than almost anything else in horse racing. At least it can be measured. And some, including me, think that how fast horses run, and how long they can run fast, is about as good a definition of "class" as anything else out there.

[Capper Al]

To measure anything one needs a standard. Yes, class is ambiguous, but one could do the Beyer thing and say over 120 races of $10,000 claiming the average speed for 6 furlongs was xxxx. And that's a standard. And that's the whole point that many of these speed figs are based on some such standard usually a measure of some sort of class. So are our speed figs really class performance figs for the most?

[raybo]

I don't know if "our figures" are really class performance figures or not. I suppose it depends on what figures you are using, and whether they are strictly times and variant and a time to figure chart or not.

If you're using a dirt chart and a turf chart, and you are only using times and variant, then they are pure speed figures. Anything else is not a pure speed figure. "Class" has nothing to do with pure speed figures, IMO. They only measure the time run and the variant, they measure "speed" not "class". How the speed was earned is the next step in the handicapping process, IMO.

[cj]

So are you saying that when you compare a horse to itself that you are using a feet per second standard rather than a class or race type standard? A standard of some sort is needed to measure. Did Beyer also change his ways?

I measure the horse's times against prior times. It doesn't matter whether you use parallel speed charts or FPS ratings or whatever, it is measuring speed. I only note the class briefly on my work sheets, like CLM or ALW or MCL or MSW. But that is for other reasons that have nothing to do with pars.

As best I remember Beyer has been advocating projections for a long time, not using pars. I'm certain Thorograph doesn't use pars either.

[Capper Al]

Aren't projections just saying this field of grade 3 (or whatever race type) should run 6 furlongs on the turf at xxxx?

[cj]

Aren't projections just saying this field of grade 3 (or whatever race type) should run 6 furlongs on the turf at xxxx?

No, that is using pars. Projecting is saying based on what horse X has run in the past, this is what he might be expected to run today.

I'll add an example of what I do and that is it on projections from me.

[cj]

Here is a snippet of how I do it below. This is just one race and obviously I combine the races to get an overall variant.

The first info is just race info, and you see I do note it is a claiming race (CLM). I also note Age, Sex, Distance, Surface, Run Up, rail position, etc.

I then have the top 4 in each race and the raw figures at the first two calls and the finish and my combined TimeformUS Speed Figure (denoted as PF, pace and speed). I say raw but in reality the pace variants are already calculated when the worksheet is generated. I only need to make the overall final time variant from here.

I then list a projection for each horse from the four listed past races. This is done by the program I wrote. I can and often do select a different one manually, but this is a good starting point. These are not just the speed figures the horse ran in the last four. They are adjusted for any weight changes and there are some other factors built in like maturation. I'm not going to give them all away. The key here is that it is VERY easy to underestimate projections, so I've built in some safeguards to try to avoid figure shrinkage. At least one well known figure maker, Beyer, has suffered that for years now IMO.

The last four numbers inside []s are a score for reliability. Things like distance, surface, finish position, track condition, days prior to today's race, number of races prior to today's, etc. The last sentence simply means that a more recent race is factored more heavily than an older one. All races that score below .70 are tossed from consideration. I can, however, manually use one if I feel it is warranted.

So in this case, the 4 won the race and was given a raw TimeformUS Speed Figure of 111. I projected to a 79 so the variant for this horse is a 32. The runner up ran a raw 109 and was also projected to a 79, so a 30 is projected. I do this for the top 4 horses. I can add more if needed but that is rarely the case.

Finally, I calculate a variant for the entire race based on the top four. It is the number on the far right, 33. It weights the top four finishers with the winner more heavily than second and so on.

I hope that helps. As you can see at no point is class ever factored in. I just use the projected variants for all the horses that score high enough on my reliability rating and try to figure out the track speed for the day.

(The race below is the 1st from TuP on 3-22)

[EMD4ME]

So do the fluctuations of a horses odds allow you to predict outcome?

As a NYRA player, there is rarely an angle, trip note, trainer stat that is NOT overbet.

With that said, there are many races that have either early money, non public late money, consistent money that is there for NO visible reason. In those cases, the money is hot money and does very well.

I have seen DD's that should pay $200, pay $70 (and of course come in). It's not just win money that I am referring to.

To answer your question, yes.

[EMD4ME]

Here is a snippet of how I do it below. This is just one race and obviously I combine the races to get an overall variant.

The first info is just race info, and you see I do note it is a claiming race (CLM). I also note Age, Sex, Distance, Surface, Run Up, rail position, etc.

I then have the top 4 in each race and the raw figures at the first two calls and the finish and my combined TimeformUS Speed Figure (denoted as PF, pace and speed). I say raw but in reality the pace variants are already calculated when the worksheet is generated. I only need to make the overall final time variant from here.

I then list a projection for each horse from the four listed past races. This is done by the program I wrote. I can and often do select a different one manually, but this is a good starting point. These are not just the speed figures the horse ran in the last four. They are adjusted for any weight changes and there are some other factors built in like maturation. I'm not going to give them all away. The key here is that it is VERY easy to underestimate projections, so I've built in some safeguards to try to avoid figure shrinkage. At least one well known figure maker, Beyer, has suffered that for years now IMO.

The last four numbers inside []s are a score for reliability. Things like distance, surface, finish position, track condition, days prior to today's race, number of races prior to today's, etc. The last sentence simply means that a more recent race is factored more heavily than an older one. All races that score below .70 are tossed from consideration. I can, however, manually use one if I feel it is warranted.

So in this case, the 4 won the race and was given a raw TimeformUS Speed Figure of 111. I projected to a 79 so the variant for this horse is a 32. The runner up ran a raw 109 and was also projected to a 79, so a 30 is projected. I do this for the top 4 horses. I can add more if needed but that is rarely the case.

Finally, I calculate a variant for the entire race based on the top four. It is the number on the far right, 33. It weights the top four finishers with the winner more heavily than second and so on.

I hope that helps. As you can see at no point is class ever factored in. I just use the projected variants for all the horses that score high enough on my reliability rating and try to figure out the track speed for the day.

(The race below is the 1st from TuP on 3-22)

CJ, I say this as a neutral party, your attention to detail and thought process is fantastic. Not that you need my brownie points. :lol:

[cj]

CJ, I say this as a neutral party, your attention to detail and thought process is fantastic. Not that you need my brownie points. :lol:

Took a long time and plenty money lost to get there. Thanks!

[Dave Schwartz]

CJ,

Now, that is what I love about this board: Even advanced guys find people here who can teach us things.

That was a great post, and worth a tweet!

:ThmbUp: :ThmbUp:

[steveb]

Here is a snippet of how I do it below. This is just one race and obviously I combine the races to get an overall variant.

The first info is just race info, and you see I do note it is a claiming race (CLM). I also note Age, Sex, Distance, Surface, Run Up, rail position, etc.

I then have the top 4 in each race and the raw figures at the first two calls and the finish and my combined TimeformUS Speed Figure (denoted as PF, pace and speed). I say raw but in reality the pace variants are already calculated when the worksheet is generated. I only need to make the overall final time variant from here.

I then list a projection for each horse from the four listed past races. This is done by the program I wrote. I can and often do select a different one manually, but this is a good starting point. These are not just the speed figures the horse ran in the last four. They are adjusted for any weight changes and there are some other factors built in like maturation. I'm not going to give them all away. The key here is that it is VERY easy to underestimate projections, so I've built in some safeguards to try to avoid figure shrinkage. At least one well known figure maker, Beyer, has suffered that for years now IMO.

The last four numbers inside []s are a score for reliability. Things like distance, surface, finish position, track condition, days prior to today's race, number of races prior to today's, etc. The last sentence simply means that a more recent race is factored more heavily than an older one. All races that score below .70 are tossed from consideration. I can, however, manually use one if I feel it is warranted.

So in this case, the 4 won the race and was given a raw TimeformUS Speed Figure of 111. I projected to a 79 so the variant for this horse is a 32. The runner up ran a raw 109 and was also projected to a 79, so a 30 is projected. I do this for the top 4 horses. I can add more if needed but that is rarely the case.

Finally, I calculate a variant for the entire race based on the top four. It is the number on the far right, 33. It weights the top four finishers with the winner more heavily than second and so on.

I hope that helps. As you can see at no point is class ever factored in. I just use the projected variants for all the horses that score high enough on my reliability rating and try to figure out the track speed for the day.

(The race below is the 1st from TuP on 3-22)

interesting.
if there are 10 races on the card i would be very surprised if your projections for each race were similar.
what happens then?

[cj]

interesting.
if there are 10 races on the card i would be very surprised if your projections for each race were similar.
what happens then?


Sometimes yes, sometimes no. I could write a book on all the things I see. But, and I hope you understand, I've gone about as far as I'm comfortable with sharing. It should at the very least give people something new to think about.

[Cratos]

Aren't projections just saying this field of grade 3 (or whatever race type) should run 6 furlongs on the turf at xxxx?
The question which has yet to be answered is whether this discussion is about projections or estimates and if it is about projections then are the statistical characteristics voluntary or nonvoluntary?

Finally, if anyone asserts that they don't use "pars"; my question then is describe this unique distribution.

Remember that estimates and projections are similar.

[steveb]

Sometimes yes, sometimes no. I could write a book on all the things I see. But, and I hope you understand, I've gone about as far as I'm comfortable with sharing. It should at the very least give people something new to think about.

i can understand that you don't want to go further, but at the same time, i have so many things going through my head( i guess that's the "something new to think about" bit!)

i would not call it new, different maybe, but my first thoughts....are they really speed figures at all?
that is a different subject to if they are good or not, as only you and your clients would know that, and of no interest to me as such.

if you are projecting like that for an entire card, then i am sure you are going to mostly get races that don't fit, unless you are making the adjustments on a race to race basis.

as for the book, i have been asked more than a few times, when i am going to write mine.
the answer is still...never.

[jasperson]

Let's as stated by timeform the final time figure is adjusted,then final speed figure is adjusted this is just too much adjustment for me.

2) Final Time Figures: Final-time figures made by hand and made in a pure manner that keeps them uninfected by pace adjustments.
By combining a horse’s final-time figure, pace figures, and running style into one number, we form what we believe is a state-of-the-art single-number measure of all-around performance.

[Tom]

The question which has yet to be answered is whether this discussion is about projections or estimates and if it is about projections then are the statistical characteristics voluntary or nonvoluntary?

Finally, if anyone asserts that they don't use "pars"; my question then is describe this unique distribution.

Remember that estimates and projections are similar.

Those who are using CJ's figs and making money every week really don't have time to speculate on theory. It would cost us money to divert our time.

[castaway01]

Thanks for the post CJ. Interesting to learn more about the process you go through.

[cj]

Let's as stated by timeform the final time figure is adjusted,then final speed figure is adjusted this is just too much adjustment for me.

2) Final Time Figures: Final-time figures made by hand and made in a pure manner that keeps them uninfected by pace adjustments.
By combining a horse’s final-time figure, pace figures, and running style into one number, we form what we believe is a state-of-the-art single-number measure of all-around performance.

The final time figures, not adjusted by pace, are available as well in the running lines. If you don't want the ones that include pace, you can still see the final time numbers.

[cj]

i can understand that you don't want to go further, but at the same time, i have so many things going through my head( i guess that's the "something new to think about" bit!)

i would not call it new, different maybe, but my first thoughts....are they really speed figures at all?
that is a different subject to if they are good or not, as only you and your clients would know that, and of no interest to me as such.

if you are projecting like that for an entire card, then i am sure you are going to mostly get races that don't fit, unless you are making the adjustments on a race to race basis.

as for the book, i have been asked more than a few times, when i am going to write mine.
the answer is still...never.

To be clear, I don't really care what people call them. My boss calls them TimeformUS Speed Figures and clearly lays out what they are on our blog. If people want to call them something differen't, does it matter?

I can say with certainly I don't get "mostly races that don't fit". Having to go race by race is the exception, not the rule. Often when that happens the races fit when broken down by sprint / route, or even by specific distance. But in the instances you mention where races don't fit, that is when we flag the race for ourselves and for customers.

[highnote]

I think by far the best way to make variants is with projections. You look at each and every horse that ran on a card. Instead of having between 1 and 12 data points (based on surface, inner/outer, etc.), you have anywhere from 10 to 100 or more.

I wrote speed figure software for a client about 15 years ago that used a method of looking at all the horses on the card in order to make a projected variant. I think it was limited to the fastest (i.e., most competitive) horses on the card. If a horse was far behind the top finishers the time of that slow horse was not used.

I seem to recall that the differences from some sort of a par time were determined. Then by looking at all the differences a variant could be determined.

It is a good method for turf races given the small number of races on a card.

If I have time I'll look at the code and try to remember how the system worked and I'll share it here. Maybe it will spark some innovation?

[thaskalos]

To measure anything one needs a standard. Yes, class is ambiguous, but one could do the Beyer thing and say over 120 races of $10,000 claiming the average speed for 6 furlongs was xxxx. And that's a standard. And that's the whole point that many of these speed figs are based on some such standard usually a measure of some sort of class. So are our speed figs really class performance figs for the most?
As long as these speed figures keep assigning high ratings to outclassed horses...you cannot call them "class performance figures", IMO. They may use a class measure as a starting point...but the subsequent figure-making process leads someplace else.

[steveb]

To be clear, I don't really care what people call them. My boss calls them TimeformUS Speed Figures and clearly lays out what they are on our blog. If people want to call them something differen't, does it matter?

