3 Year Best Times 2017-2020

[classhandicapper]

Beyer eventually dismissed using 10k claimers as a way of getting the tracks in sync because he found that 10k claiming horses were NOT equal from track to track. I always found that odd because you'd think it would possible for some smart group to identify the strongest 10k horses in the country, claim those, and then move them to weaker tracks where they'd have a edge without much fear of getting claimed. But I guess the difference in purse sizes and costs to maintain horses may have something to do with the quality differences.

Now I think Beyer just uses a DRF internal system to track every shipper from track to track to make sure the figures are in sync from track to track.

[classhandicapper]

Originally Posted by Tom

I use those link myself, for the speed figs, they are fine, but understand, some tracks will have a different values for a certain distance, like Aqu, at 6.5 f, I think it is like 6 pts off, but I would have to check.

Here is the latest I have for BL, but it was last update in July 2019.
I update every year, so it is almost time - I will try to get to it soon, as long as someone is interested. As more tracks open it will easier to do - I need PPS to do it.

A few distances have several values for a BL - this is because DRF PPs had a lot of really out of wack BL values. Bit it is close enough for government work.

Nice chart.

I always found varying the value a beaten length from surface to surface by so much to be a bit unsettling even though the races develop differently. If you do that, you are no longer making speed figures. You are making performance figures. Not that there's anything wrong with trying to create numbers that more accurately reflect ability/performance. They just aren't final time speed figures.

[steveb]

Originally Posted by classhandicapper

Nice chart.

I always found varying the value a beaten length from surface to surface by so much to be a bit unsettling even though the races develop differently. If you do that, you are no longer making speed figures. You are making performance figures. Not that there's anything wrong with trying to create numbers that more accurately reflect ability/performance. They just aren't final time speed figures.

don't know what beyer does now, but you should absolutely use NO constant.


length is an outdated measure, but nevertheless, a variable number is correct.

if a horse is beaten .5 sec in 68 sec race, then that deserves bigger penalty than beaten .5 secs in 73 sec race, all else being equal.

speed is all about proportion.

[MJC922]

Originally Posted by classhandicapper

Beyer eventually dismissed using 10k claimers as a way of getting the tracks in sync because he found that 10k claiming horses were NOT equal from track to track. I always found that odd because you'd think it would possible for some smart group to identify the strongest 10k horses in the country, claim those, and then move them to weaker tracks where they'd have a edge without much fear of getting claimed. But I guess the difference in purse sizes and costs to maintain horses may have something to do with the quality differences.

Now I think Beyer just uses a DRF internal system to track every shipper from track to track to make sure the figures are in sync from track to track.

Once you have good speed charts and decent variants then you can go back and revisit the pars you used in the first place to create the speed charts because maybe those class levels aren't actually a fifth of a second apart at six furlongs etc, maybe they're two fifths apart, and really what is that filly and mare adjustment, what are the true age adjustments. Maybe 10k at one track isn't 10k at another. These are all things we need to ask and not just accept on theory coming from books.

Which brings me to the next step. Iteration. You know how Nietzsche said we should let the superman be the meaning of the earth? in this case we should let iteration be the meaning of the earth. So how about going back and starting all over again but now with variant adjusted speed figures as a baseline you can refine all of the class pars and all of the adjustments that will allow you to construct better speed charts and better class pars in the first place. From those you'll make all new variants and new speed figures that are more accurate. So that's your first iteration. Unfortunately one iteration took a long time didn't it, weeks, months, maybe even years? When you look at something like my class ratings you're looking at output generated from hundreds of iterations every night.

Eventually you might say why even use class pars. If you have speed figures derived from thousands of iterations. Why not project a figure for each horse based upon its previous races and not just the race winners. Now you have sample sizes that start to become more robust, especially if track variants are your cup of tea, because track speed is always changing... or maybe not so much, lol. Though I suppose a variant is a handy crutch that can be used to adjust out of line races where final time isn't a good measure of performance after all.

Eventually you might ask why do I need the clock? You honestly don't, you need to iterate your way out of it.

[classhandicapper]

Originally Posted by steveb

don't know what beyer does now, but you should absolutely use NO constant.


length is an outdated measure, but nevertheless, a variable number is correct.

if a horse is beaten .5 sec in 68 sec race, then that deserves bigger penalty than beaten .5 secs in 73 sec race, all else being equal.

speed is all about proportion.

I agree with you are saying, but that's not what I was talking about or what figure makers are adjusting for here.

