VARIANTS

[PaceAdvantage]

I've been meaning to start a topic like this for some time.

I'd like to know, particularly from the guys who make their own figs, why you think that the variants you come up with on a day to day basis are grounded in any sort of reality.

On a nine or 10 race card, comprised of a couple of sprints, a couple of routes, and a turf race or two, how can you possibly come up with accurate variants? Variants are supposed to measure the inherent speed of the racetrack, are they not? They are supposed to tell you how fast or slow the track was playing on any particular day.

But, how are we to assume the TRACK was the cause of the difference in "speed." Maybe it was the horses themselves?

Why are variants any different than the elusive TRACK BIAS people like to deride every now and then on this board. A lot of people are of the belief that track bias is a myth. These same people believe that in the rare instance a track bias TRULY exists, the track super will fix it so that the bias disappears as soon as it is detected.

In any event, tracking a BIAS, to me, seems easier than creating an ACCURATE variant, because a bias can be detected over MANY DAYS, while a variant has to be created based on ONE day's worth of races (which are further split into a handful of sprints, routes and turf races).

How do you speed fig guys reconcile this? Do you just come to the conclusion that any variant is better than no variant at all? Or, if your numbers enable you to handicap successfully, do you assume the variants you are producing are accurate?

I take it the speed fig makers here have experimented with creating figs using NO variant and found the resultant speed figures to be miserable failures, correct?

I guess the reason I started this thread is to see why variant making isn't considered in the same vein as "track bias hunting." A mostly futile effort based on a tiny sample size.

Thanks for any replies in advance.

==PA

[cj]

...Thanks for any replies in advance.

==PA

Kiss my ass!

[cj]

Sorry, couldn't resist.

I don't think you are only looking at 9 or 10 races. You are looking at 80 to 100 horses for the dirt races, maybe 20 for turf. Each horse can be viewed as a potential "variant" maker.

This is why I am supremely confident that the Beyer for Bellamy Road was inflated as I said within a couple days after the race. You have to look not only at the winners and the final time, you have to look at the internal fractions and the also rans.

As for unadjusted figures, I've tried them, they are called the DRF Speed Ratings, and they stink. And yes, I know any Tom, Dick, of Harry can go find some race where they outperformed Beyers and paid boxcars, but in the long haul, they still stink.

[Valuist]

There could be a real long response to this but I'll try to be concise. For me, I like to compare apples to apples, oranges to oranges etc. I am not a big fan of speed charts comparing 6f times to 1 mile times. I won't say they can't work but I think its a possible source for errors. A track like Oaklawn is great because 95% of their sprints are all at 6f. Tracks like Bel, CD or Arlington are tougher because they may run 6 dirt races at 5 different distances and 3 grass races all at different distances. All one can do then is project.

I think any time you see a fig that looks suspicious, go back thru the charts and see if its the only race at the distance that day. If it is, it could be a mistake.

Other sources for errors are changing track conditions or wind shifts. Litfin in the DRF Weekly charts is pretty good at documenting that stuff.

I do think one has to be flexible and willing to alter if they find they make a mistake. The DRF will make adjustments after the fact to the Beyers. The older a race in the pps, the more likely it is accurate.

[Turfday]

At the Southern California tracks, my take is that too many players jump to a fast conclusion about the existence of a bias. The "want to believe." But even if there is a bias, it usually doesn't last long enough, as the track super's are continually tinkering with the surface.

The only way to take advantage is to fully acknowledge to yourself beyond a shadow of a doubt that there were certain results on a given day that were primarily caused and solely influenced by a bias and to play on or against those horses coming back NEXT TIME.

MANY years ago, the Del Mar main track was totally biased favoring horses on the outside "rally wide types." It was totally attributable to the way the turns were banked. Sharp players made big money back in the day knowing that bias because they could virtually count out it every day.

When they re-did the track, they over compensated to where it was just the reverse....inside speed was king. And sharp players realized it at the time as well.

In recent years, it has leveled out.

As far as variants, it's also very subjective, just like much of using other handicapping factors.

[sjk]

When I look at the different estimates of the day's variant that come from the horses that ran, the values are almost always distributed in a fairly normal range and the std dev generally falls in a certain familiar range. This gives me the impression that the average should be a meaningful statistic.

The only real test of meaningfulness is whether the variants are a useful tool in forecasting the speed of the day's horses when they run again. That can be measured in a database by making forecasts for a large number of starters and seeing how accurate your forecasts turn out to be.

Since there are many elements that you could include in a forecast and many ways you could possibly combine them it is useful to look at a variety of methods, measure their effectiveness and use the best.