I can say with certainly I don't get "mostly races that don't fit". Having to go race by race is the exception, not the rule. Often when that happens the races fit when broken down by sprint / route, or even by specific distance. But in the instances you mention where races don't fit, that is when we flag the race for ourselves and for customers.

no it does not matter what they are called.
the only thing that matters is if they make people money or not.

if i had a fair sized card of races(same surface) anywhere, there is no way known they would all fit perfectly when i am trying to figure(automatically) the speeds.
i guess too 'fit' would be in the eye of the beholder.
my measure of reasonably good fit would be somewhere less than maybe 5 speed points standard deviation between expected and actual race speeds, for all races on card, which probably happens no more than 60% of the time.
but for the raw figures i would never massage any number no matter how far from expectation, unless it was weather related.

i learnt a long time ago that adjusting for slow pace races for instance, gave the donkeys too high numbers relative to the good horses.
i would rather the good horses have slow base numbers in cases like that.

pace and all the other things(i was aware of) would be accounted for with other factors.

but don't think i am having a go, because i'm not.
i just find in interesting, even if i have no interest in betting, i still am as far as method goes.

[jasperson]

The final time figures, not adjusted by pace, are available as well in the running lines. If you don't want the ones that include pace, you can still see the final time numbers.
My objection with the final time numbers is they are man made. If a human has anything to do with the number mistakes can and will be made. I like all my factors computer generated and untouched by human hands. I can live with some inaccuracy in speed figure because I don't use just one speed rating. I look at the last 4 speed figures for each horse and can usually tell the horses that have shown enough speed to win this race. Then I start to handicap the other factors like pace,class,distance and form.

[cj]

My objection with the final time numbers is they are man made. If a human has anything to do with the number mistakes can and will be made. I like all my factors computer generated and untouched by human hands. I can live with some inaccuracy in speed figure because I don't use just one speed rating. I look at the last 4 speed figures for each horse and can usually tell the horses that have shown enough speed to win this race. Then I start to handicap the other factors like pace,class,distance and form.

Well that is different than your first post. I would never trust a computer with horse racing without human intervention. That is what you get with Equibase and they have some pretty absurd ratings. But that is just me. We all do things differently which is what makes the game great. Good luck.

[cj]

but don't think i am having a go, because i'm not.
i just find in interesting, even if i have no interest in betting, i still am as far as method goes.

Not at all, I respect your opinion.

[Capper Al]

Thanks CJ for your post #110. It appears to me that you are tweaking TimeFormUS numbers and, from the testimonies from some of your clients, successfully. Yet TimeFromUS's documentation states that it uses a parallel charts. There's a good chance that these charts use race type by distance/surface as the basis of their numbers. This would mean class is the base element in your formula before you start tweaking.

[raybo]

Thanks CJ for your post #110. It appears to me that you are tweaking TimeFormUS numbers and, from the testimonies from some of your clients, successfully. Yet TimeFromUS's documentation states that it uses a parallel charts. There's a good chance that these charts use race type by distance/surface as the basis of their numbers. This would mean class is the base element in your formula before you start tweaking.

Al, hasn't he already told you that his "speed" figures don't include class? His "performance" figures might, but not his "speed" figures. And, how could a parallel chart be based on race type or distance, unless there is more than one dirt chart and one turf chart? It's just a list of possible times that correspond to a list of consecutive speed numbers isn't it?

[cj]

Thanks CJ for your post #110. It appears to me that you are tweaking TimeFormUS numbers and, from the testimonies from some of your clients, successfully. Yet TimeFromUS's documentation states that it uses a parallel charts. There's a good chance that these charts use race type by distance/surface as the basis of their numbers. This would mean class is the base element in your formula before you start tweaking.

You clearly don't understand so I'm obviously not conveying it well. I also am a bit insulted you are basically calling me a liar. I'm not sure how else I can say things so I'm checking out.

EDIT: I'll revise that. Maybe you aren't calling me a liar. Maybe you think more about how the figures are made than I do. That is absurd, but not insulting.

[cj]

Al, hasn't he already told you that his "speed" figures don't include class? His "performance" figures might, but not his "speed" figures. And, how could a parallel chart be based on race type or distance, unless there is more than one dirt chart and one turf chart? It's just a list of possible times that correspond to a list of consecutive speed numbers isn't it?

This said it better than I could have. My charts are actually based on world records which have nothing to do with class.

Also, the performance figures are only final time figures modified by pace and to a much smaller extent weight. That is it. Again, no class.

[Capper Al]

You clearly don't understand so I'm obviously not conveying it well. I also am a bit insulted you are basically calling me a liar. I'm not sure how else I can say things so I'm checking out.

EDIT: I'll revise that. Maybe you aren't calling me a liar. Maybe you think more about how the figures are made than I do. That is absurd, but not insulting.

No way am I calling you a liar. I'm questioning the use of parallel charts.

[cj]

No way am I calling you a liar. I'm questioning the use of parallel charts.

Parallel charts have nothing to do with class, at least the ones I made for TimeformUS. How could you ever make speed figures without a parallel time chart to compare different distances?

[Capper Al]

This said it better than I could have. My charts are actually based on world records which have nothing to do with class.

Also, the performance figures are only final time figures modified by pace and to a much smaller extent weight. That is it. Again, no class.

Got it. Your parallel charts are world records and that becomes your standard. We can't compare anything without a standard to compare to.

[highnote]

My charts are actually based on world records ...


I've toyed with using world records as a baseline. If you do regression using current world records and distance, the way Charles Carroll did in his book, you still find the correlation is still about 99%.

I've done similar regressions using every distance and track record for a given track and found that even for a given track the correlation is still very high, but I also found interesting differences between individual tracks and distances -- which probably has to do with the runups, the sharpness of the turns, and the degree of banking.

Nick Mordin, in his book "Mordin On Time" laid out his system of using Standard Times as a starting point in making figures. His Standard Times allowed for a comparison to be made at every track, distance and surface. World Records would be a good place to start because world records are often set by Grade 1 horses. However, tracks that run races for low priced stock may have never had a Grade 1 horse race over it's surface, but it might be a very fast surface so the cheaper stock are actually able to run close to world record times even at route distances, whereas very few Grade 1 horses can get close to Secretariat's 12 furlong record. (That's an extreme example, but illustrates the point.) So when those horses ship to a track where they face higher priced stock, their speed figures from the lower level track might be inflated. So you have to account for those differences.

Nick solved that problem by using unique standard times for each track, surface and distance. The standards he created were meant to represent what a hypothetical Grade 1 horse could run over a certain track. It required trial and error and a lot of grunt work to come up with a set of standards, but once he got the relative standard times aligned the figures he produced were very usable. He was successful enough to be able to write about the experience in a book called "Betting For A Living". It's hard to find a copy nowadays, but it is worth a read.

Just did a search and found out it's in paperback!

Great Story!

http://www.amazon.com/Betting-Living-Nick-Mordin/dp/1904328083

[steveb]

I've toyed with using world records as a baseline. If you do regression using current world records and distance, the way Charles Carroll did in his book, you still find the correlation is still about 99%.

I've done similar regressions using every distance and track record for a given track and found that even for a given track the correlation is still very high, but I also found interesting differences between individual tracks and distances -- which probably has to do with the runups, the sharpness of the turns, and the degree of banking.

Nick Mordin, in his book "Mordin On Time" laid out his system of using Standard Times as a starting point in making figures. His Standard Times allowed for a comparison to be made at every track, distance and surface. World Records would be a good place to start because world records are often set by Grade 1 horses. However, tracks that run races for low priced stock may have never had a Grade 1 horse race over it's surface, but it might be a very fast surface so the cheaper stock are actually able to run close to world record times even at route distances, whereas very few Grade 1 horses can get close to Secretariat's 12 furlong record. (That's an extreme example, but illustrates the point.) So when those horses ship to a track where they face higher priced stock, their speed figures from the lower level track might be inflated. So you have to account for those differences.

Nick solved that problem by using unique standard times for each track, surface and distance. The standards he created were meant to represent what a hypothetical Grade 1 horse could run over a certain track. It required trial and error and a lot of grunt work to come up with a set of standards, but once he got the relative standard times aligned the figures he produced were very usable. He was successful enough to be able to write about the experience in a book called "Betting For A Living". It's hard to find a copy nowadays, but it is worth a read.

Just did a search and found out it's in paperback!

Great Story!

http://www.amazon.com/Betting-Living-Nick-Mordin/dp/1904328083

using regression may be fine in your country, but what about when the track topography is all over the joint, or the starting points are from chutes.
i was just going to tell you that you would be wasting your time, you need data and more data, but beforehand i thought i best check my standards for happy valley and shatin, that are figured by interation.
near perfect correlation it turns out surprisingly, but i still would not think it(regression) was the way to go.

[highnote]

It is not surprising that you got a high correlation coefficient. That is consistent with what Charles Carroll found with world record times at various distances.

It also seems to apply for track record times at U.S. racetracks, but with some interesting anomalies. Those anomalies must be understood if one is to make good speed ratings from par times or standard times.

I tried using regression to come up with standard times. It's starts to get complex because there are so many different tracks, distances and surfaces. The correlation coefficient might be nearly perfect at one track for a particular surface, but not perfect at another.

The presenter in this talk did a good job explaining some ways to deal with this challenge using TDA:

https://www.youtube.com/watch?v=F2t_ytTLrQ4

I've always thought that you need several "local" models to model horse racing outcomes in the U.S., rather than one "global" model that will work on every track, distance, class and surface.

The question is, how many local models do you need? Is there a way to cluster the data from all the tracks in a country and have the data displayed in such a way that the shape of the data can tell you where the clusters are that you should create the local models.

It looks like using TDA might be a way to find those localized clusters.

[steveb]

It is not surprising that you got a high correlation coefficient. That is consistent with what Charles Carroll found with world record times at various distances.

It also seems to apply for track record times at U.S. racetracks, but with some interesting anomalies. Those anomalies must be understood if one is to make good speed ratings from par times or standard times.

I tried using regression to come up with standard times. It's starts to get complex because there are so many different tracks, distances and surfaces. The correlation coefficient might be nearly perfect at one track for a particular surface, but not perfect at another.

The presenter in this talk did a good job explaining some ways to deal with this challenge using TDA:

https://www.youtube.com/watch?v=F2t_ytTLrQ4

I've always thought that you need several "local" models to model horse racing outcomes in the U.S., rather than one "global" model that will work on every track, distance, class and surface.

The question is, how many local models do you need? Is there a way to cluster the data from all the tracks in a country and have the data displayed in such a way that the shape of the data can tell you where the clusters are that you should create the local models.

It looks like using TDA might be a way to find those localized clusters.

you basically just need a way of lining up all jurisdiction strengths, which i am confident i know how to do.
but every track and distance in all the countries/states/localities i have done, all have their own set of times, directly comparable, and constantly updated from the beginning as new meetings are added to the data.

finding standards and classes is very easy(well it is to me anyway), and regression plays no part in that aspect of it.
you give me the data in the correct format, and i could give you standards(and class figures) for track & distance in a matter of minutes, for however many tracks you desired.
furthermore i would be confident that they would be more accurate than you would get any place else.

i would give you one clue.
most people when they analyse data trying to find time(and maybe class), will just lump it all together.
the secret is to look at it on a meeting by meeting basis, rather than lumped together.
the second clue is that you need to stop thinking in ways that everybody else does.
forget your mordins and carrolls and quirins and all those guys.
you need uniqueness.

as far as a global model is concerned, that is nonsensical, it would not work.
and if you did want perfect correlation you would not need to know anything about any particular track or distance, you could just have a formula that would do the same thing, and probably be more accurate!
for instance.....
75.13 * (thisdistance/1300) ^ 1.095
....would give you perfectly correlated times for any distance you required.

that particular formula i use myself as a starting point, and there are probably better starting points(as in more accurate), but i learned long ago that no matter what numbers(nor how bad) i use to start, they always end up relatively exactly the same.
and i only care about relationships, the actual times don't mean diddly squat, all that matters is how they relate to each other.

[highnote]

you basically just need a way of lining up all jurisdiction strengths, which i am confident i know how to do.
but every track and distance in all the countries/states/localities i have done, all have their own set of times, directly comparable, and constantly updated from the beginning as new meetings are added to the data.

finding standards and classes is very easy(well it is to me anyway), and regression plays no part in that aspect of it.
you give me the data in the correct format, and i could give you standards(and class figures) for track & distance in a matter of minutes, for however many tracks you desired.
furthermore i would be confident that they would be more accurate than you would get any place else.

i would give you one clue.
most people when they analyse data trying to find time(and maybe class), will just lump it all together.
the secret is to look at it on a meeting by meeting basis, rather than lumped together.
the second clue is that you need to stop thinking in ways that everybody else does.
forget your mordins and carrolls and quirins and all those guys.
you need uniqueness.

as far as a global model is concerned, that is nonsensical, it would not work.
and if you did want perfect correlation you would not need to know anything about any particular track or distance, you could just have a formula that would do the same thing, and probably be more accurate!
for instance.....
75.13 * (thisdistance/1300) ^ 1.095
....would give you perfectly correlated times for any distance you required.

that particular formula i use myself as a starting point, and there are probably better starting points(as in more accurate), but i learned long ago that no matter what numbers(nor how bad) i use to start, they always end up relatively exactly the same.
and i only care about relationships, the actual times don't mean diddly squat, all that matters is how they relate to each other.


Interesting that you mention relationships. Those videos on Topographical Data Analysis are all about finding relationships and then examining the clusters of data that form the relationships.

As far as a global model -- what I meant is a model for one country. I wouldn't expect a U.S. model to work at British racetracks.

Do syndicates betting Hong Kong races create one model that works at both tracks, or do they create more than one?

I ask because Benter made it seem like he only used one model.

[steveb]

Interesting that you mention relationships. Those videos on Topographical Data Analysis are all about finding relationships and then examining the clusters of data that form the relationships.

As far as a global model -- what I meant is a model for one country. I wouldn't expect a U.S. model to work at British racetracks.

Do syndicates betting Hong Kong races create one model that works at both tracks, or do they create more than one?

I ask because Benter made it seem like he only used one model.


you might do best to ask your friend nick mordin as far as hong kong goes.
i have seen emails between him and ciara(i think that was her name, and crikey, that should set the cat among the pigeons!) from woods and associates, so he would maybe know more than me about that.
i have long ago left that world, not that i ever really belonged in it.

i do know one guy that was the chief modellor at woods(has left now), and he was using basically the same model for south africa as they were for hong kong.
i got told, but don't know if it is true or not, that they were even using a generic model for greyhounds that was not far removed from the above models.