Beyer is essentially saying that 1/5th of a second is more significant on turf than dirt. Most people think he is correct. However, even though doing what he's doing makes for better figures, they aren't speed figures.

[classhandicapper]

Originally Posted by MJC922

So how about going back and starting all over again but now with variant adjusted speed figures as a baseline you can refine all of the class pars and all of the adjustments that will allow you to construct better speed charts and better class pars in the first place. From those you'll make all new variants and new speed figures that are more accurate.

Why not project a figure for each horse based upon its previous races and not just the race winners. Now you have sample sizes that start to become more robust, especially if track variants are your cup of tea, because track speed is always changing... or maybe not so much, lol.

The Beyer team does keep speed figure class PARs for all tracks. They are published by DRF.

I think most figure makers make projection speed figures that look at more than just the winner. That was the recommendation going all the way back to Beyer's original work. The only time Class Pars for winners are used is at the very beginning of the process.

[MJC922]

Originally Posted by classhandicapper

The Beyer team does keep speed figure class PARs for all tracks. They are published by DRF.

I think most figure makers make projection speed figures that look at more than just the winner. That was the recommendation going all the way back to Beyer's original work. The only time Class Pars for winners are used is at the very beginning of the process.

Most do of course move on to the projection method to make their daily variants. A projection is made from less accurate figures in the first place so it's essentially not even one iteration. The crude way to do it would be let's say you projected a speed figure of 80 for this race two months ago, with variant applied you gave it an 82. Now horses exit the race and given the evidence for horses running back you can see the race was probably more like a 76 than an 82. Unfortunately all of your projections since then are subsequently 'off' a certain amount as are the associated daily variants. So we really need to make that a 76 and do another iteration to refine projections/variants that are subsequently going to be based upon that, but nobody is going to do another iteration manually. I'm sure Beyer's team monitors certain races and make ad-hoc changes as needed but this isn't coming close to the kind of data processing I'm talking about.

[cj]

Originally Posted by MJC922

Most do of course move on to the projection method to make their daily variants. A projection is made from less accurate figures in the first place so it's essentially not even one iteration. The crude way to do it would be let's say you projected a speed figure of 80 for this race two months ago, with variant applied you gave it an 82. Now horses exit the race and given the evidence for horses running back you can see the race was probably more like a 76 than an 82. Unfortunately all of your projections since then are subsequently 'off' a certain amount as are the associated daily variants. So we really need to make that a 76 and do another iteration to refine projections/variants that are subsequently going to be based upon that, but nobody is going to do another iteration manually. I'm sure Beyer's team monitors certain races and make ad-hoc changes as needed but this isn't coming close to the kind of data processing I'm talking about.

This is something I actually wanted to do, but customers don't particularly like figures changing after they made bets on prior ones. I still do a lot of this in the background, but in the foreground the figures remain the same. For future projections, not always.

[classhandicapper]

Originally Posted by MJC922

Most do of course move on to the projection method to make their daily variants. A projection is made from less accurate figures in the first place so it's essentially not even one iteration. The crude way to do it would be let's say you projected a speed figure of 80 for this race two months ago, with variant applied you gave it an 82. Now horses exit the race and given the evidence for horses running back you can see the race was probably more like a 76 than an 82. Unfortunately all of your projections since then are subsequently 'off' a certain amount as are the associated daily variants. So we really need to make that a 76 and do another iteration to refine projections/variants that are subsequently going to be based upon that, but nobody is going to do another iteration manually. I'm sure Beyer's team monitors certain races and make ad-hoc changes as needed but this isn't coming close to the kind of data processing I'm talking about.

I tend to do what you are suggesting manually via Formulator (so do others). You can see all the Beyers for a race in the chart and then you can monitor the Beyers those same horses are running coming out the race very easily. However, except in certain circumstances, it gets tricky because horses change class, distance, surface, have different trips, are in different stages of their of form cycle etc.. next time. That makes it difficult to know whether the original figure should be changed or not unless you want to look up the details of every horse's subsequent start and get very subjective.

If the evidence is overwhelming or the figure was suspect to begin with, I'd be more apt to change it. Otherwise I'd leave it and move on, I'd also be more apt to change it if it was a maiden/alw race for very lightly raced horses or a race with lot of first time turfers. That where you find the most big errors to begin with because the original projection was often based on sketchy information.

[cj]

Originally Posted by classhandicapper

I tend to do what you are suggesting manually via Formulator (so do others). You can see all the Beyers for a race in the chart and then you can monitor the Beyers those same horses are running coming out the race very easily. However, except in certain circumstances, it gets tricky because horses change class, distance, surface, have different trips, are in different stages of their of form cycle etc.. next time. That makes it difficult to know whether the original figure should be changed or not unless you want to look up the details of every horse's subsequent start and get very subjective.

