Track Variants - Bogus or real?

[TrifectaMike]

Except for extreme cases, do speed handicappers really believe they are measuring surface anomalies for the day?

Is a variant based on reality or simply number crunching without merit and most likely wrong.

I personally believe track variants deter not aid speed handicappers.

Make your case... convince that I am wrong in my belief.

Mike (Dr Beav)

[HUSKER55]

FWIW, I think the track variant is mored indicative of the quality of horses and has very little, if anything, to do with the speed of the track.

The DRF takes the average par time of the day and compares it to the par time of today's entrants.

Any jump between races can be explained by class of the animals. How else could you get two different beyers for the same time-of-race in a given day at the same track?

[cj]

I personally believe track variants deter not aid speed handicappers.

Make your case... convince that I am wrong in my belief.


Make your case first. Why do you believe that?

[cj]

FWIW, I think the track variant is mored indicative of the quality of horses and has very little, if anything, to do with the speed of the track.

The DRF takes the average par time of the day and compares it to the par time of today's entrants.

Any jump between races can be explained by class of the animals. How else could you get two different beyers for the same time-of-race in a given day at the same track?

First, you are talking about two different variants and trying to relate them. The DRF variant has nothing to do with Beyer figures.

How two different Beyer figures are given for the the same time is covered below:

http://www.paceadvantage.com/forum/showthread.php?t=99477

I'd sure like to see some examples of this happening. I'm not saying it never happens, but it is a rare occurrence indeed. It was stated in the thread linked above and now by you like this happens every day, and I can assure you it does not.

[jorcus99]

Except for extreme cases, do speed handicappers really believe they are measuring surface anomalies for the day?


Is a variant based on reality or simply number crunching without merit and most likely wrong.

I personally believe track variants deter not aid speed handicappers.

Make your case... convince that I am wrong in my belief.

Mike (Dr Beav)

I guess it depends on what you consider extreme. Were the slow times at Hollywood park last weekend extreme enough to create a vairant? Most would say yes. If that is he case where do you draw the line? You pretty much have to accept it exists at some point.

[therussmeister]

Except for extreme cases, using or not using variants would not likely impact a handicappers success rate, therefore, whilst they don't help, neither would they deter. But if you apply variants every day, instead of just extreme days, you don't have to define what constitutes extreme.

[Magister Ludi]

Is a variant based on reality or simply number crunching without merit and most likely wrong.

It's a moot point. Track variants are so widely used that if there is any useful information added by them to a forecast, it is already reflected in the public odds.

[Robert Goren]

It's a moot point. Track variants are so widely used that if there is any useful information added by them to a forecast, it is already reflected in the public odds.But if you don't use them, then you are at a disadvantage to the public unless you use the odds as part of making your picks. I suppose a case could be made that they over bet and you could bet against them. Good luck with that.

[Magister Ludi]

But if you don't use them, then you are at a disadvantage to the public unless you use the odds as part of making your picks.

Bingo!

[Robert Goren]

Except for extreme cases, using or not using variants would not likely impact a handicappers success rate, therefore, whilst they don't help, neither would they deter. But if you apply variants every day, instead of just extreme days, you don't have to define what constitutes extreme.It has been my experence that if you have an extreme variant for the day, the horses don't run back to that race unless today races also have similar extreme variant. For instance a horse ran on day when the variant was 20 point slower normal. That race is useless unless today's races are running about 20 points slower than normal. Some horse only run on extremely slow or extremely fast tracks. Some horses won't run a lick on an extreme variant day. That is why you should always know the variant for the day. This is something that variant makers never tell anyone.

[cj]

It has been my experence that if you have an extreme variant for the day, the horses don't run back to that race unless today races also have similar extreme variant. For instance a horse ran on day when the variant was 20 point slower normal. That race is useless unless today's races are running about 20 points slower than normal. Some horse only run on extremely slow or extremely fast tracks. Some horses won't run a lick on an extreme variant day. That is why you should always know the variant for the day. This is something that variant makers never tell anyone.

Of course there are some horses that like a very deep track, and some a very hard one. But those horses are in the vast minority. As an example, Mountaineer winterizes its track every year, but most horses are still able to repeat figures on the non-winterized track, a markedly different surface.

[Greyfox]

Make your case first. Why do you believe that?

:ThmbUp: I'll second that notion. My guess is that we'll either wait a long time or never get a direct answer.

[RonTiller]

This question has nothing to do with anybody's beliefs or theories - it is an empirical question, pure and simple. To that end, I think it is a good question that speed figure makers should be able to answer with facts.

To rephrase the original question: How do speed numbers measure up made a) without variants, b) with variants (using whatever methodology you use)?

Jim Cramer recently measured the performance of his speed numbers for every race run in 2012 - that's EVERY RACE, 47,365, from Prineville Rodeo Roundup to Belmont. This includes all races, no filtering out 1st time starter races, 2 year old races, off tracks, races we don't like, etc. Here's the result:


NO Variant 47,365 races 26.5% winners #1 last race speed rating .84 ROI
WITH Variant 47,365 races 30.0% winners #1 last race speed rating .92 ROI

What does this mean? Using raw speed numbers, with no variant (but with track to track built in), the #1 last race Cramer speed rating horse won 26.5% of the all races run in 2012. (Note that races with all 1st time starters are included, but they automatically count as a loss in this tally).

Taking the same 47,365 races, but this time using the speed numbers with Jim's variant included in the number, the #1 last race Cramer speed rating horse won 30.0% of the all races run in 2012. In other words, the track variant that Jim produces ups the win% by 3.5%.

This is really simple engineering. Make both sets of numbers, with and without variant, measure them both over a large sample (however you choose to measure them), and compare results.

That is how you optimize a specific methodology of making variants as well - measure numbers made using different variants (and different methodlogies, weightings and go with the best performing.

If you don't measure the speed ratings in an empirical manner (however you choose to measure them), you have no firm basis to answer TrifectaMike's excellent question.

So my answer to TrifectaMike's question, philosophical beliefs, opinions, theories and anecdotal evidence aside, is that Cramer speed numbers perform significantly better (as demonstrated through large empical studies) with his variant incorporated into the number than with no variant applied.

Please note that this data is using Cramer speed numbers with and without his variants, which are projected variants, based on projections of how fast each horse in the race will run in that race. Your mileage may vary with other speed ratings and variant methodologies.

Ron Tiller
HDW

[sjk]

Many would agree that a good first step towards analyzing today's race is to assign numerical grades (adjusted speed ratings if you will) to each horse's past performance lines.

How would you begin to do this without a track variant?

[TexasDolly]

This question has nothing to do with anybody's beliefs or theories - it is an empirical question, pure and simple. To that end, I think it is a good question that speed figure makers should be able to answer with facts.

To rephrase the original question: How do speed numbers measure up made a) without variants, b) with variants (using whatever methodology you use)?

Jim Cramer recently measured the performance of his speed numbers for every race run in 2012 - that's EVERY RACE, 47,365, from Prineville Rodeo Roundup to Belmont. This includes all races, no filtering out 1st time starter races, 2 year old races, off tracks, races we don't like, etc. Here's the result:


NO Variant 47,365 races 26.5% winners #1 last race speed rating .84 ROI
WITH Variant 47,365 races 30.0% winners #1 last race speed rating .92 ROI

What does this mean? Using raw speed numbers, with no variant (but with track to track built in), the #1 last race Cramer speed rating horse won 26.5% of the all races run in 2012. (Note that races with all 1st time starters are included, but they automatically count as a loss in this tally).

Taking the same 47,365 races, but this time using the speed numbers with Jim's variant included in the number, the #1 last race Cramer speed rating horse won 30.0% of the all races run in 2012. In other words, the track variant that Jim produces ups the win% by 3.5%.

This is really simple engineering. Make both sets of numbers, with and without variant, measure them both over a large sample (however you choose to measure them), and compare results.

That is how you optimize a specific methodology of making variants as well - measure numbers made using different variants (and different methodlogies, weightings and go with the best performing.

If you don't measure the speed ratings in an empirical manner (however you choose to measure them), you have no firm basis to answer TrifectaMike's excellent question.