[highnote]

you might do best to ask your friend nick mordin as far as hong kong goes.
i have seen emails between him and ciara(i think that was her name, and crikey, that should set the cat among the pigeons!) from woods and associates, so he would maybe know more than me about that.
i have long ago left that world, not that i ever really belonged in it.

i do know one guy that was the chief modellor at woods(has left now), and he was using basically the same model for south africa as they were for hong kong.
i got told, but don't know if it is true or not, that they were even using a generic model for greyhounds that was not far removed from the above models.

I haven't talked to Nick in at least 3 years. I wouldn't even know how to reach him. He's a journalist and free-spirit and has lived all over the world. Who knows where he has landed?

I've heard Benter uses one model for all of the U.S., but it's hard to believe one model would fit all.

In the Asch and Quandt book, "Professors Guide to Racetrack Betting", their logit models had different coefficents for the same variables that were used for modeling the two or three tracks in their analysis. The tracks were in New Jersey. Because the coefficients were different at each track they recommended that anyone who wanted to make a similar model should collect data from each different track and make separate models.

I suppose the only way to know for sure is to do the analysis.

But we're straying from the topic of speed figures.

When I made my latest attempt at making speed figures, I ultimately decided to collect the class, distance, surface, and sectional and final times for each track along with the sample size for each category.

In order to boost the sample size I adjusted the final times of females and younger horses by using the Beyer or Quirin recommendations and the same with maiden claimers.

The thing I found interesting is that some lower level horses tend to average faster running times than higher level horses.

I can think of a few reasons why this is the case, but have never made a serious attempt to understand it fully. But it is a challenge because if those lower level fast horses step up in class they are not likely to beat the higher level horses who have averaged slower final times. It's an interesting phenomena.

[steveb]

I haven't talked to Nick in at least 3 years. I wouldn't even know how to reach him. He's a journalist and free-spirit and has lived all over the world. Who knows where he has landed?

I've heard Benter uses one model for all of the U.S., but it's hard to believe one model would fit all.

In the Asch and Quandt book, "Professors Guide to Racetrack Betting", their logit models had different coefficents for the same variables that were used for modeling the two or three tracks in their analysis. The tracks were in New Jersey. Because the coefficients were different at each track they recommended that anyone who wanted to make a similar model should collect data from each different track and make separate models.

I suppose the only way to know for sure is to do the analysis.

But we're straying from the topic of speed figures.

When I made my latest attempt at making speed figures, I ultimately decided to collect the class, distance, surface, and sectional and final times for each track along with the sample size for each category.

In order to boost the sample size I adjusted the final times of females and younger horses by using the Beyer or Quirin recommendations and the same with maiden claimers.

The thing I found interesting is that some lower level horses tend to average faster running times than higher level horses.

I can think of a few reasons why this is the case, but have never made a serious attempt to understand it fully. But it is a challenge because if those lower level fast horses step up in class they are not likely to beat the higher level horses who have averaged slower final times. It's an interesting phenomena.

i have no idea of anything about benter, other than what i have read in the public domain.
but i can't see why the same model can't be used for everything.
just because the factors are the same, does not mean that there isn't different coefficients and whatnot.
although i am not proficient at that type of thing, so am probably out of my depth there.

as for the low levels averaging faster times than higher level, then i would doubt any methodology that came to that conclusion.
it could simply be the way you are measuring?
not to mention that not all horses racing at low levels are actually low level horses, they need to start somewhere.
there are odd instances, where this can happen, but there are logical reason for it, that i don't care to mention here.

crikey, while i am feeling generous, here is something straight from what was(and maybe still is) one of the biggest syndicates going.
it will give you an idea of how the guys that make all the money find standards/classes. of course it's not as good as my way, but at the price i am asking for it you can't complain!

[highnote]

I will need some time to study this. I'll get back to you! THANKS!

[classhandicapper]

I have an interesting story on making figures.

My best friend is a very good poker player (he gave up horse racing to play poker). During the poker craze I started playing Texas hold'em. When I spoke to him about the game he kept encouraging me to play all the other games. My feeling was that I wasn't even very good at hold'em yet and I should try to specialize and get good at that game before moving to the next. His view was that there are certain things that come up in the other games fairly often that rarely come up in hold'em (but they do come up). He felt that playing those other games would make me a better hold'em player because I would quickly gain expertise in those areas that I might not notice in a lifetime of just playing hold'em. I'm sure he is right.

For decades I've either made my own figures in NY or analyzed everyone else's. Two years ago I started betting mules on the fair circuit just for fun. So of course I decided to make my own speed figures. Within a couple of months I learned a couple of things about speed figures from mules that I never really noticed or thought through in decades of analyzing data on horses.

There's some kind of lesson in both those stories about how we should go about learning and building models for horses etc...

[Capper Al]

I've toyed with using world records as a baseline. If you do regression using current world records and distance, the way Charles Carroll did in his book, you still find the correlation is still about 99%.

I've done similar regressions using every distance and track record for a given track and found that even for a given track the correlation is still very high, but I also found interesting differences between individual tracks and distances -- which probably has to do with the runups, the sharpness of the turns, and the degree of banking.

Nick Mordin, in his book "Mordin On Time" laid out his system of using Standard Times as a starting point in making figures. His Standard Times allowed for a comparison to be made at every track, distance and surface. World Records would be a good place to start because world records are often set by Grade 1 horses. However, tracks that run races for low priced stock may have never had a Grade 1 horse race over it's surface, but it might be a very fast surface so the cheaper stock are actually able to run close to world record times even at route distances, whereas very few Grade 1 horses can get close to Secretariat's 12 furlong record. (That's an extreme example, but illustrates the point.) So when those horses ship to a track where they face higher priced stock, their speed figures from the lower level track might be inflated. So you have to account for those differences.

Nick solved that problem by using unique standard times for each track, surface and distance. The standards he created were meant to represent what a hypothetical Grade 1 horse could run over a certain track. It required trial and error and a lot of grunt work to come up with a set of standards, but once he got the relative standard times aligned the figures he produced were very usable. He was successful enough to be able to write about the experience in a book called "Betting For A Living". It's hard to find a copy nowadays, but it is worth a read.

Just did a search and found out it's in paperback!

Great Story!

http://www.amazon.com/Betting-Living-Nick-Mordin/dp/1904328083

I read the mentioned books also and found them very interesting. What Mordin ends up doing is creating a track variant (he might not call it that) to adjust for track differences. It seemed to work okay for him. But I can't remember if later he said he gave up on his speed figs or not.

[cj]

using regression may be fine in your country, but what about when the track topography is all over the joint, or the starting points are from chutes.
i was just going to tell you that you would be wasting your time, you need data and more data, but beforehand i thought i best check my standards for happy valley and shatin, that are figured by interation.
near perfect correlation it turns out surprisingly, but i still would not think it(regression) was the way to go.

For me at least, the world records and regression highnote speaks of was just a starting point. Of course there are adjustments from track to track. You have to start somewhere. My only point was that it wasn't about class which seemed to be a sticking point with the original poster.

[Capper Al]

It is not surprising that you got a high correlation coefficient. That is consistent with what Charles Carroll found with world record times at various distances.

It also seems to apply for track record times at U.S. racetracks, but with some interesting anomalies. Those anomalies must be understood if one is to make good speed ratings from par times or standard times.

I tried using regression to come up with standard times. It's starts to get complex because there are so many different tracks, distances and surfaces. The correlation coefficient might be nearly perfect at one track for a particular surface, but not perfect at another.

The presenter in this talk did a good job explaining some ways to deal with this challenge using TDA:

https://www.youtube.com/watch?v=F2t_ytTLrQ4

I've always thought that you need several "local" models to model horse racing outcomes in the U.S., rather than one "global" model that will work on every track, distance, class and surface.

The question is, how many local models do you need? Is there a way to cluster the data from all the tracks in a country and have the data displayed in such a way that the shape of the data can tell you where the clusters are that you should create the local models.

It looks like using TDA might be a way to find those localized clusters.

Enjoyed the YouTube.

[Capper Al]

Interesting that you mention relationships. Those videos on Topographical Data Analysis are all about finding relationships and then examining the clusters of data that form the relationships.

As far as a global model -- what I meant is a model for one country. I wouldn't expect a U.S. model to work at British racetracks.

Do syndicates betting Hong Kong races create one model that works at both tracks, or do they create more than one?

I ask because Benter made it seem like he only used one model.

I have been posting some speed figure comparison in my thread 'What's up with the Capper?' here in PA. My AMS and Euro speed figures are being compared to BRIS speed figures. (I'm a one man part time operation. So development is slow.) But anyway, my figures are proving to be able to compete with BRIS. I believe that I have the beginnings of cracking the global model.

What I can say is that I have found a way to interpret a global distance/surface chart. I needed to have some speed figures for Euro horse appearing in North American races.

[Capper Al]

For me at least, the world records and regression highnote speaks of was just a starting point. Of course there are adjustments from track to track. You have to start somewhere. My only point was that it wasn't about class which seemed to be a sticking point with the original poster.

Once you have a standard you have a class standard. You can't get away from it. Even if it is that the average horse runs 8 furlongs on the dirt at XXX, you know if your horse is above or below average. That's class my friend.

[cj]

Once you have a standard you have a class standard. You can't get away from it. Even if it is that the average horse runs 8 furlongs on the dirt at XXX, you know if your horse is above or below average. That's class my friend.


Well than for you speed is class. It isn't for me. Obviously you think you know more about my speed figures than I do since I've repeatedly said class isn't used. I used world records, periods. How much slower horses run than that baseline is a measure of speed, not class, my friend.

[classhandicapper]

Once you have a standard you have a class standard. You can't get away from it. Even if it is that the average horse runs 8 furlongs on the dirt at XXX, you know if your horse is above or below average. That's class my friend.

You are talking about different issue.

Most pure final time handicappers are not particularly interested in whether a race was faster or slower than average for the class. They just want to know which horse ran faster. How fast they ran in the past etc...

If you are a class handicapper, you not only have to know the pecking order of the classes, the more critical component is separating the strong races from the weak ones within a specific class (because that's tougher). When you are trying to do that, a quick glance at the speed figure for the race vs. the PAR for the class is one piece of evidence. I prefer to dig deep into the quality of the field and see what the accomplishments of the horses in the field were, what their recent form looks like, what their trips were like, etc... and compare that to the average field for that class.

I've always done that manually, but I recently created an automated way of generating standards for that too and then generating ratings (stronger than average or weaker than average) that are of good enough quality that I now know when to really dig when not to bother.

[Capper Al]

Well than for you speed is class. It isn't for me. Obviously you think you know more about my speed figures than I do since I've repeatedly said class isn't used. I used world records, periods. How much slower horses run than that baseline is a measure of speed, not class, my friend.

CJ, I let it be until you jumped on the OP. I accepted that what you meant was that you don't use any class rating factors and left it at that. It is the standard that's class, however you make it as long as it correlates.

[Capper Al]

You are talking about different issue.

Most pure final time handicappers are not particularly interested in whether a race was faster or slower than average for the class. They just want to know which horse ran faster. How fast they ran in the past etc...

If you are a class handicapper, you not only have to know the pecking order of the classes, the more critical component is separating the strong races from the weak ones within a specific class (because that's tougher). When you are trying to do that, a quick glance at the speed figure for the race vs. the PAR for the class is one piece of evidence. I prefer to dig deep into the quality of the field and see what the accomplishments of the horses in the field were, what their recent form looks like, what their trips were like, etc... and compare that to the average field for that class.

I've always done that manually, but I recently created an automated way of generating standards for that too and then generating ratings (stronger than average or weaker than average) that are of good enough quality that I now know when to really dig when not to bother.

Let's say we look at things differently.

[classhandicapper]

Let's say we look at things differently.

That's fine.

My point is that you have some kind of class standard and compare the horse's speed to that. Most final time handicappers don't care about the standards and many class handicappers don't care about the final times.

[Capper Al]

That's fine.

My point is that you have some kind of class standard and compare the horse's speed to that. Most final time handicappers don't care about the standards and many class handicappers don't care about the final times.

When you can make the statement that horse A runs 3 lengths(or any other unit measure) faster than the rest of the field, you are in essence saying that horse A outclasses the field. There is no way to measure speed without a standard and once the standard is established there is a classification.

[Capper Al]

BTW, I think Quinn published a speed and class correlation chart showing that the two correlate.

[classhandicapper]

When you can make the statement that horse A runs 3 lengths(or any other unit measure) faster than the rest of the field, you are in essence saying that horse A outclasses the field. There is no way to measure speed without a standard and once the standard is established there is a classification.

I understand where you are coming from.

[raybo]

When you can make the statement that horse A runs 3 lengths(or any other unit measure) faster than the rest of the field, you are in essence saying that horse A outclasses the field. There is no way to measure speed without a standard and once the standard is established there is a classification.

A "classification" is not the same as "class". A "classification" like you are presenting, is one dimensional (speed, or lengths, or whatever), "class" is multi-dimensional, big difference. Take that "speed" horse and stick him in a higher "class" and that "speed" will not hold up.

[Capper Al]

A "classification" is not the same as "class". A "classification" like you are presenting, is one dimensional (speed, or lengths, or whatever), "class" is multi-dimensional, big difference. Take that "speed" horse and stick him in a higher "class" and that "speed" will not hold up.

This isn't black and white. But overall I would agree with the multi-dimensional usage for class when judging class as an entity onto itself.

[raybo]

This isn't black and white. But overall I would agree with the multi-dimensional usage for class when judging class as an entity onto itself.

Well, are you talking about "class" or a "classification" like speed or time or lengths? Speed, time, and lengths, etc., do not accurately define "class". Courage, determination, intelligence, refusal to lose, style flexibility, and pace plus final run ability, more accurately define "class".