If the evidence is overwhelming or the figure was suspect to begin with, I'd be more apt to change it. Otherwise I'd leave it and move on, I'd also be more apt to change it if it was a maiden/alw race for very lightly raced horses or a race with lot of first time turfers. That where you find the most big errors to begin with because the original projection was often based on sketchy information.

It also doesn't help that it takes horses so long to run back anymore. It isn't all that rare to check a race from six weeks ago and few if any horses have run back.

[steveb]

Originally Posted by MJC922

Once you have good speed charts and decent variants then you can go back and revisit the pars you used in the first place to create the speed charts because maybe those class levels aren't actually a fifth of a second apart at six furlongs etc, maybe they're two fifths apart, and really what is that filly and mare adjustment, what are the true age adjustments. Maybe 10k at one track isn't 10k at another. These are all things we need to ask and not just accept on theory coming from books.

Which brings me to the next step. Iteration. You know how Nietzsche said we should let the superman be the meaning of the earth? in this case we should let iteration be the meaning of the earth. So how about going back and starting all over again but now with variant adjusted speed figures as a baseline you can refine all of the class pars and all of the adjustments that will allow you to construct better speed charts and better class pars in the first place. From those you'll make all new variants and new speed figures that are more accurate. So that's your first iteration. Unfortunately one iteration took a long time didn't it, weeks, months, maybe even years? When you look at something like my class ratings you're looking at output generated from hundreds of iterations every night.

Eventually you might say why even use class pars. If you have speed figures derived from thousands of iterations. Why not project a figure for each horse based upon its previous races and not just the race winners. Now you have sample sizes that start to become more robust, especially if track variants are your cup of tea, because track speed is always changing... or maybe not so much, lol. Though I suppose a variant is a handy crutch that can be used to adjust out of line races where final time isn't a good measure of performance after all.

Eventually you might ask why do I need the clock? You honestly don't, you need to iterate your way out of it.

like that do you think?

[steveb]

Originally Posted by classhandicapper

I agree with you are saying, but that's not what I was talking about or what figure makers are adjusting for here.

Beyer is essentially saying that 1/5th of a second is more significant on turf than dirt. Most people think he is correct. However, even though doing what he's doing makes for better figures, they aren't speed figures.

i would say beyer was wrong.
dirt generally more tiring than turf, thus what might equate to 1 second or maybe 6 lengths, beaten on turf, would be something more on dirt.

[cj]

Originally Posted by steveb

i would say beyer was wrong.
dirt generally more tiring than turf, thus what might equate to 1 second or maybe 6 lengths, beaten on turf, would be something more on dirt.

Being beaten a second or a length is a bigger loss in a shorter race than a longer race. And, for the most part, races on turf at the same distance are shorter timewise because they are almost always run faster. I'm sure you know this of course, but I don't really see a difference in what you said and what Beyer is doing.

[steveb]

Originally Posted by cj

Being beaten a second or a length is a bigger loss in a shorter race than a longer race. And, for the most part, races on turf at the same distance are shorter timewise because they are almost always run faster. I'm sure you know this of course, but I don't really see a difference in what you said and what Beyer is doing.

i was already wondering if i expressed it wrongly, as was looking at tom's
attachment towards start of thread.


i would only partly agree with your first sentence.
whereas it's certainly true, most seem to think the difference is bigger than what it actually is.


as a rough guide of my opinion...
if the race was run in 124 seconds and an also ran went 125 then that would be about 16 slower than winner.
if race was 58 secs and also ran was 59 seconds then that would be about 16.9 slower than winner.
that's if the respective race distances were 2000 metres and 1000 metres, otherwise numbers would vary slightly.


it's easy to prove that as fact, with simple math.


essentially all that beyers in their original form, are only expression of distance expressed in his numbers.
most, probably do not understand this(and most likely him either), but it is inarguable fact.


it's probably harder to see this fact if using lbs and furlongs/yards rather than metrics which is much more intuitive.
A beyer point is in reality equal to one metre at 1000 metres.
so if you had one beyer of 100 and another of 90 then that is in fact 10 metres at 1000m.

where the mistakes of beyer(originally....not sure of the now) was that that a difference of 10m at one distance would become twice that, at twice the distance.
it doesn't work like that.

[steveb]

mistake on my part
when i say it does not work like that, i only mean in the sense of beaten runners, not of winning time differences.