So my answer to TrifectaMike's question, philosophical beliefs, opinions, theories and anecdotal evidence aside, is that Cramer speed numbers perform significantly better (as demonstrated through large empical studies) with his variant incorporated into the number than with no variant applied.

Please note that this data is using Cramer speed numbers with and without his variants, which are projected variants, based on projections of how fast each horse in the race will run in that race. Your mileage may vary with other speed ratings and variant methodologies.

Ron Tiller
HDW


It may be that the variant produces more winners but it seems to me that the ROI is really what's important ,not win percentage as a gauge.
I had also read that Cramer assigns a number to a horse and then as the horse runs more races goes back and adjusts the original(or previous numbers) based on subsequent races. I don't know if this a fact so I would appreciate your comment on this as well. If so, how does that affect the variant etc.?
Thank you,
TD

[cj]

:ThmbUp: I'll second that notion. My guess is that we'll either wait a long time or never get a direct answer.

I agree, and if we don't, this thread will just close.

[TrifectaMike]

I agree, and if we don't, this thread will just close.

Close it

[raybo]

My opinion on the use of daily variants is that, I don't care if the times were faster due to a faster surface, or, faster horses (class). All I want to know is, were the times faster or slower and by how much. I use them to adjust the raw times and don't care what the perceived class of the horses were, or what the perceived surface speed was, I just want to know what kind of pace the horses contended against. There are more than just surface condition and class of horses included in the variant, it also includes wind, trouble, or lack thereof, during the running of the race, jockey maneuverings, paths taken/ground lost or saved, etc., etc.. I like having all that info included in the variant.

Most would say that the variant has no "specific" meaning, I happen to believe that; it doesn't matter that it isn't specific, regarding a particular factor like surface speed, for example. The way I use it simply works, for what I am using it for, adjusting raw fractional and final times.

[cj]

Close it

I'll leave it open, just won't let you participate.

[MightBeSosa]

This would have been a better question for April 1.

Are we asking if variant adjusted figs are better than raw figs?

And if so, what would be the point? That Thorograph and Ragozin and Beyer are all wasting our time?:confused:

[traynor]

Except for extreme cases, do speed handicappers really believe they are measuring surface anomalies for the day?

Is a variant based on reality or simply number crunching without merit and most likely wrong.

I personally believe track variants deter not aid speed handicappers.

Make your case... convince that I am wrong in my belief.

Mike (Dr Beav)

Suppose one looks at 1000 six-furlong races at Track A, and discovers (after correcting for outliers) that the majority of times are between 1:10 and 1:12. One might then surmise that a "normal" time for six-furlongs at Track A is 1:11. What is missing is the realization that the base figure is derived from a cluster of values around a central point, not a specific, carved-in-stone value.

A six-furlong race at Track A in 1:10 is NOT "fast" nor is a six-furlong race at Track A in 1:12 "slow."

I think everyone who uses speed in his or her judgments regarding past performances starts out "overadjusting" and essentially massaging numbers. With experience, one tends to realize that a "normal" time is a range--not a specific value--and determines variants accordingly.

I think there is a great deal of value in accurate variants, with the emphasis on "accurate." That presupposes substantially more effort than simplistic "par times" and labeling anything that does not exactly match that artificial value as being "fast" or "slow."

[Robert Fischer]

the track variants - (whether intentional or not), in reality, add a class element.

This should result in figures that perform slightly better than the raw times figures.

However that isn't really a true answer to the actual question that Mike was asking in this thread.

[cj]

I think there is a great deal of value in accurate variants, with the emphasis on "accurate." That presupposes substantially more effort than simplistic "par times" and labeling anything that does not exactly match that artificial value as being "fast" or "slow."

NO competent figure maker is doing so using pars.

[Robert Fischer]

...
Is a variant based on reality[?...]

lets suppose that the variant is not based on reality. Or to be more specific, lets suppose the variant does a pretty fair job as a class measurement, but that it is very hit or miss as an actual adjustment in pure speed.

What is the alternative?
You could take all the raw unadjusted times and throw out a good amount of the outliers.

Then you could treat the remaining raw time's (that fall within the qualifying cluster) as a their "real" speed figure.

you could then use your "real" speed figure as a stand-alone and in combination with other factors such as a "real" class figure.

maybe after all that work you could come up with something slightly different.

[traynor]

NO competent figure maker is doing so using pars.

I think that is what I just said.

[MightBeSosa]

NO competent figure maker is doing so using pars.

I used par times for a week once back in the 70's , then had enough data to do it right for the next 40 years.

[RonTiller]

I guess I must have missed something - not sure why the controversy. This is an important topic that speed figure makes should be able to intelligently address to customers and the general public.

It may be that the variant produces more winners but it seems to me that the ROI is really what's important, not win percentage as a gauge.

ROI has no bearing in making our numbers. It is what it is. Speed ratings are intended to measure the speed of each horse. I get it that handicappers care about ROI but that is a separate issue. We and others use these numbers to buy and sell horses as well as handicap. If somebody else makes a speed rating meant to optimize ROI rather than performance (however that is measured) that is their prerogative. I certainly wouldn't want to claim a horse off numbers optimized for maximum ROI.

I had also read that Cramer assigns a number to a horse and then as the horse runs more races goes back and adjusts the original(or previous numbers) based on subsequent races.

This is in general correct but it is easy to get the wrong impression. Variants are automatically recalculated for a period after the race. The main reason for this is that the projected final times for each race vary in their strength. For instance, a race at CD with 5 and 6 year old horses who just run 6F races month after month will have a stronger, more reliable projection than a 6F race of 2YOs with 5 1st time starters and 2 horses that ran in a 2F race at SA. In this case, there is so little to go on that the projection is more iffy than in the HAW case. In fact, the strength of the projection is scored and that score can be used for weightings if desired.

When the 2YOs race back, this gives more data points for the original race, where there was only enough data for a less reliable projection. So the variants, when recalculated, may cause the original race to have a higher or lower speed rating. Since the number calculated for that original race is itself used in future projections, it is vital to make that adjustment to the original race. It is also important to note that the new variant for the original race can impact the variant applied to the rest of the races on that original card, whose numbers will also be used in future projections. Thus, modifying the variant in one race can have a widespread impact on a lot of speed ratings.

In the end, Jim has done the work to test and measure speed ratings with recalculated variants versus speed ratings with variants set in stone on the day of the race and the recalculated ones win hands down.

I think there is a great deal of value in accurate variants, with the emphasis on "accurate." That presupposes substantially more effort than simplistic "par times" and labeling anything that does not exactly match that artificial value as being "fast" or "slow."
Yes, although I would be more inclined to say "optimized" rather than "accurate", as I'm not sure there is a "correct" variant that is there waiting to be approximated or guessed. Engineer numbers, test, adjust, retest, etc. is about all you can do, with an eye to improving the numbers based on an objective measurement, whatever that is. The heart of it all is in the projections for each race - if they are bad, obviously the variants will be bad as well.

Also, it is important to note that there is no subjectivity to these numbers - in other words, Jim does NOT simply change a number because it looks out of whack. Nor does he pour over today's races with an eye to changing old speed ratings, say from 2 weeks ago. The methodology he has developed over almost 20 years is designed to take all subjectivity out. Problematic numbers are dealt with by modifying the methodology that led to the problematic number in the first place, NOT by typing in a "better" number and crossing his fingers.

Ron Tiller
HDW

[Saratoga_Mike]

Jim Cramer recently measured the performance of his speed numbers for every race run in 2012 - that's EVERY RACE, 47,365, from Prineville Rodeo Roundup to Belmont. This includes all races, no filtering out 1st time starter races, 2 year old races, off tracks, races we don't like, etc. Here's the result:


NO Variant 47,365 races 26.5% winners #1 last race speed rating .84 ROI
WITH Variant 47,365 races 30.0% winners #1 last race speed rating .92 ROI


So my answer to TrifectaMike's question, philosophical beliefs, opinions, theories and anecdotal evidence aside, is that Cramer speed numbers perform significantly better (as demonstrated through large empical studies) with his variant incorporated into the number than with no variant applied.

Ron Tiller
HDW

Great info - thanks.

[Saratoga_Mike]

Close it

Let's hope it stays open - some interesting data being presented refuting your main assertion.