My personal experience tells me that one can make a good enough variant to be the first step towards successful handicapping.

[JustRalph]

I am fascinated by this pace capping stuff.......and the idea of variants.

I can do a bunch of things.........but this stuff seems hard for me. Why, I don't know. I can fix a gazillion computer problems, but when it comes to understanding this......... I get a little whacked.

Example:
See the pic below: The circled numbers are speed figures (BRIS) and next to them the DRF variant from that day. As you can see this horse had a distinct difference in variant on different days he raced (or is it that distinct? 12-18, is a 6 point variant, that much different?) I realize the last one was on turf.....but therein lies the rub for me. What am I really supposed to garner from a 12, or an 18 variant? I have read in some places that the higher the number, the better. I have been told by others that a higher number isn't that big a deal depending on what track it is run on. Anybody care to take a shot at explaining this to me? I am little thick headed when it comes to numbers...........but I feel like it shouldn't be this hard to understand. All things considered.


http://www.realworldtechsupport.com/variant.jpg

[kenwoodallpromos]

True, but just like pars, variants likie DRF's can be recorded over any number of days to see if trends develop. But who else does that?
And no one is stopping anyone else from analyzing fractions and paths taken as well as final times.

[cj]

JR,

The DRF published variant is about as worthless as the Speed Rating in the DRF. I would guess you are at least as well off using no variant at all as you are using that thing.

For the record, the higher the DRF variant, the slower the track was supposed to be that day.

This is from DRF:




Daily Racing Form's Speed Rating and Track Variant provide an "old style" gauge of a horse's speed in a race.

The Speed Rating is a comparison of a horse's final time with the best time at the distance at that track in the last three years. The best time is given a rating of 100. One point is deducted for each fifth of a second by which a horse fails to equal that time. Thus, in a race where the winner equals the best time (a Speed Rating of 100), another horse who is beaten 12 lengths gets a Speed Rating of 88 (100 minus 12). As a companion to the Speed Rating, Daily Racing Form's Track Variant takes into consideration all races run on a particular day under the same conditions of distance and track surface. The Speed Ratings of all winners in each type of race are added toget her and an average is computed. This average is deducted from the par of 100 and the difference is the Track Variant. (Example: if the average Speed Rating of winners sprinting on the main track is 86, the Track Variant is 14 (par of 100 minus 86). The lower the Track Variant, the faster the track, or the better the overall quality of competition that day.

[JustRalph]

Cj, enough said........basically I haven't been using it. I am doing ok without it. I have looked at coming up with my own variant and have had some others send me their work on variants. Either way, it doesn't change the way I play much. And from what I can tell, it takes a ton of time to do it right. I was wondering how many different opinions I would get on the drf variant. Looks like not many use it.........or at least they aren't willing to jump in here and talk about it.

[KingChas]

I do use them but in a very minor role.Since I cannot devote every day to
crunching numbers at every track I play.I use them as a final adjustment to my speed figs.I do not apply the DRF rules for adjusting.I add or deduct points from my ratings.Points not times.There is no exact science to this,kind of comes with experience.Basically pay attention to variants 10-09-08 etc to lower rating.Variants of to 30-31-32 etc to raise rating.Usually doesn't make major difference variants in between,minor adjs.Highest +/- I adjust is 7pts either way.
Of course this is pace and closing adj which could total a 14 pt difference.It really depends on the type of handicapper you are,I do agree at times they(DRF Variants are BS) ala this years prep races.I have been using the Racing Advance lately anyway-at least its available every day even Derby Day! ;)

In closing I do believe in a track bias,I also believe on a winning day my adjustments were correct and on a losing day they weren't.That can be said for Beyers, Sheets,$500 computer race programs, etc..............You still have to handicap.(Read between them lines) :)

[grahors]

King,
What are the DRF "rules for adjusting?
Sartin (I think) recomended using 17 as an average and subtracting or adding from that point.
I have Cynthia's "ave" variants for each track and have tried adj with those?????
Wm Scott has a chart in his books and uses a 1 for 1 adj over 17-18. and a varied adjustment for below 16?????
Who knows???
Like you said, ya gotta read between the lines after ya use the figs to get you into the ballpark.
Good handicapping.
Grahors

[KingChas]

Yes I did start the scale based on Wm. Scott.But over the years have made adjustments.The point of speed ratings is to see what horses are coming into form.Or throwouts.If I were to send you figs for a race card,and you just bet highest figs without looking at form.I might have a highest fig of 95 for a horse that is coming off a 6 month layoff that was vanned off last race.The point of the figs is to see what range of speed the horses in the race will run.(projecting a top effort in the making)If I just played highest figs I would lose alot-don't like that chalky taste either-lol.

oops forgot -see CJ's reply above explaining DRF rules for variants.