Even "class" classifications (as in man-made class) aren't terribly accurate class definitions. That's one of the reasons performance figures are created, to more accurately define the relationships between final time speed and class, etc., how the final time figure was earned.

[Capper Al]

Well, are you talking about "class" or a "classification" like speed or time or lengths? Speed, time, and lengths, etc., do not accurately define "class". Courage, determination, intelligence, refusal to lose, style flexibility, and pace plus final run ability, more accurately define "class".

Even "class" classifications (as in man-made class) aren't terribly accurate class definitions. That's one of the reasons performance figures are created, to more accurately define the relationships between final time speed and class, etc., how the final time figure was earned.

Sometimes one needs to challenge convention. I don't disagree as much with the class for class sake definitions that you give, but still stand that any horse that can out run the field by a second is more than not the class of the field as one dimensional as this is.

[cj]

CJ, I let it be until you jumped on the OP. I accepted that what you meant was that you don't use any class rating factors and left it at that. It is the standard that's class, however you make it as long as it correlates.

My only point was that it wasn't about class which seemed to be a sticking point with the original poster.

The above certainly wasn't jumping on anyone, wasn't intended that way at all. I just was saying we seemed not be communicating well on this...a sticking point.

[Capper Al]

The above certainly wasn't jumping on anyone, wasn't intended that way at all. I just was saying we seemed not be communicating well on this...a sticking point.

NP

[cj]

When you can make the statement that horse A runs 3 lengths(or any other unit measure) faster than the rest of the field, you are in essence saying that horse A outclasses the field. There is no way to measure speed without a standard and once the standard is established there is a classification.

What do three lengths (which are nothing more than a measure of time) have to do with class? Nobody is saying it means a horse outclassed the field, at least not among speed figure makers. We are saying he is faster...speed.

[cj]

NP

Cool.

[Capper Al]

What do three lengths (which are nothing more than a measure of time) have to do with class? Nobody is saying it means a horse outclassed the field, at least not among speed figure makers. We are saying he is faster...speed.

I picked a number. Let use 10 lengths better than the field to make my point. Lengths probably are converted to speed points by most cappers.

[NorCalGreg]

I picked a number. Let use 10 lengths better than the field to make my point. Lengths probably are converted to speed points by most cappers.


Al...are you secretly laughing your ass off, seeing how many times you can get CJ to respond to the same "point" :D Now you're gonna rephrase it---WITH 10 LENGTHS INSTEAD OF THREE :lol: :lol: :lol:

[Capper Al]

Al...are you secretly laughing your ass off, seeing how many times you can get CJ to respond to the same "point" :D Now you're gonna rephrase it---WITH 10 LENGTHS INSTEAD OF THREE :lol: :lol: :lol:

No. These discussions are difficult, and CJ has a lot of good input to offer. We all look at the same numbers a little differently. I believe the discussion has been very good in this thread, and we were able to bridge our misunderstandings.

[Capper Al]

Aqu speed rankings are posted in the 'What's up with the Capper' thread. Post your speed figures/selections there if you like.

Good luck!

[cj]

I picked a number. Let use 10 lengths better than the field to make my point. Lengths probably are converted to speed points by most cappers.

Any number doesn't matter to me. Lengths are time, not class. I can't say I've ever heard class defined with time.

[Capper Al]

Any number doesn't matter to me. Lengths are time, not class. I can't say I've ever heard class defined with time.

Beyer trashed the traditional idea of class in one of his books saying that class is speed. I don't go that far.

[steveb]

Any number doesn't matter to me. Lengths are time, not class. I can't say I've ever heard class defined with time.

class can have many meaning as far as racing goes.
in my way, class(one aspect of) and speed are both the same thing.
class is the centrepoint(mean usually) of the individual speeds for that grade of race.
speed is the individual efforts.

i realise it's only semantics, and thus unimportant, but i am in al's corner as far as that topic goes.

[InsideTheRaces.com]

To measure anything one needs a standard. Yes, class is ambiguous, but one could do the Beyer thing and say over 120 races of $10,000 claiming the average speed for 6 furlongs was xxxx. And that's a standard. And that's the whole point that many of these speed figs are based on some such standard usually a measure of some sort of class. So are our speed figs really class performance figs for the most?

I don't want to debate what Class means, but you can't just lump horses into groups by claiming price.
A non conditional 10,000 Claimer at Charles Town is not the same kind of horse as a non conditional 10,000 Claimer at Parx.
A non conditional 10,000 Claimer has a purse of $15,000 at Charles Town vs $30,000 at Parx. I guarantee you the 10,000 Parx horses will Smoke the 10,000 Charles Town horses. You can go even further when comparing horses at the same track. A 10,000 claimer nw3L would get Smoked by a winning non conditional 5,000 Claimer at either track. So what does a $10,000 claimer really mean?

[steveb]

I don't want to debate what Class means, but you can't just lump horses into groups by claiming price.
A non conditional 10,000 Claimer at Charles Town is not the same kind of horse as a non conditional 10,000 Claimer at Parx.
A non conditional 10,000 Claimer has a purse of $15,000 at Charles Town vs $30,000 at Parx. I guarantee you the 10,000 Parx horses will Smoke the 10,000 Charles Town horses. You can go even further when comparing horses at the same track. A 10,000 claimer nw3L would get Smoked by a winning non conditional 5,000 Claimer at either track. So what does a $10,000 claimer really mean?

i would have no idea of the relative strengths of those joints, but it would take about 2 minutes to figure it, given sufficient data, and the know how.



another way, you could run all those joints together(starting from the same baseline).

once you got the speed equivalent to class for them all as a group, then you could check each venue off against the other.

then you would have the same class figure for them all, BUT if you tallied the speeds and they were indeed different strengths, the better venues would have faster average speeds than the lesser joints.

guaranteed

[Tom]

The Beyer pars for $10K claimers varies greatly track to track, generally between 73 and 83. I looked at all $10K claimers last year.

[traynor]

i would have no idea of the relative strengths of those joints, but it would take about 2 minutes to figure it, given sufficient data, and the know how.



another way, you could run all those joints together(starting from the same baseline).

once you got the speed equivalent to class for them all as a group, then you could check each venue off against the other.

then you would have the same class figure for them all, BUT if you tallied the speeds and they were indeed different strengths, the better venues would have faster average speeds than the lesser joints.

guaranteed

The real question should be, will comparing single instances to an average or mean indicate a "deviation"--faster or slower--by that single instance? In the "serious wagering" software apps I have seen (and worked on), that is one of the first concepts abandoned as "interesting, but not especially useful."

An average (or mean), however carefully constructed, edited, cleaned, and manicured, is still only an expression "simplifying" a range of values. Specifically, any value within the range of values used to construct the average (or mean) is "valid."

Consider all times at a given grade/distance/track as input, with a sample large enough to be considered a "normal distribution." That does NOT mean that a horse finishing in a time a second less or more than the average (or mean) is "faster" or "slower" if that time falls within the normal distribution range.

That is one of the major reasons that make the use of "par times" more deceptive than predictive. Or even useful.

[Capper Al]

The trick with the use of pars is in the variant. Even Beyer figs will be readjusted afterwards.

[InsideTheRaces.com]

i would have no idea of the relative strengths of those joints, but it would take about 2 minutes to figure it, given sufficient data, and the know how.



another way, you could run all those joints together(starting from the same baseline).

once you got the speed equivalent to class for them all as a group, then you could check each venue off against the other.

then you would have the same class figure for them all, BUT if you tallied the speeds and they were indeed different strengths, the better venues would have faster average speeds than the lesser joints.

guaranteed

My point was that grouping horses by claiming price is flawed. I would think grouping horses by total purse would yield better results. I wish I had a database to validate which method produces a tighter grouping across all tracks.

[traynor]

The trick with the use of pars is in the variant. Even Beyer figs will be readjusted afterwards.

Not quite. There is a cascade effect, in which the flawed base corrupts everything that generates from it. "Variant" implies that some base exists from which another data point differs. If on a given day, in six six-furlong races, same "class level," the the times are 1:10, 1:10.1, 1:10.2, 1:10.3, 1:10.4,and 1:11 (or an even greater spread of values) what does it tell you about the "speed" of the track on that particular day? Not much. All values fall within a performance envelope for some other categorization (like "class level")--and all are "normal."

It is the conceptual error of assuming that variations of values from an average or mean are significant when they are not. Not just in horse racing.

[Capper Al]

Not quite. There is a cascade effect, in which the flawed base corrupts everything that generates from it. "Variant" implies that some base exists from which another data point differs. If on a given day, in six six-furlong races, same "class level," the the times are 1:10, 1:10.1, 1:10.2, 1:10.3, 1:10.4,and 1:11 (or an even greater spread of values) what does it tell you about the "speed" of the track on that particular day? Not much. All values fall within a performance envelope for some other categorization (like "class level")--and all are "normal."

It is the conceptual error of assuming that variations of values from an average or mean are significant when they are not. Not just in horse racing.

Not so fast. The averages or means do set a point for comparisons. One just has to figure out what happened that race day at that particular track. That's the variant and that's the hard part. Anyone can average or find the mean. And the speeds do cluster around a mid-point over the long run.

[Cratos]

Not so fast. The averages or means do set a point for comparisons. One just has to figure out what happened that race day at that particular track. That's the variant and that's the hard part. Anyone can average or find the mean. And the speeds do cluster around a mid-point over the long run.
I hope you are responding through the "lens" of a nonlinear curve and non-homogeneous data.

[raybo]

Not so fast. The averages or means do set a point for comparisons. One just has to figure out what happened that race day at that particular track. That's the variant and that's the hard part. Anyone can average or find the mean. And the speeds do cluster around a mid-point over the long run.

I think that is the point of contention between "pars" people and "projections" people. The pars people are restricted by the size of the sample, because they are looking at the winners' times to get their averages/means, while the "projections" people are looking at individual horses' past performances and projecting what they should produce today, all things being equal of course, and then combining several horses from each race to get an idea of how many underperformed or overperformed that day, at that track. That may be all hogwash, but that's how I understand it anyway.

[steveb]

The real question should be, will comparing single instances to an average or mean indicate a "deviation"--faster or slower--by that single instance? In the "serious wagering" software apps I have seen (and worked on), that is one of the first concepts abandoned as "interesting, but not especially useful."

An average (or mean), however carefully constructed, edited, cleaned, and manicured, is still only an expression "simplifying" a range of values. Specifically, any value within the range of values used to construct the average (or mean) is "valid."

Consider all times at a given grade/distance/track as input, with a sample large enough to be considered a "normal distribution." That does NOT mean that a horse finishing in a time a second less or more than the average (or mean) is "faster" or "slower" if that time falls within the normal distribution range.

That is one of the major reasons that make the use of "par times" more deceptive than predictive. Or even useful.

when one wants to find the strengths of one jurisdiction to another, single instances are worthless, so basically i am at a loss to know what you mean?
i use the normal distribution all the time, but i can't see any relevance here, when one wants to figure the different relative strengths of one place to another.
if the average of one place for the same type race was maybe 5 points (or whatever) stronger to another, then that place WILL be stronger on average and will give you the baselines needed.
that the slow place would have races run faster than the other is neither here nor there.

your last sentence had me smiling, than you very much.
if they were not predictive, then the people that make all the money would not have factors based on them i don't think.

[traynor]

Not so fast. The averages or means do set a point for comparisons. One just has to figure out what happened that race day at that particular track. That's the variant and that's the hard part. Anyone can average or find the mean. And the speeds do cluster around a mid-point over the long run.


The averages or means do NOT set a point for comparison. They MAY (help a bit to) define a normal distribution--a range of values, not a single figure. Any value that falls within the normal distribution is "normal." The error is created by defining (or interpreting) the average or mean as an "expected performance standard" that may be used for individual comparisons.

On the contrary, the variant is the easy part, if one gets past the notion of "par times" as some kind of absolute value and looks at what is actually going on.

Almost any set of (roughly) similar values will seem to cluster around a mid-point over the long run. That is the essence of a bell curve.

[traynor]

when one wants to find the strengths of one jurisdiction to another, single instances are worthless, so basically i am at a loss to know what you mean?
i use the normal distribution all the time, but i can't see any relevance here, when one wants to figure the different relative strengths of one place to another.
if the average of one place for the same type race was maybe 5 points (or whatever) stronger to another, then that place WILL be stronger on average and will give you the baselines needed.
that the slow place would have races run faster than the other is neither here nor there.

your last sentence had me smiling, than you very much.
if they were not predictive, then the people that make all the money would not have factors based on them i don't think.

I don't know about jurisdictions, but if you are referring to a track-to-track "adjustment" (as widely defined and/or used) the notion of comparing one absolute value to another absolute value and applying the difference as an "adjustment" is about as useful as Professor Gordon Jones' parallel speed charts. Or the various "equalization adjustments" made by most handicapping software.

[steveb]

I don't know about jurisdictions, but if you are referring to a track-to-track "adjustment" (as widely defined and/or used) the notion of comparing one absolute value to another absolute value and applying the difference as an "adjustment" is about as useful as Professor Gordon Jones' parallel speed charts. Or the various "equalization adjustments" made by most handicapping software.

this is the problem with forums.
i have no idea what you just said, thus it is meaningless to me.

just because what YOU know does not work(or you think it doesn't), that does not mean what others know doesn't
i always have a chuckle when people get presumptuous about what others can do.

there is only one evidence needed for the worth of any factor.
i don't think i need tell you what it is.

[classhandicapper]

If I wanted to compare horses from different circuits on a final time basis, I'd make speed figures for each circuit in isolation and then monitor every shipper in either direction until I eventually had enough of a sample size to create an adjustment to bring them into sync.

If I was more class oriented, I'd define the pecking order at each track and then monitor the class level of all the shippers.

[traynor]

this is the problem with forums.
i have no idea what you just said, thus it is meaningless to me.

just because what YOU know does not work(or you think it doesn't), that does not mean what others know doesn't
i always have a chuckle when people get presumptuous about what others can do.

there is only one evidence needed for the worth of any factor.
i don't think i need tell you what it is.