[cj]

This is in general correct but it is easy to get the wrong impression. Variants are automatically recalculated for a period after the race. The main reason for this is that the projected final times for each race vary in their strength. For instance, a race at CD with 5 and 6 year old horses who just run 6F races month after month will have a stronger, more reliable projection than a 6F race of 2YOs with 5 1st time starters and 2 horses that ran in a 2F race at SA. In this case, there is so little to go on that the projection is more iffy than in the HAW case. In fact, the strength of the projection is scored and that score can be used for weightings if desired.

Ron Tiller
HDW

While there is certainly NOTHING wrong with adjusting figures later down the road to get the most accurate numbers possible, I would say doing this means printing Win percentages and ROIs on figures that may have very well changed after the horse has subsequently run are misleading.

For example, Horse A wins his debut and is given a 55. It is the top figure in the field next time he runs. After that race though, the figure is adjusted to 53 and is no longer the top figure in that race.

If you want to measure ROI and W%, you have to base it on the figures that were in the PPs at the time, not what they were later changed to reflect. Maybe you are doing that, but it isn't clear to me.

[TexasDolly]

While there is certainly NOTHING wrong with adjusting figures later down the road to get the most accurate numbers possible, I would say doing this means printing Win percentages and ROIs on figures that may have very well changed after the horse has subsequently run are misleading.

For example, Horse A wins his debut and is given a 55. It is the top figure in the field next time he runs. After that race though, the figure is adjusted to 53 and is no longer the top figure in that race.

If you want to measure ROI and W%, you have to base it on the figures that were in the PPs at the time, not what they were later changed to reflect. Maybe you are doing that, but it isn't clear to me.

This was my point in asking the question originally about Cramer "redoing" the ratings. In addition, to the valid point CJ made, consider a horse that had an inferior figure going in and wins the race with an improved figure . The subsequent adjustment may make the horse the choice because of the improved rating. How it affects the variant process is anyone's guess. It is no surprise to me that Cramer variant and race adjusted speed ratings produce a higher percentage of winners. In fact, I would think that considerably more horses do run an improved number to win than run an inferior one . The subsequent adjustment to the ratings insures that the new speed ratings have a higher win percentage.

[Cratos]

Except for extreme cases, do speed handicappers really believe they are measuring surface anomalies for the day?

Is a variant based on reality or simply number crunching without merit and most likely wrong.

I personally believe track variants deter not aid speed handicappers.

Make your case... convince that I am wrong in my belief.

Mike (Dr Beav)

Mike, you have raised a most interesting topic with a salient question that needs to be answered by track variant users with following explanations:

1. A definition of the daily variant
2. An application of the daily variant
3. An expectation of the result of the application of the daily variant.

My criticism of the daily variant is that it lacks the science of physics and use very elementary math and statistics to solve a very complex problem.

But let’s explore the daily variant concept from an application and resultant point of view. A horse moving through space over a racetrack surface which is static and the expected resultant of the racetrack surface is a variant measured in time increments with the magnitude of velocity (speed).

That relationship between the horse hooves and the track surface is the relationship between kinetic friction and normal force which results in the coefficient of friction; simply stated is resistance. It is not a variant of track surface speed because the track is static and do not have speed; again it have resistance to speed.

However what is largely missing in the speed figure/variant calculation is how the track turns relate to the distance of the race in terms of percent of the race.

Through calculations, It can easily be concluded that for sheer speed of the horse with respect to race length, the 7f distance at Saratoga would be the best distance because it has the longest run to the turn and the smallest percent of the race in the turn.

Therefore on a day when the dominant race distance is the 7f at Saratoga how do the “speed guys” handle the variant since “speed “is their measuring tool so to speak and turn impact is a major detriment to how fast a horse run over a racetrack.

I will now answer your question: “Is a variant based on reality or simply number crunching without merit and most likely wrong?”
The variant in my opinion is based on simply number crunching without merit and most likely wrong.

To calculate a surface variant there will be a need for some sophisticated devices and operated by very well train technicians. I say this because the most influential change to the race track static surface comes from the environment in terms of thermal, wind, and moisture. This cannot and is not being done by taking racing times of the horses.

[cj]

This was my point in asking the question originally about Cramer "redoing" the ratings. In addition, to the valid point CJ made, consider a horse that had an inferior figure going in and wins the race with an improved figure . The subsequent adjustment may make the horse the choice because of the improved rating. How it affects the variant process is anyone's guess. It is no surprise to me that Cramer variant and race adjusted speed ratings produce a higher percentage of winners. In fact, I would think that considerably more horses do run an improved number to win than run an inferior one . The subsequent adjustment to the ratings insures that the new speed ratings have a higher win percentage.

Yes, I figured it was. I think it is just as likely a big figure winner steps up and loses and has his figure reduced using that method, but it doesn't really matter. Both cases are not only possible, but likely.

[Greyfox]

I will now answer your question: “Is a variant based on reality or simply number crunching without merit and most likely wrong?”
The variant in my opinion is based on simply number crunching without merit and most likely wrong.

.

If we were talking about calculating how to go to the moon, I'd say you are on a more accurate trajectory.

In the meanwhile, if a "dark and dirty number calculated by simple arithmetic" number increases my chances of picking today's winner, then there is definitely merit in it and while not technically totally accurate it has merit.

[barn32]

I personally believe track variants deter not aid speed handicappers.
I wonder what Tom Brohamer would say about that.

[Cratos]

If we were talking about calculating how to go to the moon, I'd say you are on a more accurate trajectory.

In the meanwhile, if a "dark and dirty number calculated by simple arithmetic" number increases my chances of picking today's winner, then there is definitely merit in it and while not technically totally accurate it has merit.

One of the reason this sport is dying is because it fails immensely to keep up with technology.
I suppose you believe that data mining and sablemetrics are also ways to the moon

The point of my post is either do it right or don't do it at all and there are some people who do my way and make a very nice profit.

[raybo]

Isn't the bottom line, regarding using or not using a variant, whether or not it improves your play?

Don't the majority of players use mass produced figures, from 1 of several commercial sources, without further adjustments?

If that is the case, then, shouldn't you be doing something different?

If that is true, then why not use something like variants to further adjust whatever pace/speed/velocity figure you prefer, and shouldn't you be using those adjustments as a means of arriving at a method that does not mirror what most players do?

What does it matter if the variants, or whatever you decide to use for further adjustments, are based in reality or not, as long as it moves you either closer to profitability or towards better profitability?

And, shouldn't these adjustments be dynamic, changing to reflect a new view of what is happening in reality?

[TexasDolly]

Yes, I figured it was. I think it is just as likely a big figure winner steps up and loses and has his figure reduced using that method, but it doesn't really matter. Both cases are not only possible, but likely.
They are both likely and probably constitute the majority of the races since there are fewer races that are won with the original rating . The interesting thing is that both of the cases will serve to enhance the " new variant and new speed rating". One by lowering the original rating( the losing case ) and the other by enhancing it (the winning case.)

[Greyfox]

The point of my post is either do it right or don't do it at all and there are some people who do my way and make a very nice profit.

Even with your high tech knowledge, my guess is some Einstein somewhere would say your approach is wrong - "Do it right or don't do it at all."

And there are some people who do it my simple way and make a reasonable profit too.

I can appreciate some of you brighter mathematical minds try to turn this into astrophysics. Good for you. But in reality, it isn't.

Anything that increases my probability of winning is fine by me.

[cj]

One of the reason this sport is dying is because it fails immensely to keep up with technology.
I suppose you believe that data mining and sablemetrics are also ways to the moon

The point of my post is either do it right or don't do it at all and there are some people who do my way and make a very nice profit.

The point of all your posts is to be an asshole. All you ever do is try to find flaw with everything others do, while offering nothing yourself in return, similar to the guy that started this thread.

[Magister Ludi]

The point of all your posts is to be an asshole. All you ever do is try to find flaw with everything others do, while offering nothing yourself in return, similar to the guy that started this thread.

I've learned FAR more from "the guy that started this thread" than the combined contributions of everyone else on this site.

[cj]

I've learned FAR more from "the guy that started this thread" than the combined contributions of everyone else on this site.

Well, I'm happy for you, but I must have missed those. I know he started some secret fraternity for handicappers, but I think that fizzled out pretty quickly. Regardless, his attitude in this thread shows exactly what I'm talking about.