[pandy]

I made variants for all the major tracks for many years, and I found that you have to use some common sense. For instance, say it is sunny, dry, no wind, and the temperature is 75 degrees at Santa Anita all week. On Wednesday your par-based variant comes out to +5 (second fast). Under similar weather conditions on Thursday and Friday the variant is +4 and +6. Then on Saturday it is also 75 degrees and sunny and your variant comes out to +10.

Common sense says that the Saturday variant may be higher because the horses were faster. At this point, you have to start tinkering and maybe do some projections to check your track variant. I can tell you that if you see variants like this (higher on weekends) then the variant is questionable, especially under the same weather conditions.

[Valuist]

I'm trying to work on an a performance rating that encompasses a number of factors, starting with the speed figure but also taking into account pace, bias in dirt races, ground loss in grass races, and last but not least, any judgements or opinions I have on a given race (i.e. a given field being very strong or weak for that level, or a race with a large amount of speed or definite lack of speed). While I use speed and pace figures, many of my wagers are based on other judgements I make. I would like to have my figures incorporate those opinions into them. How practical will this be? It could be extremely time consuming. I may try only using one circuit.

[skate]

good topic

ive not the time to read all post," ill be back", but for now.

i do think the varient should be considered. as a mater of fact (derby) i thru out the so. cal. horses because of slower times (usually fast times) boong.
i was not thinking of the rain slowing down the tracks in so.cal. i use TSN figures for ease, but no varients since they allow for the varient already.
when i run my own figs. i take a guess at the varient, just to be in the ballpark. it bugs me, but for now...

anyhow, i think that the varients can be over examined, but a positive direction exist with them.
keeping in mind, pace is one figure and speed another. it is the pace figure that gets the most help from a varient. yes, the varient is elusive, but we are not looking for exactness here. if the horses themselves changed dramatically, then the varient would be useless. but the track changes, thats what we need to know.

so.cal. tracks are so consistent, as is turf paradise, no rain. i really forgot about looking for a true figure. and guess what, maybe the figure is true.

track varients and front wraps are elusive, my situation.

[Tuffmug]

Variants = VooDoo

[AQUEBUCKS]

Interesting subject...One that we could spend a lifetime figuring out. I worked a few summers during college at Belmont cutting grass , and I remember the routine with the "harrow yard" employees. The guys that drive the tractors...I remember that their instructions were sacred, and that they worked a lot of overtime. I always wondered why they got the extra hours at night. Thought at the time it was a scam or something, but come to realize now that trimming dirt, into the rail and outside was priveliged info, and done at night. Has anyone-and I'm sure some of you have-wondered about track maintenance, the harrows, rollers, water trucks, shavers etc...and what they do to the track throughout the course of the day and night. It's a relatively simple procedure to change the angle of the harrows to push a 1/4, 1/2 or 3/4 inch of dirt inside or outside during one or two revolutions. The track Superintendent instructs the crew according to the Racing Secretary's recommendations. This occurs several times throughout a race card. A bit more so in the past few years at NYRA. They have since obtained the services of a new track Superintendent. I think he's from Maryland, and his instructions are less sporadic as were his predecessor's. All-in-all, there are just too many variables when trying to constuct a perfect variant. I believe the median/average approach will work for the most part. For those constructing their own figures, a more subjective analysis by using video and comparing large databases. Rumor has it that they are trying to get rid of Belmonts notorious outside bias...We'll see this weekend...

[RonTiller]

PaceAdvantage,

Good question.

I take it the speed fig makers here have experimented with creating figs using NO variant and found the resultant speed figures to be miserable failures, correct?

Correct. Once we have a scale of speed points per length (how many lengths does 1 speed point equal, at each distance), we can make raw speed ratings (SR Raw), with no track to track adjustments or daily variants. Now, measure the performance of the last race SR Raw, as always, measuring 100,000 or more races. The win % of these raw speed numbers must now be a benchmark, as you correctly imply. Can they be beat by adjusting for track to track and daily variant? The answer is yes.

Many people use pars to construct both daily variants and track to track adjusments. OK, make the pars, several diffierent ways, create par adjusted speed ratings (SR Pars) and measure the win % of SR Raw versus SR Pars over the same samples. The win % is higher than for SR Raw. However, pars is just the first iteration of the game Beat Raw Speed Ratings.