" only effects that fall much farther than two standard deviations away from what would have been expected are considered statistically significant—normal random error or variation in the measurements is in this way distinguished from causal variation. "
https://en.wikipedia.org/wiki/Standard_deviation

There is little point in calculating "variants" from data points that are little more than semantic noise.

[traynor]

If I wanted to compare horses from different circuits on a final time basis, I'd make speed figures for each circuit in isolation and then monitor every shipper in either direction until I eventually had enough of a sample size to create an adjustment to bring them into sync.

If I was more class oriented, I'd define the pecking order at each track and then monitor the class level of all the shippers.

Good strategy. Conceptually much better than using "differences between par times of equivalent class levels"--without regard for how much comfort and solace is generated by the latter in using values that seem to mean something (whether they actually do or not).

[steveb]

If I wanted to compare horses from different circuits on a final time basis, I'd make speed figures for each circuit in isolation and then monitor every shipper in either direction until I eventually had enough of a sample size to create an adjustment to bring them into sync.

If I was more class oriented, I'd define the pecking order at each track and then monitor the class level of all the shippers.

what i do is start from the same baseline, and then the differences come from what most refer to as the variant.
maybe it is simplistic and i don't know how to explain it in depth, not that i want to
does it work.....it certainly does.
and you don't need any common horses to figure it.

[steveb]

" only effects that fall much farther than two standard deviations away from what would have been expected are considered statistically significant—normal random error or variation in the measurements is in this way distinguished from causal variation. "
https://en.wikipedia.org/wiki/Standard_deviation

There is little point in calculating "variants" from data points that are little more than semantic noise.

another chuckle thank you.
what has the normal distribution got to do with figuring relative strengths of different jurisdictions?
hang on i will tell you....sweet sweet nothing
i have no idea if other people do that, but it is irrelevant my way.

anyway, thank you for telling this 64 year old, that he has been going the unacceptable since he was 25 or 26.
maybe i just dreamed my life and i will wake up shortly. :)

[Capper Al]

what i do is start from the same baseline, and then the differences come from what most refer to as the variant.
maybe it is simplistic and i don't know how to explain it in depth, not that i want to
does it work.....it certainly does.
and you don't need any common horses to figure it.

Your method makes sense to me.

[Dave Schwartz]

If I wanted to compare horses from different circuits on a final time basis, I'd make speed figures for each circuit in isolation and then monitor every shipper in either direction until I eventually had enough of a sample size to create an adjustment to bring them into sync.

Did this. it was completely uncorrelated.

[Capper Al]

what i do is start from the same baseline, and then the differences come from what most refer to as the variant.
maybe it is simplistic and i don't know how to explain it in depth, not that i want to
does it work.....it certainly does.
and you don't need any common horses to figure it.

One can't compare without a baseline. If no baseline, what are they comparing to?

[Capper Al]

I hope you are responding through the "lens" of a nonlinear curve and non-homogeneous data.

Cratos,

How are you?

[Capper Al]

I think that is the point of contention between "pars" people and "projections" people. The pars people are restricted by the size of the sample, because they are looking at the winners' times to get their averages/means, while the "projections" people are looking at individual horses' past performances and projecting what they should produce today, all things being equal of course, and then combining several horses from each race to get an idea of how many underperformed or overperformed that day, at that track. That may be all hogwash, but that's how I understand it anyway.

Helpful post. One needs a baseline to compare and a horse can't be compared to itself. The best that can be done is use speed figs to project which are themselves built from a baseline. There's no escaping it.

[Capper Al]

The averages or means do NOT set a point for comparison. They MAY (help a bit to) define a normal distribution--a range of values, not a single figure. Any value that falls within the normal distribution is "normal." The error is created by defining (or interpreting) the average or mean as an "expected performance standard" that may be used for individual comparisons.

On the contrary, the variant is the easy part, if one gets past the notion of "par times" as some kind of absolute value and looks at what is actually going on.

Almost any set of (roughly) similar values will seem to cluster around a mid-point over the long run. That is the essence of a bell curve.

The best we can do is use an average or a top or a mean as a standard. We need a starting point to measure from. On the contrary, the variant is the hard part-- explaining why the horse did what it did in today's race and quantifying it is difficult.

[whodoyoulike]

My point was that grouping horses by claiming price is flawed. I would think grouping horses by total purse would yield better results. I wish I had a database to validate which method produces a tighter grouping across all tracks.


... BUT if you tallied the speeds and they were indeed different strengths, the better venues would have faster average speeds than the lesser joints.

guaranteed


I agree with both of these comments which is the reason I've separated my handicapping to "A", "B", "C" and "D" type of tracks by keeping track of only the "A" tracks and adjusting for the others. I've found the characteristics of the horses racing at the different track ratings has been consistent for my type of handicapping.

[whodoyoulike]

The best we can do is use an average or a top or a mean as a standard. We need a starting point to measure from. On the contrary, the variant is the hard part-- explaining why the horse did what it did in today's race and quantifying it is difficult.

This is why I've concluded that you really need to evaluate a horse's ability in a range versus comparing it to a specific average number.

What has the horse performed at and what is it projected to be able to perform at in the upcoming race?

[steveb]

The best we can do is use an average or a top or a mean as a standard. We need a starting point to measure from. On the contrary, the variant is the hard part-- explaining why the horse did what it did in today's race and quantifying it is difficult.

i think of the variant as just part of the same process that gets the speeds, and in and of itself meaningless.
it's just the bit left over, after the calculations and is the difference between expected and actual.
i find the standard deviation of the differences between actual and expected more helpful, and certainly much more informative than the variant.

although in some of my methods the 'variant' is a way to measure the difference between things when everything else is equal.

[Cratos]

Cratos,

How are you?
I am doing well; thanks for asking.

However you are correct, a baseline or frame of reference is need to compare. What me and my associates do is start with a daily environmental impact curve built from 4 variables. The reason being is that whether it is the rate of motion of man, animal, or machine you absolutely must have an understanding of the resistance to motion due to the environment.

Then we assess the motion by distance with respect to the environment where the motion occurred.

What we have in the end after making some esoteric adjustments is a standardize downward sloping curve that becomes our reference curve.

The comparative output is very dynamic due to the environmental influence.

This is done with our model.

[steveb]

My point was that grouping horses by claiming price is flawed. I would think grouping horses by total purse would yield better results. I wish I had a database to validate which method produces a tighter grouping across all tracks.

my apologies, i should have responded to this.

one thing that really irritates me, is when somebody asks you a question, and then you answer them as best you can, but they don't acknowledge that you have answered the question they asked.

i know that you did not ask a question, but it still deserves a response.

if you meant grouping races (rather than horses) at claiming price, then i would say that it depends.
you don't know until you have done the calculations.
i am not familiar with racing in your part of the world, but places where they have 10 zillion tracks and racing is run far and wide, then it is usually obvious that some places will be stronger than others.

we have that situation over here the same as you, and class 'a' at a metro venue or close to a major training centre, will naturally nearly always be stronger than a more remote place.
often times a theoretical lower class at the stronger area will be in actual fact much stronger than a theoretical higher class at a far flung joint.
but no matter, you find out by doing the calculations.
the problem though is that how does one do the calculations if it is speed based?
my thoughts from reading this board is that most have no idea how to do it.

not being familiar with the prize money structure over there i can't comment knowingly.
but surely some races will have humungous prize and be won by second raters?
i won't use it myself, because over here they will often have the same prize for a race of a certain type run at woop woop, as they will for that type at a major track.
they are even weighted the same and get the same penalites it they win at the 2nd rate joint as if they did at the major joint.
too ambiguous for me.
or all the races at some tracks will be of the same prize despite hugely different class levels.

thus to me it does not make sense to do it the way you think it should be.

[highnote]

if you meant grouping races (rather than horses) at claiming price, then i would say that it depends.

....

you don't know until you have done the calculations.

thus to me it does not make sense to do it the way you think it should be.

Hi Steveb,

Below is some data from Fort Erie racetrack in Canada.

Columns are:

TrackAbb, Furlongs, Dirt, ClaimingPrice, FinalTime, 2furlongTime, 4furlongTime, 6furlongTime, SampleSize

Notice how sometimes the lower priced claiming races have faster times than the higher priced claiming races.*

Compare the average final time of the $4,000 claiming races with the average time of the $5,000 claiming races at 8.32 furlongs.

Both have a fairly large sample size, yet, the more expensive horses run slower final times.

At shorter distances, the lower priced horses run faster early fractions than the older horses, but slower final times. (Maybe less expensive horses can't carry their speed as far, even though they are quicker, and that's why they are racing at the bottom level?)

I've seen this at many lower level U.S. tracks, too.


FE 500 D 4000 58.51 22.28 45.78 0 84
FE 550 D 4000 65.09 22.44 45.87 0 55
FE 600 D 4000 71.79 22.62 46.07 0 129
FE 650 D 4000 78.68 23.08 46.32 0 125
FE 832 D 4000 103.88 0 47.28 72.9 126
FE 850 D 4000 106.34 0 47.81 73.21 72
FE 450 D 5000 52.14 22.49 46.14 0 26
FE 500 D 5000 58.43 22.36 45.87 0 215
FE 550 D 5000 65.05 22.47 45.92 0 95
FE 600 D 5000 71.69 22.57 46.02 0 287
FE 650 D 5000 78.62 23.13 46.42 0 156
FE 832 D 5000 104.05 0 47.19 73.32 235
FE 850 D 5000 106.5 0 47.84 73.63 176
FE 900 D 5000 113.46 0 48.64 74.34 17
FE 500 D 10000 58.31 22.49 45.95 0 98
FE 550 D 10000 64.68 22.43 45.75 0 30
FE 600 D 10000 71.48 22.55 45.99 0 134
FE 650 D 10000 78.22 23.23 46.38 0 61
FE 832 D 10000 103.57 0 46.9 73.08 89
FE 850 D 10000 105.76 0 47.67 73.28 63


* In order to have large sample sizes I converted all race times to older male horse times using the adjustments Beyer and Quirin suggest using.
For example, the winning time of a 3 year old filly would have X amount of time deducted from it to bring it into alignment with older male horses. It's possible this caused errors, but I have noticed the same phenomena even without these artificial adjustments.

In order to make par times, I will just use the average time of the higher level claiming times as the par time and extrapolate downwards, otherwise, the lower level horses will have faster pars than the higher level, and I don't think that would be correct.

How do deal with this?

[steveb]

Hi Steveb,

Below is some data from Fort Erie racetrack in Canada.

Columns are:

TrackAbb, Furlongs, Dirt, ClaimingPrice, FinalTime, 2furlongTime, 4furlongTime, 6furlongTime, SampleSize

Notice how sometimes the lower priced claiming races have faster times than the higher priced claiming races.*

Compare the average final time of the $4,000 claiming races with the average time of the $5,000 claiming races at 8.32 furlongs.

Both have a fairly large sample size, yet, the more expensive horses run slower final times.

At shorter distances, the lower priced horses run faster early fractions than the older horses, but slower final times. (Maybe less expensive horses can't carry their speed as far, even though they are quicker, and that's why they are racing at the bottom level?)

I've seen this at many lower level U.S. tracks, too.


FE 500 D 4000 58.51 22.28 45.78 0 84
FE 550 D 4000 65.09 22.44 45.87 0 55
FE 600 D 4000 71.79 22.62 46.07 0 129
FE 650 D 4000 78.68 23.08 46.32 0 125
FE 832 D 4000 103.88 0 47.28 72.9 126
FE 850 D 4000 106.34 0 47.81 73.21 72
FE 450 D 5000 52.14 22.49 46.14 0 26
FE 500 D 5000 58.43 22.36 45.87 0 215
FE 550 D 5000 65.05 22.47 45.92 0 95
FE 600 D 5000 71.69 22.57 46.02 0 287
FE 650 D 5000 78.62 23.13 46.42 0 156
FE 832 D 5000 104.05 0 47.19 73.32 235
FE 850 D 5000 106.5 0 47.84 73.63 176
FE 900 D 5000 113.46 0 48.64 74.34 17
FE 500 D 10000 58.31 22.49 45.95 0 98
FE 550 D 10000 64.68 22.43 45.75 0 30
FE 600 D 10000 71.48 22.55 45.99 0 134
FE 650 D 10000 78.22 23.23 46.38 0 61
FE 832 D 10000 103.57 0 46.9 73.08 89
FE 850 D 10000 105.76 0 47.67 73.28 63


* In order to have large sample sizes I converted all race times to older male horse times using the adjustments Beyer and Quirin suggest using.
For example, the winning time of a 3 year old filly would have X amount of time deducted from it to bring it into alignment with older male horses. It's possible this caused errors, but I have noticed the same phenomena even without these artificial adjustments.

In order to make par times, I will just use the average time of the higher level claiming times as the par time and extrapolate downwards, otherwise, the lower level horses will have faster pars than the higher level, and I don't think that would be correct.

How do deal with this?

i know this won't be the answer you want highnote, but it's the best i can do.

you can't tell anything in particular as far as i can tell with that info you provide.
there is nothing unusual about lower class races often running faster raw times than higher class.

but you are doing, as far as i can tell, something i would never recommend.....lumping all the data together.
it needs looking at on a meeting by meeting basis.

roughly speaking i would make EVERY race the same class to begin with.
my way that would be 100 for each and every race.
then i would make a rough set of standard times for every distance using that formula i put on this thread a few days ago.
then meeting by meeting i would iterate through all the data
after each iteration i would compare the class figures with the speed figures.
i would then make the class figures for the second iteration equal to the average speeds of the first iteration, and change the standard times in line with that.
i would just keep doing that until class and speed with each type were the same.
at the end you would know how fast each class is, and as a bonus you would have accurate standards.
where you don't have enough data, you will likely be let down, but that in itself is not a big problem.

if you have expected figures for most races on any card, and also accurate standard times, then you won't have any problems figuring how fast any particular race has been run.
because you have baselines to help you

now i know there is no normal distribution involved here, and some guys that are much more intelligent than me will tell you it's not possible.
unfortunately i never listen to the experts, and just go my own way.
probably if i did listen , i would be bankrupt now, and never have had the life i had, and left on my own terms, to do things that now interest me more.