[DeltaLover]

I've learned FAR more from "the guy that started this thread" than the combined contributions of everyone else on this site.

This makes two of us...

The guy who started this thread is among the top contributors in this site and I am confused by the comments and the animosity.

[cj]

This makes two of us...

The guy who started this thread is among the top contributors in this site and I am confused by the comments and the animosity.

Maybe I've been unfair to TriM. I haven't read much of his stuff since the last time he started basically the same exact topic.

What are some things about horse racing you have learned from him?

[Greyfox]

This makes two of us...

The guy who started this thread is among the top contributors in this site and I am confused by the comments and the animosity.

And what specific threads could we go to for that info?

[PaceAdvantage]

Did not Ron Tiller basically dismantle the thread starter's premise in one fell swoop? He proved without a doubt that applying a variant improved the usefulness and accuracy of the speed figure.

I'm not even sure that cj's point about them going back and changing numbers even takes away from Ron's point.

[cj]

As for cratos' nonsense about turns, what makes him think that isn't factored into variants, and therefore figures also? Of course it is.

[cj]

Did not Ron Tiller basically dismantle the thread starter's premise in one fell swoop? He proved without a doubt that applying a variant improved the usefulness and accuracy of the speed figure.

I'm not even sure that cj's point about them going back and changing numbers even takes away from Ron's point.
Well, it probably does to a point, but the numbers he posted look very familiar so I suspect the changes are small and don't have much effect over a large sample.

[DeltaLover]

Maybe I've been unfair to TriM. I haven't read much of his stuff since the last time he started basically the same exact topic.

What are some things about horse racing you have learned from him?

For example:

http://www.paceadvantage.com/forum/showthread.php?t=92151

http://www.paceadvantage.com/forum/showthread.php?t=87406

http://www.paceadvantage.com/forum/showthread.php?t=85592

http://www.paceadvantage.com/forum/showthread.php?t=78576

I think these are among the most most interesting and provocative concepts posted in PA

[cj]

For example:

http://www.paceadvantage.com/forum/showthread.php?t=92151

http://www.paceadvantage.com/forum/showthread.php?t=87406

http://www.paceadvantage.com/forum/showthread.php?t=85592

http://www.paceadvantage.com/forum/showthread.php?t=78576

I think these are among the most most interesting and provocative concepts posted in PA

To each his own. I've already taken a few Stats classes. I'm more interested in horse racing and betting.

[Greyfox]

For example:

http://www.paceadvantage.com/forum/showthread.php?t=92151

http://www.paceadvantage.com/forum/showthread.php?t=87406

http://www.paceadvantage.com/forum/showthread.php?t=85592

http://www.paceadvantage.com/forum/showthread.php?t=78576

I think these are among the most most interesting and provocative concepts posted in PA

If you read those and learned from them great.

I read all of those threads at the time and honestly got diddly squat out of them.
I called them "deja mu" - the same old bull over again.

Tom called him a math troll.
Sure, it's fun to pick the math of horse racing apart.
But when challenged to handicap a race for the day, he offered zilch.

There comes a point where the rubber meets the road.
When it came to actually demonstrating any of that Bayesian gobbly gook in action, absolutely nothing was presented, zero, zil, nil, nothing.

[raybo]

I've learned FAR more from "the guy that started this thread" than the combined contributions of everyone else on this site.

Maybe you have more to learn than others here? But, probably, you're just another math/statistics guru who only participates when those things are discussed. The game is horse racing, or maybe you have mistakenly clicked on the wrong forum?

[Magister Ludi]

Maybe you have more to learn than others here? But, probably, you're just another math/statistics guru who only participates when those things are discussed. The game is horse racing, or maybe you have mistakenly clicked on the wrong forum?

The game that you play is horse racing. The work that I do is making money.

[raybo]

The game that you play is horse racing. The work that I do is making money.

Cute! :rolleyes:

[TexasDolly]

Did not Ron Tiller basically dismantle the thread starter's premise in one fell swoop? He proved without a doubt that applying a variant improved the usefulness and accuracy of the speed figure.

I'm not even sure that cj's point about them going back and changing numbers even takes away from Ron's point.

I originally raised the question in post #15 about Cramer redoing speed ratings
and Ron replied to that concern and I thought he said that the variants were constructed after the fact of the redo of the speed rating. Which makes their effect suspect to me.

Maybe others with large databases can post their findings using variants from the actual unadjusted speed ratings.
TD

[HUSKER55]

THAT WON'T HAPPEN until the testoterone level drops a few points :D

[RonTiller]

While there is certainly NOTHING wrong with adjusting figures later down the road to get the most accurate numbers possible, I would say doing this means printing Win percentages and ROIs on figures that may have very well changed after the horse has subsequently run are misleading.

For example, Horse A wins his debut and is given a 55. It is the top figure in the field next time he runs. After that race though, the figure is adjusted to 53 and is no longer the top figure in that race.

If you want to measure ROI and W%, you have to base it on the figures that were in the PPs at the time, not what they were later changed to reflect. Maybe you are doing that, but it isn't clear to me.
The specific question I was addressing was justifying incorporating variants into speed ratings versus not incorporating variants at all. I agree that if one asks about the performance of a published number (perhaps to compare various speed ratings), the answer should be the number published before the race and not subsequently recalculated numbers. I wasn't answering that question, though I see now that others may not distinguish those questions. Jim Cramer measures the performance of his speed ratings based on the current state of the database and that is what I reported in answer to the thread's question. We agree that these numbers are not the same as the numbers published the day before the race, but when Jim improves the win % of the numbers in the current state of the database, the published numbers demonstrably improve as well.

The last race speed ratings, published in our handicapping data files the day before each race in question are this: #1 last race Cramer Speed Rating win % = 28.25%. None of this is a secret. There are lots of people, including many in this forum, with the historical data and tools who can answer that same question.

This does NOT mean that WITH variant, the win % is 28.25% (the published number) and WITHOUT variant, it is 26.5%, because that 26.5% represents the raw numbers as they are now, not the raw numbers corresponding to the published numbers. Jim informs me that the relationship between speed ratings with and without variant remains about the same whether taking a snapshot of the current state of the numbers or the published state of the numbers. That is the question I was answering so my original answer still stands.

Although variant is in fact important, the track to track is much more important to the final speed rating.

Ron Tiller
HDW

[traynor]

This does NOT mean that WITH variant, the win % is 28.25% (the published number) and WITHOUT variant, it is 26.5%, because that 26.5% represents the raw numbers as they are now, not the raw numbers corresponding to the published numbers. Jim informs me that the relationship between speed ratings with and without variant remains about the same whether taking a snapshot of the current state of the numbers or the published state of the numbers. That is the question I was answering so my original answer still stands.

Ron Tiller
HDW

That is why business analysts do so well handicapping thoroughbreds. They continually ask, "Does this stuff really mean what we think it means?"

Values expressed in hundredths of a percent imply detailed accuracy. The term "about the same " implies something else.

[TexasDolly]

The specific question I was addressing was justifying incorporating variants into speed ratings versus not incorporating variants at all. I agree that if one asks about the performance of a published number (perhaps to compare various speed ratings), the answer should be the number published before the race and not subsequently recalculated numbers. I wasn't answering that question, though I see now that others may not distinguish those questions. Jim Cramer measures the performance of his speed ratings based on the current state of the database and that is what I reported in answer to the thread's question. We agree that these numbers are not the same as the numbers published the day before the race, but when Jim improves the win % of the numbers in the current state of the database, the published numbers demonstrably improve as well.

The last race speed ratings, published in our handicapping data files the day before each race in question are this: #1 last race Cramer Speed Rating win % = 28.25%. None of this is a secret. There are lots of people, including many in this forum, with the historical data and tools who can answer that same question.

This does NOT mean that WITH variant, the win % is 28.25% (the published number) and WITHOUT variant, it is 26.5%, because that 26.5% represents the raw numbers as they are now, not the raw numbers corresponding to the published numbers. Jim informs me that the relationship between speed ratings with and without variant remains about the same whether taking a snapshot of the current state of the numbers or the published state of the numbers. That is the question I was answering so my original answer still stands.