But, how are we to assume the TRACK was the cause of the difference in "speed." Maybe it was the horses themselves?

Exactly. Here is where further iterations come in. Try projecting what the final time of the race should be based on the horses' previous races, the composition of the race, etc. You work out models, methodologies and techniques, measure 100,000 races, and see what you get. If you get a statistically significant improvement, THIS is now the benchmark to beat.

On a nine or 10 race card, comprised of a couple of sprints, a couple of routes, and a turf race or two, how can you possibly come up with accurate variants?

Well, there are easy cases and there are hard cases. If ALL the races on a card are at leat 2 seconds slower than the projections, at least prima facie, the track is probably slow. The hard cases are when the some races are slow and some are fast. What we have found is that really tracking down and analyzing these anomalous cases results in improved projections, which in turn reduces the number of hard cases. And as SJK points out, they usually do fall whin a normal range. If there were no reality behind variants, this would be extremely puzzling.

If one incorporates races after the fact into the projection, one can catch stinkers as well as improve the performance, especially with lightly raced horses have the least amount of data with which to make a projection. This does in fact make our speed ratings from a year ago more reliable than those from yesterday. In fact, the improvement is a couple of % points.

why you think that the variants you come up with on a day to day basis are grounded in any sort of reality.

I'm not sure that the question of what is THE correct track variant for the day is similar to the question what is THE correct charge to mass ratio of an electron. I do think there is an optimum value (or values) for each race, that, when used to calculate speed ratings, yields the highest win % when appropriately measured. The grounding in reality is statistically significant testing that shows adjustments to the SR Raw for track to track and daily variants result in speed ratings with better performance.

I should also point out that at some tracks, pure raw unadjusted speed ratings don't do appreciably worse than fancy high tech speed ratings. I recall that Hollywood summer meet did pretty good with raw numbers - few shippers and unvarying weather. Not so at a track like Turfway, where the difference is dramatic.

Ron Tiller
HDW

[Tuffmug]

Ron,

You left out the most important part!

What is the % improvement in win % in using speed figures vs raw times? 2%, 5% 10%, 30%????

[sjk]

Ron,

Pleased to see someone come out in favor of backwards projections and recalculating past variants periodically. I recalculate all variants on a weekly basis for all past days and treat forward and backward projections on an equal footing. I would argue that those who permanently fix the variant the day after a race are missing out on important information.

[Blind Pursuit]

I want to reiterate CJ's point that it's not just a variant based on "one day's" races but on the ~100 horses who ran that day. In fact, I think the biggest mistake is to give a single variant to all races at a particular distance on a particular day. There's no reason why the variant shouldn't change race-to-race as well as day-to-day. Actually, I think the variant should change from race-to-race. After all, the track crew works on the strip after every race; the weather changes; etc.

When you make figures for a race you look at the winner's raw figure (0 variant) and the raw figures for the horses who ran behind him, based on whatever methods you use. Find the variant that makes the most sense in the context of the horses' form coming into the race. Repeat that process for a day's races and the individual race variants should also end up following some kind of logical progression (track slowing down, speeding up, suddenly changing following a downpour, etc).

And, as others said, when the horses run back again, you can check your work and adjust as needed.

[Show Me the Wire]

PA:

My answer to your thought provoking question is one (variant) can be easily determined through mathematical exercises while the other one (bias) cannot. Constructing a variant is easily done with published data generated through the running of the actual race and other published summaries such as pars and pps. Constructing a track bias is not easily done or easily verifiable because the published data from the race does not contain information needed to determine the existence of a bias.

For example it can be easily construed from a day with eight sprint races the winners actually average 2/5 off the three-year track record and using mathematical formulas it can be calculated the surface was so many lengths or slower than day x. However, the information necessary to calculate the variant does not tell you out of eight sprints six out of the eight races were won by post position number 5 or 6, and that one race was won by post position 7 so you know that 7/8 of all the races won that day came from post position 5 through 7, with 6/8 of the winners came from post position 5 or 6. Or that the aberration from this day’s results the winner came from post 11 and the winner was a big class dropper. Additionally the data needed to construct a variant does not show the 7/8 of the winners were similar in position at certain points of calls during the race.

PA, I agree with you observation regarding the difference in the variant may be attributable to the horses themselves and not the track surface. More than likely it is a combination of both of the above-mentioned factors. Since the process of making a track variant assumes the difference is caused by the surface and the variant maker justifies this assumption by relating today’s performance to the horse’s past performances the creation of a daily variant is generally nothing more than a mathematical exercise based on a flawed assumption. It is flawed because the actual physical variations of the track surface are not quantifiable through the current information generated by the actual race.