[Capper Al]

Okay projection folks, let's see if we are on the same page. To me, speed projection is just bringing form/cycle into the figuring. I do like this(not there yet), but what is being projected are speed figures which are based on some sort of baseline or standard which in turn are categorized into a class structure. You can't escape.

[Capper Al]

Why do you think commercial speed figs like Beyer's are recalculated later? You have two choices and one doesn't count. To adjust a speed afterwards indicates that either there was a miscalculation in the parallel charts or the daily variant. I serious doubt the Beyer is adjusting his parallel charts on a daily basis for errors. It has to be the daily variant that needs adjusting. The Daily variant is the hard part.

[cj]

Okay projection folks, let's see if we are on the same page. To me, speed projection is just bringing form/cycle into the figuring. I do like this(not there yet), but what is being projected are speed figures which are based on some sort of baseline or standard which in turn are categorized into a class structure. You can't escape.

Nope, not on the same page, at least with this projection folk.

[Capper Al]

Nope, not on the same page, at least with this projection folk.

Thanks. When I get to projecting speed, my interpretation would be through form/cycle. What else is there to project than will this horse improve or will this horse decline. Anyway, that's the way I see it.

[ubercapper]

Well that is different than your first post. I would never trust a computer with horse racing without human intervention. That is what you get with Equibase and they have some pretty absurd ratings. But that is just me. We all do things differently which is what makes the game great. Good luck.

Years ago, we used to review lists of outliers (for winners) and "try" subjectively to come up with more reasonable figures looking at what the competition did before and after as well as the horse's history, but decided there was more downside than upside to human intervention.

What we take away from this was to find ways to look at races in which there were outliers as part of the automated speed figure creation process, analyze those for patterns, create rules to deal with those, check those rules historically for accuracy, then implement if appropriate.

[highnote]

i know this won't be the answer you want highnote, but it's the best i can do.

you can't tell anything in particular as far as i can tell with that info you provide.
there is nothing unusual about lower class races often running faster raw times than higher class.

but you are doing, as far as i can tell, something i would never recommend.....lumping all the data together.
it needs looking at on a meeting by meeting basis.

roughly speaking i would make EVERY race the same class to begin with.
my way that would be 100 for each and every race.
then i would make a rough set of standard times for every distance using that formula i put on this thread a few days ago.
then meeting by meeting i would iterate through all the data
after each iteration i would compare the class figures with the speed figures.
i would then make the class figures for the second iteration equal to the average speeds of the first iteration, and change the standard times in line with that.
i would just keep doing that until class and speed with each type were the same.
at the end you would know how fast each class is, and as a bonus you would have accurate standards.
where you don't have enough data, you will likely be let down, but that in itself is not a big problem.

if you have expected figures for most races on any card, and also accurate standard times, then you won't have any problems figuring how fast any particular race has been run.
because you have baselines to help you

now i know there is no normal distribution involved here, and some guys that are much more intelligent than me will tell you it's not possible.
unfortunately i never listen to the experts, and just go my own way.
probably if i did listen , i would be bankrupt now, and never have had the life i had, and left on my own terms, to do things that now interest me more.

Your method makes sense for a particular track, but what do you do when there are multiple tracks that form a circuit and some of the horses ship back and forth between them to race?

For example, Grade 1 (Group 1) horses race at Santa Anita in California, but rarely would a Grade 1 horse race at Turf Paradise in Arizona. However, less expensive stock will be shipped between the two tracks to race.

If 100 is Grade 1 at Santa Anita then what is 100 at Turf Paradise and is 100 at SA the same as 100 at TP?

[Dave Schwartz]

roughly speaking i would make EVERY race the same class to begin with.
my way that would be 100 for each and every race.
then i would make a rough set of standard times for every distance using that formula i put on this thread a few days ago.
then meeting by meeting i would iterate through all the data
after each iteration i would compare the class figures with the speed figures.
i would then make the class figures for the second iteration equal to the average speeds of the first iteration, and change the standard times in line with that.
i would just keep doing that until class and speed with each type were the same.
at the end you would know how fast each class is, and as a bonus you would have accurate standards.
where you don't have enough data, you will likely be let down, but that in itself is not a big problem.

So, this is what you WOULD do but you've never actually done it?

I have done it. In fact, I have just done it for the 25th annual time.

While this is pretty close to what I actually do, it is not a trivial thing to accomplish.

What I do works like this:

1. Take the last 2 years of data at a track and analyze the class levels vs the pars, looking for degree of errors.

2. Then I use those class levels to project new pars. (These pars will not be very good, but will serve the purpose in the interim.)

3. Steps 1 & 2 are continued until the change from the previous iteration is very slight. With each iteration, I note the amount of error at the most common classes at the most common distances.

4. When the error/change is small enough, I then toss the created pars and use these established class levels to create new pars from scratch.

5. I take into consideration the history of the pars to determine what the "pattern" is (i.e. distance between the common distances).

6. Then each "key" distance is hand-crafted and adjusted until it correlates well with past races. (Note: A key distance is a distance where there is a relatively large sample of races at class levels that are solid (thinking back to steps 1 & 2).

7. Once the key distances make sense, then I fill in the blanks with whatever races I can find and adjust the times to fit, but using the actual races to extract a "pace multiplier."

Step 7 is how I can create a pretty good par time for a distance that older none-maiden horses may never run: By using the pattern to decide what the final time should be, creating par figs then changing the final time to fit.

8. Create all the class pars by extrapolating them from the theoretical $10k claimer par "Track Par."


Summary
This is not a perfect approach but, as our Speed Reliability Index, a metric I developed to gauge the effectiveness of the pars. This SRI relates how well the horses run back to the par times.


Why do horses not run back well at some tracks?

Sometimes I cannot tell. The most obvious reason is that the pars are just not very good, but that begs the question "Why are the pars not very good?"

To determine that, to begin with, I look for correlation of at least one other set of metrics, completely disconnected from my speed numbers. If the other source is having a difficult time, then I assume that it is not the figs themselves but some other influence.


What are those other influences?

An obvious one is wide-spread discoloration of form at the track by the trainers. Another is nearly open cheating - i.e. horses simply not performing in correlation to their somewhat obvious abilities.

However, the most important is simply the dynamics at the race track in question. There are tracks that are highly formful - with the top speed figs winning a preponderance of races. There are other tracks that just aren't.

Turf tracks can vary significantly from one year to another, especially because of how the cut the grass and, indirectly, the weather. (Dry years cause races to run faster.)

A a general rule, boutique meets generally are highly predictive because the horses (and trainers) are really aiming to win. The obvious boutique meets being DMR, SAR, KEE - but there are others - some being quite small. A few of the smaller fair circuits qualify.

KEE has traditionally been very difficult. I think that is because there are such a wide variety of class levels that are atypical when compared to the backbone of American racing, the Claiming Race.

Since the system of pars I have built depends greatly upon claiming horses - and especially the lower levels - and KEE is mostly for better allowance-grade stock - this provides a ready-made challenge.

However, the pars usually shine at DMR and, to a lesser degree, SAR. The fair circuit in CA is also easy pickings for par-based speed numbers, as are the tiny fairs like FER, TIL, etc.


What about turf racing?
Turf racing is just different. Because of the "slow early-fast late" pace and the big question, "Does the horse like the turf?" speed ratings are simply not as important. This usually results in a lower-than-normal SRI.

Regards,
Dave Schwartz

[InsideTheRaces.com]

not being familiar with the prize money structure over there i can't comment knowingly.
but surely some races will have humungous prize and be won by second raters?
i won't use it myself, because over here they will often have the same prize for a race of a certain type run at woop woop, as they will for that type at a major track.
they are even weighted the same and get the same penalites it they win at the 2nd rate joint as if they did at the major joint.
too ambiguous for me.
or all the races at some tracks will be of the same prize despite hugely different class levels.

thus to me it does not make sense to do it the way you think it should be.

Steve,
Thanks for the response. I have no knowledge about racing in your neck of the woods so I really can't comment on it. It's obviously different here. There are 7 tracks including Parx within a 3 hour drive of Parx. Each track has it's own claiming price structure with different purses even for the same claiming price. I'm just stating that a competitive 10k claimer at one track doesn't make it a competitive 10k claimer somewhere else. It's more about the purse than claiming price.

My original response in this thread was to Capper AL who suggested lumping together all 10k claimers across all tracks to create some kind of par values. As I have stated I don't think that will work.

[highnote]

Thanks, Dave. Sounds like a full-time job!


So, this is what you WOULD do but you've never actually done it?

I have done it. In fact, I have just done it for the 25th annual time.

While this is pretty close to what I actually do, it is not a trivial thing to accomplish.

What I do works like this:

1. Take the last 2 years of data at a track and analyze the class levels vs the pars, looking for degree of errors.

2. Then I use those class levels to project new pars. (These pars will not be very good, but will serve the purpose in the interim.)

3. Steps 1 & 2 are continued until the change from the previous iteration is very slight. With each iteration, I note the amount of error at the most common classes at the most common distances.

4. When the error/change is small enough, I then toss the created pars and use these established class levels to create new pars from scratch.

5. I take into consideration the history of the pars to determine what the "pattern" is (i.e. distance between the common distances).

6. Then each "key" distance is hand-crafted and adjusted until it correlates well with past races. (Note: A key distance is a distance where there is a relatively large sample of races at class levels that are solid (thinking back to steps 1 & 2).

7. Once the key distances make sense, then I fill in the blanks with whatever races I can find and adjust the times to fit, but using the actual races to extract a "pace multiplier."

Step 7 is how I can create a pretty good par time for a distance that older none-maiden horses may never run: By using the pattern to decide what the final time should be, creating par figs then changing the final time to fit.

8. Create all the class pars by extrapolating them from the theoretical $10k claimer par "Track Par."


Summary
This is not a perfect approach but, as our Speed Reliability Index, a metric I developed to gauge the effectiveness of the pars. This SRI relates how well the horses run back to the par times.


Why do horses not run back well at some tracks?

Sometimes I cannot tell. The most obvious reason is that the pars are just not very good, but that begs the question "Why are the pars not very good?"

To determine that, to begin with, I look for correlation of at least one other set of metrics, completely disconnected from my speed numbers. If the other source is having a difficult time, then I assume that it is not the figs themselves but some other influence.


What are those other influences?

An obvious one is wide-spread discoloration of form at the track by the trainers. Another is nearly open cheating - i.e. horses simply not performing in correlation to their somewhat obvious abilities.

However, the most important is simply the dynamics at the race track in question. There are tracks that are highly formful - with the top speed figs winning a preponderance of races. There are other tracks that just aren't.

Turf tracks can vary significantly from one year to another, especially because of how the cut the grass and, indirectly, the weather. (Dry years cause races to run faster.)

A a general rule, boutique meets generally are highly predictive because the horses (and trainers) are really aiming to win. The obvious boutique meets being DMR, SAR, KEE - but there are others - some being quite small. A few of the smaller fair circuits qualify.

KEE has traditionally been very difficult. I think that is because there are such a wide variety of class levels that are atypical when compared to the backbone of American racing, the Claiming Race.

Since the system of pars I have built depends greatly upon claiming horses - and especially the lower levels - and KEE is mostly for better allowance-grade stock - this provides a ready-made challenge.

However, the pars usually shine at DMR and, to a lesser degree, SAR. The fair circuit in CA is also easy pickings for par-based speed numbers, as are the tiny fairs like FER, TIL, etc.


What about turf racing?
Turf racing is just different. Because of the "slow early-fast late" pace and the big question, "Does the horse like the turf?" speed ratings are simply not as important. This usually results in a lower-than-normal SRI.

Regards,
Dave Schwartz

[InsideTheRaces.com]

Hi Steveb,

Below is some data from Fort Erie racetrack in Canada.

Columns are:

TrackAbb, Furlongs, Dirt, ClaimingPrice, FinalTime, 2furlongTime, 4furlongTime, 6furlongTime, SampleSize

Notice how sometimes the lower priced claiming races have faster times than the higher priced claiming races.*

Compare the average final time of the $4,000 claiming races with the average time of the $5,000 claiming races at 8.32 furlongs.

Both have a fairly large sample size, yet, the more expensive horses run slower final times.

At shorter distances, the lower priced horses run faster early fractions than the older horses, but slower final times. (Maybe less expensive horses can't carry their speed as far, even though they are quicker, and that's why they are racing at the bottom level?)

I've seen this at many lower level U.S. tracks, too.


FE 500 D 4000 58.51 22.28 45.78 0 84
FE 550 D 4000 65.09 22.44 45.87 0 55
FE 600 D 4000 71.79 22.62 46.07 0 129
FE 650 D 4000 78.68 23.08 46.32 0 125
FE 832 D 4000 103.88 0 47.28 72.9 126
FE 850 D 4000 106.34 0 47.81 73.21 72
FE 450 D 5000 52.14 22.49 46.14 0 26
FE 500 D 5000 58.43 22.36 45.87 0 215
FE 550 D 5000 65.05 22.47 45.92 0 95
FE 600 D 5000 71.69 22.57 46.02 0 287
FE 650 D 5000 78.62 23.13 46.42 0 156
FE 832 D 5000 104.05 0 47.19 73.32 235
FE 850 D 5000 106.5 0 47.84 73.63 176
FE 900 D 5000 113.46 0 48.64 74.34 17
FE 500 D 10000 58.31 22.49 45.95 0 98
FE 550 D 10000 64.68 22.43 45.75 0 30
FE 600 D 10000 71.48 22.55 45.99 0 134
FE 650 D 10000 78.22 23.23 46.38 0 61
FE 832 D 10000 103.57 0 46.9 73.08 89
FE 850 D 10000 105.76 0 47.67 73.28 63


* In order to have large sample sizes I converted all race times to older male horse times using the adjustments Beyer and Quirin suggest using.
For example, the winning time of a 3 year old filly would have X amount of time deducted from it to bring it into alignment with older male horses. It's possible this caused errors, but I have noticed the same phenomena even without these artificial adjustments.