Although variant is in fact important, the track to track is much more important to the final speed rating.

Ron Tiller
HDW

Ron, I guess you are saying that the published numbers with and without a variant have roughly the same improvement as the current numbers demonstrate with and without a variant. What is the pct. winners then without the variant applied . As I understand it the pct. is 28.5 with the variant(which I assume was done before the fact.)
Ron, thanks for taking the time to address these questions.
TD

[Saratoga_Mike]

Well, I'm happy for you, but I must have missed those. I know he started some secret fraternity for handicappers, but I think that fizzled out pretty quickly. Regardless, his attitude in this thread shows exactly what I'm talking about.

Your take on both posters is exactly right - thought I was the only one who noticed it.

[RonTiller]

That is why business analysts do so well handicapping thoroughbreds. They continually ask, "Does this stuff really mean what we think it means?"

Values expressed in hundredths of a percent imply detailed accuracy. The term "about the same " implies something else.
I don't know what THAT response means. For the current state of the database, the numbers are as stated. "about the same" in this case means that we do not save the exact data that would PRECISELY answer the related question for the time period in question, so it would be irresponsible for me to say they are identical, even though 20 years of research has shown in the past that they are. Rats, we just don't save data with an eye to defending statements I make on PaceAdvantage. I agree though that my statement looks odd so I'll take the rebuke.

Ron, I guess you are saying that the published numbers with and without a variant have roughly the same improvement as the current numbers demonstrate with and without a variant. What is the pct. winners then without the variant applied . As I understand it the pct. is 28.5 with the variant(which I assume was done before the fact.)
28.25% is the win % of #1 last race speed ratings that we put in the files we make the day before the race, the same files our customers download and handicap with.

The current set of numbers shows a 3.5% difference between numbers with and without variants, 30% versus 26.5%. If one uses the numbers published before the race, in our data files, with variants, it is 28.25%. Because we don't store the base number, without variant, that the final speed number is built from, I cannot give a precise win % for speed ratings as we put in data files, only without variant. That was the meaning of my "about the same" comment that Traynor jumped on. Jim has measured the differences between published and current speed numbers in a thousand different ways over 20 years and says unequivocally that the same difference holds with the published numbers, i.e. 3.5%. But just for you Traynor, lets make it 3.5028594783% :)

Concerning whether variants improve performance, I've said all I have to say. The case is closed at HDW but others can continue the debate.

Concerning variants being recalculated, I don't want to get into an endless discussion. We have discussed it here and in other forums many times before. However, Jim wanted to provide this example of how a set of numbers might evolve:

Day 1 race 1 Track A

horse A projected 100 ran 90
horse b projected 95 ran 85
horse c projected 80 ran 70
horse d projected 75 ran 65

Day 1 race 2 Track A

horse A projected 90 ran 100
horse b projected 85 ran 95
horse c projected 70 ran 80
horse d projected 65 ran 75

Day 1 race 3 Track A

horse A projected 95 ran 90
horse b projected 85 ran 80
horse c projected 75 ran 70
horse d projected 65 ran 60


At the end of race 1, it appears that horse A ran 90 (original number = 90)
At the end of race 2, it appears that horse A in race 1 did not run 90 (original number = 90) but 100
At the end of race 3, it appears that horse A in race 1 did not run 100 but 95
And after races 4, 5, 6, 7, 8 and 9 it may change 4 or 5 more times.

If horse A does not run again for 2 months and 50 races since day 1 race 1 provides more data, the more accurate number for horse A could be 85, and will be changed at HDW.

I hope all our competition keeps their original numbers.

Ron Tiller
HDW

[eurocapper]

The current set of numbers shows a 3.5% difference between numbers with and without variants, 30% versus 26.5%.

Is it possible this is caused by mud or other extreme weather conditions? (The original question referred to non-obvious debatable track variants I think.)

[jeebus1083]

My stance may have changed over the years, but here is my take: All tracks react differently, and all horses react differently to different surfaces. Boiling the resiliency of a track to a single number or adjective ("fast" or "slow") requires both skill and intuition. Some are more skilled and/or profitable at it than others. Par tables are useful in "getting an idea" of how a race "should" have played, but they are not the end all. An observant horseplayer that follows the ABC's of a track on an everyday basis will have a better handle on this than someone relying solely on a par table, because he will base the races on today's reality rather than last year's information.

So yes, I believe that track variants do exist. However, they are only as cut and dry as each individual makes them out to be.

[traynor]

I don't know what THAT response means. For the current state of the database, the numbers are as stated. "about the same" in this case means that we do not save the exact data that would PRECISELY answer the related question for the time period in question, so it would be irresponsible for me to say they are identical, even though 20 years of research has shown in the past that they are. Rats, we just don't save data with an eye to defending statements I make on PaceAdvantage. I agree though that my statement looks odd so I'll take the rebuke.

Ron Tiller
HDW

No rebuke was intended. Nor was any criticism of the material in question intended. I meant only to point out that people who invest on the basis of their judgment calls would do well to think critically about every detail of the data or information used as a basis for their decisions.

[TexasDolly]

Thanks Ron for elaborating on the HDW technique.In your scheme using after the fact adjusted speed ratings and after the fact adjusted variants, the variants represent a plus of 3+ pct.. I guess I am still curious,like others I'm sure, as to the conclusion others have reached when the data
is more basically handled . I guess I just restated the original question.
TD

[RonTiller]

I'd just like to clarify one thing before signing off this thread.

Pars have been mentioned on several occasions in this thread, but as CJ pointed out, pars are not commonly used by professional speed figure makers (he said competent but I'll say professional). HDW does not use pars to make variants or in track to track.

This is kind of a dumbed down sequence of events (I mean dumbed down for me):

1. Project a time for every horse in the race (not just the winner, as some have inaccurately suggested in other contexts). This itself could constitute a lifetime of work. The better the projection, the better the variants, the better the speed number. As I stated, some projections will have lots of highly predictive data points, some virtually none, some a rather iffy bag.

2. Compare those individual projections against what really happened. Combine those individual projections versus reality into a variant for the race. Many many ways to do this. Lots of trial and error, testing and retesting.

3. Now each race has a variant - what to do? Did it start raining after race 1 but race 1 was mostly 1st time starters? What if race 8 was a mile and 7/16th, run once every 2 years? It is startling (and depressing) how many what ifs there are that can alter the landscape, depending on how one decides to handle them. One can split variants, assign a myriad of weighting schemes, calculate averages, medians, standard deviations, apply variant for each race, for combinations of races, etc. There are a seemingly endless list of possibilities to test when trying to optimize the "what do I do with all these individual race variants now" phase. At some point, "a" variant is applied to each race's base speed rating (not necessarily the same variant). By base speed rating, I mean the speed with all the other stuff already incorporated, like whatever speed points per length scheme is used, whatever track to track, whatever other factors go into the speed rating, excluding the variant.

4. For a set period of days (I don't know the current value) variants for those same races will be recalculated, using projections that incorporate races run AFTER the race. Those are more data points to get the projections for those earlier races a little better. That CAN result in a different variant for that race, due to the new projection, which CAN affect the speed ratings for other races on that same card, depending on how individual race variants propagate throughout a race card.

Imagine a race originally had a calculated variant of +15 (15 slow) but several weeks later, with new projections, it turns out the horses in the race had been originally projected too high. With a new projection in hand, the race variant might now be +2. If a weighting scheme or averaging is used to make a variant for the day or a cluster of races for the day, to apply to races, this could change EVERY horses speed number for the day.

I should point out that Jim originally started doing this because testing showed the performance improved. And I'm not saying its not tricky and full of potential minefields. You gotta know what you are doing and as Traynor said in another context, you must think critically about every detail

I don't say any of this to start a new debate or to be provocative in any way. Enough people seemed genuinely curious that I thought I'd share. For those who insist that the variants, and hence the speed ratings calculated immediately after the race card is done should be written in stone and never changed, we have no quarrel. You gotta do what you gotta do.

One final thought. Anybody who takes the time to make figures, whether professionally or for private use, whether top notch or lousy, deserves some respect, if for no other reason than they have the patience and fortitude to show up every day and keep working at a thankless, never ending task. (It is of course possible that there is a whole community of lobotomized former figure makes in Tahiti, betting on which coconut will fall next).

Ron Tiller
HDW

[Greyfox]

One final thought. Anybody who takes the time to make figures, whether professionally or for private use, whether top notch or lousy, deserves some respect, if for no other reason than they have the patience and fortitude to show up every day and keep working at a thankless, never ending task.

Ron Tiller
HDW

Well said and thanks for your explanations here Ron. :ThmbUp:

[barn32]

Close itI'll leave it open, just won't let you participate.Wow.

You assume he won't answer, and then based on that you won't even give him a chance to respond in case he would like to after he's had a chance to evaluate other responses.

Wow.

[raybo]

Wow.

You assume he won't answer, and then based on that you won't even give him a chance to respond in case he would like to after he's had a chance to evaluate other responses.

Wow.

Some people learn from history, others don't but keep making the same errors over and over.

[Greyfox]

Some people learn from history, others don't but keep making the same errors over and over.

Exactly. :ThmbUp:

Tri M's response was "Close it."
There was no indication in that reply that he was going to substantiate his original pooh-poohing of variant makers and users.

[cj]

Wow.

You assume he won't answer, and then based on that you won't even give him a chance to respond in case he would like to after he's had a chance to evaluate other responses.

Wow.

Assume? He said close it. He has done this before. It isn't our first rodeo.

[jeebus1083]

Well said, Ron Tiller. One of the most intelligent posts on speed handicapping that I've read on this board.

[Buchan]

Pars have been mentioned on several occasions in this thread, but as CJ pointed out, pars are not commonly used by professional speed figure makers (he said competent but I'll say professional).

well i would say you are both wrong then.
i don't know what a "professional" speed figure maker is, but i know that entropy's group and others use pars extensively.
how else can you quickly figure class and many other things?
although what you guys,and us, define as "pars" may be two different things.

don't place much store in variants though, as that is just what is left over after the calculations are done.
it's just an indication of how the weather and other stuff affected things, but of not much use in itself imo.

[098poi]

The point of all your posts is to be an asshole. All you ever do is try to find flaw with everything others do, while offering nothing yourself in return, similar to the guy that started this thread.

CJ I must say on the one hand I laughed out loud when I saw this on the other I think that kind of statement is unnecessary. I enjoy Cratos's posts even if I don't understand them sometimes. Just my opinion.

Now a question for you and Ron Tiller (who should definitely keep posting).
What are projections for what a horse should run based on if not par times? Is it based on what the horse has run in the past only? I thought that there was a basic concept that $20,000 claimers or NW1X at track X run a certain time and then after the race the difference between the winner and what was expected was analyzed. I thought what was expected or projected was based on a par time of that condition at that track. Thanks

[TrifectaMike]

The point of all your posts is to be an asshole. All you ever do is try to find flaw with everything others do, while offering nothing yourself in return, similar to the guy that started this thread.

CJ,

You are so right. Therefore, I'll stop posting on PA. The pay was is lousy, I quit.

Mike (Dr Beav)

[Capper Al]

Thanks RonTiller for the info. The variant like speed figures fluctuate, they help more than not as your stats proved.

[cj]

CJ I must say on the one hand I laughed out loud when I saw this on the other I think that kind of statement is unnecessary. I enjoy Cratos's posts even if I don't understand them sometimes. Just my opinion.

Now a question for you and Ron Tiller (who should definitely keep posting).
What are projections for what a horse should run based on if not par times? Is it based on what the horse has run in the past only? I thought that there was a basic concept that $20,000 claimers or NW1X at track X run a certain time and then after the race the difference between the winner and what was expected was analyzed. I thought what was expected or projected was based on a par time of that condition at that track. Thanks

It was just my opinion as well, and I've never really been one to hold mine back even when I should.

For me, projections are based on what the horse has run in the past, but also with lightly raced horses I do project improvement based on a huge database and how those horses have improved in the past. The quickest way to get "shrinking figures" is to underestimate the improvement of horses winning races.

Using pars was probably a decent start many years ago. But, in 2012, it could take a few decades to compile pars with all the various classes, conditions, statebred races, etc.

[PaceAdvantage]

CJ I must say on the one hand I laughed out loud when I saw this on the other I think that kind of statement is unnecessary. I enjoy Cratos's posts even if I don't understand them sometimes. Just my opinion.cj's reply, while colorful to the extreme, brings home the point that there are some on here who enjoy sitting in their ivory tower, looking down their nose at the unwashed masses. They will point out perceived "flaws" and tell you (in their subtle way) that you are an idiot, and really offer nothing more than that.

cj could have worded his reply with a bit more tact, but then again, you would expect those sitting in ivory towers to have thicker skin...especially if they are as well versed in this endeavor as claimed.

[cj]

cj's reply, while colorful to the extreme, brings home the point that there are some on here who enjoy sitting in their ivory tower, looking down their nose at the unwashed masses. They will point out perceived "flaws" and tell you (in their subtle way) that you are an idiot, and really offer nothing more than that.

cj could have worded his reply with a bit more tact, but then again, you would expect those sitting in ivory towers to have thicker skin...especially if they are as well versed in this endeavor as claimed.

I heard that word on NYPD Blue, so I knew it was ok. I can't believe that word would make someone leave. Thank goodness for them they didn't grow up where I did.

[RonTiller]

HDW makes pars, updated every day incorporating yesterday's races. Many of our customers and some of our developers use pars effectively in their handicapping and in their handicapping programs.

This thread was about speed ratings and track variants, and HDW does not use pars at all in calculating speed ratings. Our projection method uses horses in the race, not horses in the barn.

Regarding several questions about how our projections are done, I'm afraid I cannot elaborate, as that is extremely complex and proprietary.

Ron Tiller
HDW

[Cratos]

cj's reply, while colorful to the extreme, brings home the point that there are some on here who enjoy sitting in their ivory tower, looking down their nose at the unwashed masses. They will point out perceived "flaws" and tell you (in their subtle way) that you are an idiot, and really offer nothing more than that.

cj could have worded his reply with a bit more tact, but then again, you would expect those sitting in ivory towers to have thicker skin...especially if they are as well versed in this endeavor as claimed.


Pace, I had overlooked the obnoxious comment made to me by CJ, but the your post with the incredulous comment about that offensive remark which CJ’s made was both egregious and disingenuous to say the least.

CJ’s comment was insulting and without any value or pertinent to the thread subject matter at all.

However you are the administrator and can allow whatever comments that you deem fit, but to say “while colorful to the extreme, brings home the point that there are some on here who enjoy sitting in their ivory tower” makes me wonder if intellectual discourse is acceptable on this forum.

I don’t understand the “ivory tower” comment because what I know was taught to me and others very well at some very good academic institutions. Additionally, I go back over 40 years in thoroughbred racing and handicapping.

Disagreement is inherent to debate and we will never all think alike because if we did we would think the same and where would that get us.

[Saratoga_Mike]

I don’t understand the “ivory tower” comment because what I know was taught to me and others very well at some very good academic institutions. Additionally, I go back over 40 years in thoroughbred racing and handicapping.



You just made his point w/o even knowing it.

[Tom]

Here's the difference - CJ has posted many real life examples of what he talks about. He has scores of satisfied customers using the very stuff he posts about here.

I have yet to see anything of any substance come out of those ivy towers, by multiple residents.

This a game of action, not theory and talking.

Do you have any idea how utterly stupid some of these guys sound? I'm reading how speed figures are shit, blah blah blah, while watching races that I just won, using CJ numbers. I'm reading the post while waiting for the exacta payouts to be posted! :lol:

[Saratoga_Mike]

Yes Tom, but your winnings are made in such a crude fashion. Wouldn't you feel better if you had an MIT-approved method? You might not win, but you could feel better about yourself - intellectually superior.

[PaceAdvantage]

makes me wonder if intellectual discourse is acceptable on this forum.Don't you know? Intellectual discourse is completely outlawed here.

I don’t understand the “ivory tower” comment because what I know was taught to me and others very well at some very good academic institutions. Additionally, I go back over 40 years in thoroughbred racing and handicapping.Perhaps I'm confusing you with someone else, but when I think of you, I think of someone who is quick with the criticism...quick with the "your way is flawed and I won't hesitate to tell you so"...but not so quick with "here, let me explain a better way that I have found to go about things."