In order to make par times, I will just use the average time of the higher level claiming times as the par time and extrapolate downwards, otherwise, the lower level horses will have faster pars than the higher level, and I don't think that would be correct.

How do deal with this?

Highnote,

Is your data set all races at a certain claiming price. Meaning does it include maiden, NW2, NW3, NW1 in 6months, NW2 in 6months, NW1 in a year, etc.
If your data includes all conditions under a claiming price then that's why you are seeing funky data(like faster times for lower claiming price).

Try restricting the data to claiming price with no conditions and no restricted races like State Bred and see what you get.

[CincyHorseplayer]

Great post Dave. A lot of questions that pop to a player just playing all over the country and wondering about this and that were answered for me here. And you divulged a lot without giving away the keys to the kingdom. Nicely done!


So, this is what you WOULD do but you've never actually done it?

I have done it. In fact, I have just done it for the 25th annual time.

While this is pretty close to what I actually do, it is not a trivial thing to accomplish.

What I do works like this:

1. Take the last 2 years of data at a track and analyze the class levels vs the pars, looking for degree of errors.

2. Then I use those class levels to project new pars. (These pars will not be very good, but will serve the purpose in the interim.)

3. Steps 1 & 2 are continued until the change from the previous iteration is very slight. With each iteration, I note the amount of error at the most common classes at the most common distances.

4. When the error/change is small enough, I then toss the created pars and use these established class levels to create new pars from scratch.

5. I take into consideration the history of the pars to determine what the "pattern" is (i.e. distance between the common distances).

6. Then each "key" distance is hand-crafted and adjusted until it correlates well with past races. (Note: A key distance is a distance where there is a relatively large sample of races at class levels that are solid (thinking back to steps 1 & 2).

7. Once the key distances make sense, then I fill in the blanks with whatever races I can find and adjust the times to fit, but using the actual races to extract a "pace multiplier."

Step 7 is how I can create a pretty good par time for a distance that older none-maiden horses may never run: By using the pattern to decide what the final time should be, creating par figs then changing the final time to fit.

8. Create all the class pars by extrapolating them from the theoretical $10k claimer par "Track Par."


Summary
This is not a perfect approach but, as our Speed Reliability Index, a metric I developed to gauge the effectiveness of the pars. This SRI relates how well the horses run back to the par times.


Why do horses not run back well at some tracks?

Sometimes I cannot tell. The most obvious reason is that the pars are just not very good, but that begs the question "Why are the pars not very good?"

To determine that, to begin with, I look for correlation of at least one other set of metrics, completely disconnected from my speed numbers. If the other source is having a difficult time, then I assume that it is not the figs themselves but some other influence.


What are those other influences?

An obvious one is wide-spread discoloration of form at the track by the trainers. Another is nearly open cheating - i.e. horses simply not performing in correlation to their somewhat obvious abilities.

However, the most important is simply the dynamics at the race track in question. There are tracks that are highly formful - with the top speed figs winning a preponderance of races. There are other tracks that just aren't.

Turf tracks can vary significantly from one year to another, especially because of how the cut the grass and, indirectly, the weather. (Dry years cause races to run faster.)

A a general rule, boutique meets generally are highly predictive because the horses (and trainers) are really aiming to win. The obvious boutique meets being DMR, SAR, KEE - but there are others - some being quite small. A few of the smaller fair circuits qualify.

KEE has traditionally been very difficult. I think that is because there are such a wide variety of class levels that are atypical when compared to the backbone of American racing, the Claiming Race.

Since the system of pars I have built depends greatly upon claiming horses - and especially the lower levels - and KEE is mostly for better allowance-grade stock - this provides a ready-made challenge.

However, the pars usually shine at DMR and, to a lesser degree, SAR. The fair circuit in CA is also easy pickings for par-based speed numbers, as are the tiny fairs like FER, TIL, etc.


What about turf racing?
Turf racing is just different. Because of the "slow early-fast late" pace and the big question, "Does the horse like the turf?" speed ratings are simply not as important. This usually results in a lower-than-normal SRI.

Regards,
Dave Schwartz

[Dave Schwartz]

HighNote,

It can take some time. My tools have grown and expanded over the years but I still subscribe to the theory "If it ain't broke, don't fix it."

This round of pars saw some tracks take 2+ hours while others took 5-10 minutes.

The process I outlined that starts with rebuilding all the class levels can be quite lengthy.


Cincy,

I actually have a video seminar out there for people who use our pars. Anyone on our email list can simply ask for it and I will send them a link. Truthfully, my business strategy for the pars is based upon:

1. Here is how you do it.
2. Now that you see how difficult it is, wouldn't you rather pay some money to have someone else do it for you?


InsideTheRaces,

Try restricting the data to claiming price with no conditions and no restricted races like State Bred and see what you get.

There are distances which are not run by open claiming. That is kind of the point to building par times: If the entire model doesn't work, your pars will not work.

In other words, if you cannot build pars for the track across multiple race types (i.e. MSW, MdClm, Handicaps, young, etc) then the model is wrong and the pars will be weak.

It is a Catch-22: You need good class levels to make good pars, but first you need good pars to build good class levels, but first you need good class levels...

[highnote]

Highnote,

Is your data set all races at a certain claiming price. Meaning does it include maiden, NW2, NW3, NW1 in 6months, NW2 in 6months, NW1 in a year, etc.
If your data includes all conditions under a claiming price then that's why you are seeing funky data(like faster times for lower claiming price).

Try restricting the data to claiming price with no conditions and no restricted races like State Bred and see what you get.


There are some conditioned claiming races, but they are adjusted per Quirin's recommendations. Some conditioned claiming races like NY1 in a year are not included. State bred races are in their own category.

[green80]

The problem with speed figures is that the horse only runs as fast as it needs to, to win the race.

[highnote]

The problem with speed figures is that the horse only runs as fast as it needs to, to win the race.


That's a good point.

Comprehensive handicapping will always be important.

[InsideTheRaces.com]

InsideTheRaces,



There are distances which are not run by open claiming. That is kind of the point to building par times: If the entire model doesn't work, your pars will not work.

In other words, if you cannot build pars for the track across multiple race types (i.e. MSW, MdClm, Handicaps, young, etc) then the model is wrong and the pars will be weak.

It is a Catch-22: You need good class levels to make good pars, but first you need good pars to build good class levels, but first you need good class levels...

I get that. I was trying to point out to highnote that you can't lump all 4k claiming races together at the same track and come up with an accurate average final time. A open 4k claiming race will be run faster on average than a 4kNW2L claiming race. He's using some method to adjust between the open, nw2,nw3, etc final times so he can lump them all together under a single level labelled 4k claimer. In my opinion that's why his data shows lower level claimers at that track have faster times than higher level claimers. He's taking a basket of oranges, apples, and pineapples putting them all together adjusting the recipe and calling them oranges.

[InsideTheRaces.com]

The problem with speed figures is that the horse only runs as fast as it needs to, to win the race.

And that's why 80% to 90% of horses who finally win their NW3L claiming condition can't win an open claiming race at the same claiming price. That 80-90% were running as fast as they could the whole time but it's not fast enough in non-conditional claiming races. The 10-20% who do go on to win at open claiming still might not be running as fast as they can and might be able to move up a level or two.

[green80]

And that's why 80% to 90% of horses who finally win their NW3L claiming condition can't win an open claiming race at the same claiming price. That 80-90% were running as fast as they could the whole time but it's not fast enough in non-conditional claiming races. The 10-20% who do go on to win at open claiming still might not be running as fast as they can and might be able to move up a level or two.

Thats why many horses are finished when they win that NW3 or the more uncommon NW4. Conditions are way more important than claiming price.

[Dave Schwartz]

I get that. I was trying to point out to highnote that you can't lump all 4k claiming races together at the same track and come up with an accurate average final time.

:ThmbUp: :ThmbUp:

[Tom]

I've done Dave's methods before.
Believe me, it is far easier to buy them from him.
No wonder he gave away that method! :lol:

[traynor]

The best we can do is use an average or a top or a mean as a standard. We need a starting point to measure from. On the contrary, the variant is the hard part-- explaining why the horse did what it did in today's race and quantifying it is difficult.

Starting points are fine. Par times are fine. The difficulty is that handicappers who use either as an absolute reference rather than a (convenient) expression of a normal distribution is likely to adjust himself or herself into the poorhouse.

If a set of values (normal distribution) for given track/class/distance is comprised of data points of which one standard deviation is a full second (or whatever number) plus or minus a center point, a value that is one second slower than "par" is NOT slow, nor a value that is one second faster than "par" fast. Both values fall within the range comprised of the majority of values used to create the "par" (or whatever other average). A valid argument can be made for ignoring values within TWO standard deviations as "normal, expected variations from the baseline value."

With realization that pars, baselines, or starting points--expressed as a single value--represent nothing more than the average of a set of data points rather than an absolute value, making accurate variants is pretty simple. It is regarding the average value as an absolute from which any difference is significant that makes it seem complex.

[Capper Al]

Starting points + Daily variants. It the daily variants that try to guess at the difference. That's the hard part. Why did that horse run fast or why did that horse run slow, needs to be added to the base.

[raybo]

Regarding pars, it would seem to me that a more accurate "starting point", would be a mathematical expression that uses the highest, the lowest, and the median (the point at which there are just as many below it as there are above it) rather than the average (where isolated highs or lows can skew the average in either direction) to determine a time based starting point. This method would also appear to me to get you started well in differentiating between tracks/classes. I'm sure there are better, more complex processes for this, but for those not so mathematically/statistically advanced maybe this stream of thought would be a better starting point.

[Cratos]

Starting points + Daily variants. It the daily variants that try to guess at the difference. That's the hard part. Why did that horse run fast or why did that horse run slow, needs to be added to the base.
Indirectly what I understand you to be addressing with the mentioning of "Daily Variant" is the environmental impact which has not been much of a part of this thread discussion, but it is fundamental and essential in deriving the rate of motion of the racehorse; without that metric you really don't have a "variant."

Also why not derive the predictive curve and evaluate the horses' performance with respect to the AUC or AOC.as oppose to the point estimate of the mean, median, or maximum.

[Capper Al]

Indirectly what I understand you to be addressing with the mentioning of "Daily Variant" is the environmental impact which has not been much of a part of this thread discussion, but it is fundamental and essential in deriving the rate of motion of the racehorse; without that metric you really don't have a "variant."

Also why not derive the predictive curve and evaluate the horses' performance with respect to the AUC or AOC.as oppose to the point estimate of the mean, median, or maximum.

Cratos,

Hope you're okay. I thought with your wind theory that you would be the Daily variant POC.

[traynor]

Starting points + Daily variants. It the daily variants that try to guess at the difference. That's the hard part. Why did that horse run fast or why did that horse run slow, needs to be added to the base.

I agree. That is why it is so misleading when values are considered "off of center" when they are in reality perfectly normal. If some of the data points happen to fall on one side or the other, way too many jump to the conclusion that the "track was fast" or the "track was slow" when the track was completely neutral and the data points are well within the neutral range.

Handicappers who ignore the fact that a par time/starting point is the center point of a range, and consider it an absolute value (like a freeway speed limit--every speed has to be slower than, exactly on, or faster than) tend to spend a lot of time and effort (or to write programming algorithms) that essentially generate nonsense numbers. And to then use those numbers to perform other "adjustments" and "equalizations."

[Cratos]

Cratos,

Hope you're okay. I thought with your wind theory that you would be the Daily variant POC.
I do have a "variant" which have the acronym of SAWRA which stands for Surface Air Wind Resistance Adjustment.

[steveb]

Your method makes sense for a particular track, but what do you do when there are multiple tracks that form a circuit and some of the horses ship back and forth between them to race?

For example, Grade 1 (Group 1) horses race at Santa Anita in California, but rarely would a Grade 1 horse race at Turf Paradise in Arizona. However, less expensive stock will be shipped between the two tracks to race.

If 100 is Grade 1 at Santa Anita then what is 100 at Turf Paradise and is 100 at SA the same as 100 at TP?


i do multiple tracks the same way, and i run them all at the same time using arrays highnote.
it's not a problem.
i just gave you a rough guide what happens, i could but i am not, going to go into detail.
a 100 would be 100 where ever it is run, and the 100 by my way would not be the fastest but it is certainly group 1 speed.
thus many places, will never have horses attaining that figure.

it matters not to me what standard the track runs, the same process will figure them all out, and because the base is always equivalent to 100, then the differences will show up in the 'variant'.
which in fact is the ONLY use i can find for a 'variant'.

what you need to remember highnote, is that people will be reading these posts and taking gratis.
well the smart ones will anyway.:lol:
this board has trillions of theorists, but not too many are telling you what i am telling you.
they can fill in the finer details themselves if they are smart enough.

[steveb]

So, this is what you WOULD do but you've never actually done it?




this is all i have read above.
but why would you think i have never done it?
i started writing code for this on an old commodore 64 two lifetimes ago.

the only people doing it like me are the people i had told, either for payment, or because i was feeling generous and was no longer interested in betting on racing.
been there done that.

if you did it like me, and i bet you don't, then you would know that there are very many things you would discover seemingly unrelated to time, but in fact are.

[steveb]

HighNote,

It can take some time. My tools have grown and expanded over the years but I still subscribe to the theory "If it ain't broke, don't fix it."

This round of pars saw some tracks take 2+ hours while others took 5-10 minutes.

The process I outlined that starts with rebuilding all the class levels can be quite lengthy.


Cincy,

I actually have a video seminar out there for people who use our pars. Anyone on our email list can simply ask for it and I will send them a link. Truthfully, my business strategy for the pars is based upon:

1. Here is how you do it.
2. Now that you see how difficult it is, wouldn't you rather pay some money to have someone else do it for you?