[PaceAdvantage]

Yes Tom, but your winnings are made in such a crude fashion. Wouldn't you feel better if you had an MIT-approved method? You might not win, but you could feel better about yourself - intellectually superior.There is no reason to put down theory. Or MIT.

I am fascinated with approaches like that, even if I don't understand them all that well. I would LOVE to be able to go at this game in a more scientific, more statistically oriented fashion. But I LACK both the time and the educational foundation to tackle such an endeavor at this time.

But that doesn't mean I don't enjoy reading discussions about them...

[Saratoga_Mike]

There is no reason to put down theory. Or MIT.

I am fascinated with approaches like that, even if I don't understand them all that well. I would LOVE to be able to go at this game in a more scientific, more statistically oriented fashion. But I like both the time and the educational foundation to tackle such an endeavor at this time.

But that doesn't mean I don't enjoy reading discussions about them...

I just like to see the practical application of theories.

[Cratos]

Don't you know? Intellectual discourse is completely outlawed here.

Perhaps I'm confusing you with someone else, but when I think of you, I think of someone who is quick with the criticism...quick with the "your way is flawed and I won't hesitate to tell you so"...but not so quick with "here, let me explain a better way that I have found to go about things."

Pace, I apologize for not responding sooner, but I am responding now to let you understand that I do have a different way to handicapping thoroughbred racing.

Therefore some associates and me have developed a parametric handicapping model. This parametric handicapping model is a collection of probability distributions such that each member of this collection, Pθ, is described by a finite-dimensional parameter θ.
.
For those of you who like statistics will understand that a parametric model or parametric family or finite-dimensional model is a family of distributions that can be described using a finite number of parameters These parameters are usually collected together to form a single k-dimensional parameter vector θ = (θ1, θ2, …, θk).

The objective of the parametric handicapping model is to predict the best horse in today’s race, under today’s conditions, at today’s distance, and at today’s racetrack by fair odds.

The output is measured in fair odds compared against the tote board odds.

This is not a cumbersome effort because the software does the work (analysis) for the bettor and then the bettor is left to make a waging decision along their risk curve.

The model to date has 8 modules and is based on six (6) Functional Variables (FV) and n-number of Bettor Defined Angle Variables (BDAV). The functional variables are: Pedigree, Class, Pace, Distance, Weight, and Style.

They are interdependent and have parent child relationships. For instance, distance is a child of stamina and final time is a child of pace. BDAVs can be almost anything that a bettor deem additive to his/her probabilistic chance to win. For instance it can be a jockey change or it might be an equipment change.

The model is not a speed figure or speed rating model because speed is only a single entity in a horse’s performance and is too one dimensional. Also there aren’t variants; the parametric handicapping model does include a SSRA (Surface Speed Resistance Adjustment) for the dirt track and a TYFA (Turf Yielding Factor Adjustment) for the turf course.

A modular overview in functional sequence is:

1. The Gamma Module – this module is used for data mining

2. The Alpha Module – this module is used for normalization

3. The Sigma Module – this module is used for weighted time integration

4. The Epsilon Module – this module is used for class as determined by race shape and race sustainability

5. The Theta Module – this module is used to estimate time of each horse in the race by feet/second, fractional of feet per second or final time. Additionally, it can convert all distances to metric.

6. The Zeta Module – this module brings the FVs together and assigns probabilistic weights. The total weight for each horse can never exceed .95

7. The Omega Module – this module converts the probabilistic weights into fair odds and juxtapose them in real time alongside the current tote board odds and calculate the difference.

8. The Delta Module – this module allows the bettor to assign BDAVs to each horse odds with respect to the bettor’s risk curve without violating the .95 constraint.

This program is design to handicap thoroughbred horses on any race surface in the world and at any racetrack configuration in the world because it has taken statistical physics as the discipline for its calculations.

[Red Knave]

Cratos, I never expected to see you outline, even at this high level, your methodology. Thanks.
I wonder if I could just ask you to clarify a few points.
The functional variables are: Pedigree, Class, Pace, Distance, Weight, and Style.When you say 'Style' is that a horse's running style?
For instance, distance is a child of stamina and final time is a child of pace. When you say 'distance is a child of stamina' am I to infer that stamina is embedded in another FV like Class? Does it come from Epsilon?
The model is not a speed figure or speed rating model because speed is only a single entity in a horse’s performance and is too one dimensional.You say this and yet your Sigma and Theta modules seem to integrate and/or generate times. Are these time forecasts and are they created from data obtained from other modules?
the parametric handicapping model does include a SSRA (Surface Speed Resistance Adjustment) for the dirt track and a TYFA (Turf Yielding Factor Adjustment) for the turf course.Do you calculate these adjustments from mined data (i.e. on the fly)?

Thanks in advance.

[cj]

Sounds like what a lot of bettors I know do, minus the fancy names. Maybe Cratos just can't recognize them when described in layman's terms.

[DeltaLover]

Although I am a firm believer that speed figures and track variant are overvalued I have to disagree with the following statement:


To calculate a surface variant there will be a need for some sophisticated devices and operated by very well train technicians. I say this because the most influential change to the race track static surface comes from the environment in terms of thermal, wind, and moisture. This cannot and is not being done by taking racing times of the horses.


I agree that the analytical calculation of a TV is not possible due to the very large number of parameters affecting it. In other words it is a chaotic environment and as such it is impossible to be described by a closed function.

This is exactly where the need for a model arises.

A model is not a precise description or reality. Quite the contrary. It represents an abstraction based in data reduction up to a point that it generates acceptable signals.

It is quite possible to implement such a model and many have already accomplished it. We can easily prove the validity of such a model using elementary statistics.

The point of debate should be not the possibility of generating a valid track variant and speed figures but the approach to filter the best methodology something that requires the answer to how speed figures should be used.

[Tom]

CJ does this everyday, every track, every race.
And people are making money off of his efforts.
Why don't you theorists start posting a better product, ie variant that produce figures that predict winners?

Since this thread started, how many variants have you guys calculated, how many races have you bet and won off of your efforts?
How many people line up to PAY you for your output?

Like I said before, this the real world.

[cj]

The point of debate should be not the possibility of generating a valid track variant and speed figures but the approach to filter the best methodology something that requires the answer to how speed figures should be used.

Of course, that is the big question, isn't it? I always say those that knock speed figures the most are the ones that don't know how to use them. Nobody I know that is successful thinks speed figures should be used in isolation. However, for some reason, those that are "against" speed figures seem to attribute that to all the so-called speed guys.

Anyone can read the analysis I've given for the last several Breeder's Cups and see that I don't just pick the best last speed figure. I use them as one of many tools, nothing more, nothing less.

[ronsmac]

The reality is, unless you know the wind for every single race, whether or not the track was watered and graded between races,the runup from the gate, unless you time from the gate itself, and the differences in temperature from race 1 to race 10, there really isn't one accurate variant for a race day. Of course the more accurate you try to be, you get caught up in a slew of numbers and it's much more work than it's worth. You're final variant or variants are still only going to be a semi educated guess anyway. Speed figures are so 1970's and 1980s that they offer nothing of value today, IMHO, or maybe not so humble opinion.

[Tom]

Opinions vary.

[raybo]

Opinions vary.

Correction. Opinions vary greatly!!

[Dan Montilion]

Track Variants - Bogus or real? Yes!

[Tom]

Well, that certainly answers the question.

[SPEEDHORSE]

This question is addressed to all projected variants figure makers.

I would like to know if projected speed figures should be based only on horses that delivers consistent speed figures in order that the projected track variants to have a semblance of accuracy as opposed to a projected speed figure from a horse who earn the figure only in its last race?

[cj]

This question is addressed to all projected variants figure makers.

I would like to know if projected speed figures should be based only on horses that delivers consistent speed figures in order that the projected track variants to have a semblance of accuracy as opposed to a projected speed figure from a horse who earn the figure only in its last race?

I include as many horses as possible at first (all but 1sters), but I eliminate the top and bottom 20% from any final calculations.

[raybo]

Shouldn't the "projected" portion of that post be "calculated". Does anyone really "project" variants? Not trying to be flip here, just a little confused by that post.

[sjk]

I project for horses who run in the top 3 or within 8 lengths of the winner and project from any race within 60 days (before or after) in which that horse was top 3 or within 8 lenghts of the winner.

I don't use races for 2YOs or 3YOs (except for a few days when CD and Crc insist on nothing else).

[raybo]

I project for horses who run in the top 3 or within 8 lengths of the winner and project from any race within 60 days (before or after) in which that horse was top 3 or within 8 lenghts of the winner.

I don't use races for 2YOs or 3YOs (except for a few days when CD and Crc insist on nothing else).

Are you projecting speed figures, or variants? How can you project a daily variant?

[sjk]

I get a line for each horse as in the previous post; use his speed rating in the last race in relation to his time in today's race to get a variant. When you look at all the variants for the day they usually have a std dev that is pretty consistent so I would take a look at it if things seemed unusual. Average all.

It is a little more complicated than that but you get the idea.

[thaskalos]

I get a line for each horse as in the previous post; use his speed rating in the last race in relation to his time in today's race to get a variant. When you look at all the variants for the day they usually have a std dev that is pretty consistent so I would take a look at it if things seemed unusual. Average all.

It is a little more complicated than that but you get the idea.

When there is a semblance of order, and all the variants fall into some sort of line, then I understand. But what happens when the variants are all over the place?

I find it hard to believe that the horses all improve or decline at approximately the same level. Horses improve and decline for a variety of reasons...not just because the track surface has changed. How do the projected variants account for all these things?

I understand that some variant is better than no variant at all...but we are playing a game where a fifth of a second can easily mean the difference between a win and a loss...so some kind of precision in still a must.

[sjk]

By and large they seem reasonably well grouped (std dev 11-13) the great majority of the time.

No great precision is expected from these variants. I throw a lot more information into making the figure usable when the horse runs again. The variant is just the starting point but I wouldn't know where else to start.

[SPEEDHORSE]

Bro Raybo,

What I mean when I posted Projected Track Variant is that it should really be called the Daily Track Variant. This daily variant was derived, as a result of the Projection Method in calculating variants, based on projected speed figures.

I hope I had corrected myself from the confusion I have made in my earlier post.

[raybo]

Bro Raybo,

What I mean when I posted Projected Track Variant is that it should really be called the Daily Track Variant. This daily variant was derived, as a result of the Projection Method in calculating variants, based on projected speed figures.

I hope I had corrected myself from the confusion I have made in my earlier post.

Still a little confused about "projecting" variants. I get the part about projecting speed figures, but the daily variant projections seem strange to me, just don't see how it is possible. Are we assuming that today's variant will be the same as some day in the past?

"Projecting" to me means "predicting", by the way, maybe that's the confusion?

[sjk]

We use other races today's horses have run to project times and then the difference between the projected times and actual times are used to compute a variant.

[raybo]

We use other races today's horses have run to project times and then the difference between the projected times and actual times are used to compute a variant.

Ok, so I assume the "standard deviation" you mentioned earlier, helps correct accumulated errors from the last, regarding those past daily variants? It appears to me that one relatively large miscalculation, from the past, could accumulate, along with lesser miscalculations, to add up to quite a problem, thus the need for the use of some sort of acceptable deviation as corrections in future variant calculations.

[sjk]

I look at the standard deviation to see if there was a mis-time or to see if I should consider splitting the races (one vs two turn or first part of the card vs the rest).

In practice I rarely split the variant. As I mentioned earlier I am going to make further adjustments to the individual speed figures before I use them.

Contrary to what you might think it becomes a very stable process. I started with a handful of speed figures to get the process started in 1993 and they are working fine today.

[classhandicapper]

When there is a semblance of order, and all the variants fall into some sort of line, then I understand. But what happens when the variants are all over the place?

I understand that some variant is better than no variant at all...but we are playing a game where a fifth of a second can easily mean the difference between a win and a loss...so some kind of precision in still a must.

The dirty little secret is that if you were to take the speed figures of all the leading figure makers and line them up (adjusted for methodology issues like weight and ground loss) there would be a ton of small differences, a lot of large differences, but very little difference in results.

They all make mistakes. So it comes down to methodology preferences. What do you want in your figures?

In addition, speed figures are not all inclusive in measuring ability.

IMO the primary use of figures is to answer questions about races/horses when there's a lack of other information to go by.

In other words, despite my own mistakes, I think I can evaluate horses/performances more accurately using classing techniques when we are talking about consistent seasoned stakes caliber older horses because I have a lot of information to work with.

However, if we are talking about a maiden claimer with several dropdowns, first time starters, 2nd time starters, blinker changes etc... the range of possibilities becomes so wide, my errors will often be a lot bigger than the errors in the figures. So I really want to know whether the race was fast or slow.

IMO, where handicappers get into silly territory is when they start arguing that one horse is clearly better than another because of a small difference in figures.

[traynor]

IMO, where handicappers get into silly territory is when they start arguing that one horse is clearly better than another because of a small difference in figures.

Followed by that most profound of passive voice observations, "It can be statistically proven that the 7 should have won that race. The fact that it finished dead last is statistically irrelevant."

[thaskalos]

The dirty little secret is that if you were to take the speed figures of all the leading figure makers and line them up (adjusted for methodology issues like weight and ground loss) there would be a ton of small differences, a lot of large differences, but very little difference in results.

They all make mistakes. So it comes down to methodology preferences. What do you want in your figures?

In addition, speed figures are not all inclusive in measuring ability.

IMO the primary use of figures is to answer questions about races/horses when there's a lack of other information to go by.

In other words, despite my own mistakes, I think I can evaluate horses/performances more accurately using classing techniques when we are talking about consistent seasoned stakes caliber older horses because I have a lot of information to work with.

However, if we are talking about a maiden claimer with several dropdowns, first time starters, 2nd time starters, blinker changes etc... the range of possibilities becomes so wide, my errors will often be a lot bigger than the errors in the figures. So I really want to know whether the race was fast or slow.

IMO, where handicappers get into silly territory is when they start arguing that one horse is clearly better than another because of a small difference in figures.
I consider speed and pace figures to be a vital component of the handicapping process...and I would be lost without them. This doesn't mean that I rely on figures alone in my handicapping...but accurate figures help me in a big way.

The problem with the "methodologies" of some of the figure makers out there is that we don't always know what else is included in their figures other than what they widely report.

Beyer claims that his figures are "pure" speed ratings...and yet, my own opinion is that there is a class and a pace component involved.

The Sheets supposedly account for pace in some way...but we don't know how exactly they go about it.

Give us the accurate "pure" speed figures, I say...and we can make the rest of the adjustments ourselves.

[eurocapper]

Is not every speed figure a product of the pace scenario in the given race? Maybe pace is really what matters in horse racing (except the shorter QH where there isn't much pace).

[raybo]

Is not every speed figure a product of the pace scenario in the given race? Maybe pace is really what matters in horse racing (except the shorter QH where there isn't much pace).

IMO, a pure speed figure incorporates a representation of the final time, adjusted by a daily variant, and, ideally, a good track to track adjustment, so that it is equalized over different tracks.

So, IMO, if you incorporate pace, it is no longer a pure speed figure, it becomes a "performance" figure.

And, again IMO, pace is indeed, really what matters. Class enters the picture there too, of course.

[classhandicapper]

Give us the accurate "pure" speed figures, I say...and we can make the rest of the adjustments ourselves.

That's the way I feel also, but that's probably because we are very experienced handicappers.

Sometimes a figure maker has the dilemma of knowing that a race came up really slow because of an extreme pace or other trip issue. If he gives customers the slow figure, many will not understand the pace and trip issues and think his figures just suck. So he'll make the adjustment for them knowing full well guys like us would prefer to make judgments like that on our own.

That's why I have always preferred CJ's method. He creates a performance figure that includes pace, but he also gives the customer the actual speed figure.

I don't think Beyer includes "Class", but he does break races out. So if a race comes up too fast or slow compared the horses' typical performances, he'll make arbitrary adjustments to make the figure fit.

[Cratos]

Is not every speed figure a product of the pace scenario in the given race? Maybe pace is really what matters in horse racing (except the shorter QH where there isn't much pace).

Go to the head of the class, you get an "A"