InsideTheRaces,



There are distances which are not run by open claiming. That is kind of the point to building par times: If the entire model doesn't work, your pars will not work.

In other words, if you cannot build pars for the track across multiple race types (i.e. MSW, MdClm, Handicaps, young, etc) then the model is wrong and the pars will be weak.

It is a Catch-22: You need good class levels to make good pars, but first you need good pars to build good class levels, but first you need good class levels...

i have not yet read all your first post that i commented on, because it's long and i will read later.
but this post shows me you are doing it nothing the same, for the simple reason, that if you give me the data in the required format so that the code needed no tweaking, i could give you class and time and whole lot more in a matter of minutes, and i would need to know sweet nothing about any of them.
data is king.
for australia that would hundreds of tracks, and i would do it all at the same time
of course the tracks that don't do accurate distance or time would be problematical, but where ever the data is accurate it would be fine.
nothing ever is 100%, so i would never dream of claiming that, i do know it would be better than the theorists will ever accomplish though.

[steveb]

The problem with speed figures is that the horse only runs as fast as it needs to, to win the race.

thank you green80 for such a revelation.
and the higher the class, then generally it needs to run faster.


if you understood what you can learn from time, you would not be stating things like that.
speed figures are just one tiny aspect of what you could learn.

i once upon a time shunned 'time', even during the first couple of years that racing was my life, and that was because the books i had read canned time, and the reason put forth made sense.
and me being an idiot, listened to all that nonsense.
then i read a beyer book, and it got me thinking.
that was maybe 30 years ago, and i just in the past year or so stopped betting, because i no longer had the urge to make lots of money.
i would rather watch good looking chicks on the beach these days.(feathered ones!!)

[InsideTheRaces.com]

thank you green80 for such a revelation.
and the higher the class, then generally it needs to run faster.


if you understood what you can learn from time, you would not be stating things like that.
speed figures are just one tiny aspect of what you could learn.

i once upon a time shunned 'time', even during the first couple of years that racing was my life, and that was because the books i had read canned time, and the reason put forth made sense.
and me being an idiot, listened to all that nonsense.
then i read a beyer book, and it got me thinking.
that was maybe 30 years ago, and i just in the past year or so stopped betting, because i no longer had the urge to make lots of money.
i would rather watch good looking chicks on the beach these days.(feathered ones!!)

Steve,
You don't know Jack about racing over here. Sorry dude you don't.
Just like I don't know anything about the racing in your neck of the woods.
Your bottom level horses are running for peanuts. When you pay peanuts you get Monkeys. I'm pretty sure Horses can run faster than Monkeys.

I have a my own speed rating. I call it TOMWYIJ it stands for Time Only Matters When Your In Jail.

[steveb]

Steve,
You don't know Jack about racing over here. Sorry dude you don't.
Just like I don't know anything about the racing in your neck of the woods.
Your bottom level horses are running for peanuts. When you pay peanuts you get Monkeys. I'm pretty sure Horses can run faster than Monkeys.

I have a my own speed rating. I call it TOMWYIJ it stands for Time Only Matters When Your In Jail.

thank you sir, it must be good being such a knowledgeable person.
i dips me lid to you sir.
well done for knowing what you know 'dude'!

if you think country is any big deal, you are off with the fairies.
they are horses,they run around race tracks, they have little blokes sitting on top and the aim is to be first to reach the finish line.
is it different in you fine country 'dude'??

i am pretty sure the blokes that win all the money in your fine country 'dude' are really happy there is 'dudes' like you 'dude'.

[PaceAdvantage]

Steve,
You don't know Jack about racing over here. Sorry dude you don't.
Just like I don't know anything about the racing in your neck of the woods.
Your bottom level horses are running for peanuts. When you pay peanuts you get Monkeys. I'm pretty sure Horses can run faster than Monkeys.

I have a my own speed rating. I call it TOMWYIJ it stands for Time Only Matters When Your In Jail.You can restate what you just wrote in a much more pleasant manner.

After all, we're not talking politics or religion or even race fixing here. Why the need to get all uppity?

[Dave Schwartz]

i have not yet read all your first post that i commented on, because it's long and i will read later.
but this post shows me you are doing it nothing the same, for the simple reason, that if you give me the data in the required format so that the code needed no tweaking, i could give you class and time and whole lot more in a matter of minutes, and i would need to know sweet nothing about any of them.
data is king.
for australia that would hundreds of tracks, and i would do it all at the same time
of course the tracks that don't do accurate distance or time would be problematical, but where ever the data is accurate it would be fine.
nothing ever is 100%, so i would never dream of claiming that, i do know it would be better than the theorists will ever accomplish though.

... Another guy who knows all about how everyone else should do things but isn't doing it himself.

Why are you here?

I call. How about I give you some of that data and you just tweak it up for us in a minute or two?

[pandy]

Everyone knows the formulas being used:


Parallel Speed Chart adjustment
Track variant
Daily variant


But consider how the speed figures are adjusted when it comes to the daily variant. If speed makers are trying to peg a daily variant the first thing they say is something like this: These $20,000 Claimers should be running this distance at 1:05(example). They ran a 1:10 so they are 5 over. Then they subtract 5 for the daily variant. (It's not exactly this simple, but overall this is what they do.) So the daily variant would be adjusted with a minus 5 to equalize. Here's the catch, with all this equalizing they are negating the actual speed and going by what the class of horses should of ran. So is it speed or is it class? The speed has been factored out of it.


I made speed figures for many years, they were published in Racing Action and in the NYRA track program.

I'll say this, if you're a bettor who wants to hit as many bets as possible, and you use speed figures to try and pick winners, I'd suggest you run a test comparing fast times speed figures vs slow times speed figures.

For instance, say that at Aqueduct, or any track, a MSW horse chases the pace, is 2 to 3 lengths behind the leader, and gets up to win by a head and the fractions and time are, :22.2, :45.1, 1:09. The speed figure assigned to this effort is 90. The track is listed as fast and the variant is +10.

A few days later the track is also listed as fast, but the variant is 0. A MSW horse chases the pace 2 to 3 lengths behind and wins by a head and the fractions and time are, :23.3, :46.3, 1:11. The speed figure assigned to this effort is 90.

These two horses come back to race against each other in NW1. After adjusting for the variant, both horses have identical pace figures and identical speed figures of 90.

I've found that in these situations, the horse that actually ran faster on the track will beat the horse that ran slower on the track more often than not. Times, actual times, do matter.

[Capper Al]

Thanks Pandy. My speed figs (AMS and EURO) would most likely favor the horse that clocked the fastest time. Your example, most likely occurs when the daily variant/s were in error.

[classhandicapper]

I made speed figures for many years, they were published in Racing Action and in the NYRA track program.

I'll say this, if you're a bettor who wants to hit as many bets as possible, and you use speed figures to try and pick winners, I'd suggest you run a test comparing fast times speed figures vs slow times speed figures.

For instance, say that at Aqueduct, or any track, a MSW horse chases the pace, is 2 to 3 lengths behind the leader, and gets up to win by a head and the fractions and time are, :22.2, :45.1, 1:09. The speed figure assigned to this effort is 90. The track is listed as fast and the variant is +10.

A few days later the track is also listed as fast, but the variant is 0. A MSW horse chases the pace 2 to 3 lengths behind and wins by a head and the fractions and time are, :23.3, :46.3, 1:11. The speed figure assigned to this effort is 90.

These two horses come back to race against each other in NW1. After adjusting for the variant, both horses have identical pace figures and identical speed figures of 90.

I've found that in these situations, the horse that actually ran faster on the track will beat the horse that ran slower on the track more often than not. Times, actual times, do matter.

This is one reason why horse racing is so tough to beat.

There are some really smart guys out there (like Pandy) that are willing to share insights that run counter to conventional wisdom on the subject.

I'm pretty sure I know why what you are saying is probably true, but I don't want to elaborate because when I retire I may start making my own figures again using an entirely different model than anyone else uses that will incorporate concepts like this.

[pandy]

The reason why I know this, by the way, and those of you who have used either my Diamond System or something like the Brohamer/Sartin style velocity ratings, will know that even though the system adjusts the fractions and final time for the track variant, the horses that actually run faster are almost always ranked ahead of horses that ran slower, unless the track was a total bog and the variant was crazy slow.

And using my Diamond System as an example, I see many, and I mean, a lot, of horses that have earned a strong speed figure in slow time and are therefore not ranked well by my system, and, of course, there are horses that ran a strong speed figure in fast time and are highly ranked by the system. Assuming that both of these type of strong figure horses are heavily bet, say they are both the favorite, I can assure you, based on my experiences of analyzing these rankings, the horse that ran the fast time strong speed figure has a much better chance of not being an underlay, because it has a much better chance of winning.

Now, I know that speed figure aficionados would say this is hogwash because the speed figure makes up for the track surface speed. And I'm not saying that I don't believe in speed figures. I think they are an excellent handicapping aid. However, the top figure horses with the actual faster times beat the top figure horses with the slower times, on any surface, often enough were I feel I can positively state that it is not an anomaly.

And that's why I rarely if ever bet on a top figure horse that ran a slow time unless I'm reasonably sure that the track was a total bog that day, and even then, I need a price. If you bet top figure favorites that ran slow you will get destroyed. If you are a chalk bettor, you should only bet on horses that have strong speed figures but earned those figures by running a fast time.

Same thing when trying to find false favorites, try to beat the slow times, not the fast ones. Now of course, a track like Turf Paradise, which is ridiculous, is another story. But most of the regular tracks that have decent horses are not crazy fast.

[steveb]

... Another guy who knows all about how everyone else should do things but isn't doing it himself.

Why are you here?

I call. How about I give you some of that data and you just tweak it up for us in a minute or two?


i don't know what has rocked your boat dave, not that i care.
i read your stuff now, and it is nothing like what i do, which i already knew it wouldn't be.

i am here because mr pace advantage has not banned me, and thus lets me contribute if i feel like it.

as for calling me out, that is not too clever, i have put lots of stuff on here over time and explanations.
and what is more i am not trying to flog anything to anyone.

it is what it is, and it's not the latest and greatest, because i have been doing it for what seems to be forever.

but if you insist i will explain it all, but first YOU put all your stuff up here for gratis, that you charge people for?
deal?

every single time a meeting/s are added to data, everything is updated.
if it was time consuming do you really think i would do that?
you are a programmer i thought.
computers are pretty quick, even when used by self taught crappy code writers like me.

[ultracapper]

The reason why I know this, by the way, and those of you who have used either my Diamond System or something like the Brohamer/Sartin style velocity ratings, will know that even though the system adjusts the fractions and final time for the track variant, the horses that actually run faster are almost always ranked ahead of horses that ran slower, unless the track was a total bog and the variant was crazy slow.

And using my Diamond System as an example, I see many, and I mean, a lot, of horses that have earned a strong speed figure in slow time and are therefore not ranked well by my system, and, of course, there are horses that ran a strong speed figure in fast time and are highly ranked by the system. Assuming that both of these type of strong figure horses are heavily bet, say they are both the favorite, I can assure you, based on my experiences of analyzing these rankings, the horse that ran the fast time strong speed figure has a much better chance of not being an underlay, because it has a much better chance of winning.

Now, I know that speed figure aficionados would say this is hogwash because the speed figure makes up for the track surface speed. And I'm not saying that I don't believe in speed figures. I think they are an excellent handicapping aid. However, the top figure horses with the actual faster times beat the top figure horses with the slower times, on any surface, often enough were I feel I can positively state that it is not an anomaly.

And that's why I rarely if ever bet on a top figure horse that ran a slow time unless I'm reasonably sure that the track was a total bog that day, and even then, I need a price. If you bet top figure favorites that ran slow you will get destroyed. If you are a chalk bettor, you should only bet on horses that have strong speed figures but earned those figures by running a fast time.

Same thing when trying to find false favorites, try to beat the slow times, not the fast ones. Now of course, a track like Turf Paradise, which is ridiculous, is another story. But most of the regular tracks that have decent horses are not crazy fast.

Seriously!?!? Anybody that hasn't noticed the basic truth to this premise is blind. A horse that has continually shown an inability to break 1:11, even being assigned higher speed numbers than a horse that has proven breaking 1:11 is commonplace, will empty the handicappers bank account pretty quick. Just about everybody uses some kind of speed or pace numbers, but absolutely ignoring raw times is folly. They may not be reliable because of different conditions, but the race always goes to the first one across the line, and the horse that can run the distance in the shortest time has an advantage.

[steveb]

... Another guy who knows all about how everyone else should do things but isn't doing it himself.

Why are you here?

I call. How about I give you some of that data and you just tweak it up for us in a minute or two?

not to mention my post in this thread at 144 has something sent to me by Alan Woods, wanting to know my opinion of their attempt to code their way, what i do.

so what i do is roughly explained there, but my modesty does not permit me, to not say that their way was crap!

your way was nothing like that, because their way, even if it was not so good yet(they would have improved it over time as they figured it better), would also do all their data in minutes.

[pandy]

Seriously!?!? Anybody that hasn't noticed the basic truth to this premise is blind. A horse that has continually shown an inability to break 1:11, even being assigned higher speed numbers than a horse that has proven breaking 1:11 is commonplace, will empty the handicappers bank account pretty quick. Just about everybody uses some kind of speed or pace numbers, but absolutely ignoring raw times is folly. They may not be reliable because of different conditions, but the race always goes to the first one across the line, and the horse that can run the distance in the shortest time has an advantage.

A lot of people would disagree with us on this. I wrote something like this a few months ago and some people said I was wrong,.

[ultracapper]

A lot of people would disagree with us on this. I wrote something like this a few months ago and some people said I was wrong,.

I would strongly agree with you. Excitement for raw times must be tempered for sure, but if a horse has run 6f in 1:09 and another has not, I want the one that hasn't to prove it first, particularly the larger the sample.