Track bias exists as each racing surface is not uniform around the oval and more importantly the surface is not uniform across the width of the racing surface. Additionally, the physical configuration of the oval itself tends to influence the performance of the horse, i.e. banking of turns and/or radius of the turn. Each track surface is unique in composition and configuration.

The best use of a variant is to measure performance from track to track. The variant should lend itself to the ability to discern that a final time at one track is equivalent to a final time earned at another track.

However, as with all issues concerning horse racing there is an exception in my mind for the use of daily variants. Daily variants at winterized racing surfaces are useful for different segments of the meet. For example at Hawthorne the racing surface changes dramatically from the opening of the fall/winter meet to the end of the meet. The calculation of the variant is useful because the track surface changes are akin to horses racing on different tracks.

Summing it up the difference is the availability of generated data that makes variant making a respected pursuit and "track bias" hunting a futile quest.

[kenwoodallpromos]

1st, I like what you wrote in the Beyer thread!
2nd, variants is just that- variations. If you believe at all that there are variations in overall track speed that can change a race, then it is just a matter of how they are determined and the analysis constructed.
I know almost all handicappers believe variants occur on the dirt in conjunction with rain making an off track. If you believe all dry dirt tracks are always the same or that it is impossible to tell even if a track changes, you are entitled to that opinion.
I know why Hur thinks My methods are voodoo, but can you explain your general statement the variants=voodoo? How about off-tracks?

[RonTiller]

What is the % improvement in win % in using speed figures vs raw times? 2%, 5% 10%, 30%????
We calculate no pure raw speed rating for Theoretical Downs. The Raw Speed Rating we calculate as a work product has the track to track built in. Comparing THIS raw number to his variant adjusted numbers (i.e. the published speed ratings) increases the win % a couple of points on average. Some tracks, as I mentioned, are almost the same while others, especially in the winter, are helped much more. The bigger piece of the adjustment pie is in the track to track adjustments. Here, different methodologies yield much greater variations in measured win %s of the resulting speed ratings. I myself was surprised by this relative order of importance. But the measured difference between raw and variant adjusted is indeed real, as is the difference between backwards only projections and backwards/forwards projections. It would be to the detriment of the speed ratings' performance to leave them unadjusted, whatever one's theoretical predilections. And performance is what drives this whole mess.

So it seems to me that the data conclusively proves 1) variants improve performance and 2) different ways of making variants improve or degrade performance differently.

Ron Tiller
HDW

[mcikey01]

Seems to me that much of the research and commentary on the validity of variants presumes the existence of variability in track speed because of the need for a simple (simplistic?) explanation to the variability in thoroughbred race performance. Why presume that there are statistically valid explanations? Why not let the data do the talking???

Track variants are posited as a detectable pattern that "explains" the performance of a population or groups of thoroughbreds on a given day. If a "correct" and comprehensive set of questions can be presented and a "correct" and comprehensive set of variables identified, a database of considerable size could theoretically be "data mined" to pursue the possibility of the existence of those patterns referred to as track variants.

If anybody out there is intimately familiar with either the theory and/or practice of data mining, enlighten us with the possibilities of applying this concept to test the existence of valid track variants.

[Kreed]

YES. I wish someone responds to your idea.

[sjk]

I don't know anything about data mining but I can assure you that valid tarck variants exist. I make them every day (as do others). As to the assertion that they supply a simple explanation of a horse's perfomance I view them as only a part of the puzzle but if they were not meaningful I don't think anything else in my program would be meaningful either.

If your variants are a key element of your bet selection and you bet a lot of races with good results I take that as conclusive evidence that the variants are meaningful.

[Kreed]

Man, I need good variants. They're a necessary evil. I got the "equations"
but don't have the goods. I agree TV's are Voodoo but still what can you
say without them? maybe a program should give Velocity Figs after Each Horse's
races 2 ways: With Adjustments + RAW. Sometimes, if a track is Slow, I think
TVs OVER-ADJUST. After all, maybe the Horse is Slow also.

[sjk]

Kreed,

I would think that you need to take more care in making a variant than goes into the TV. It sounds like HDW has put considerable effort into optimizing their approach and I'm sure their customers appreciate the quality.

If you use a quantitative approach there are a limited number of elements you can put into the calculation so the care you take in finding the best way to define, compute and combine your elements can make a big difference in your results.

If you want to distinguish as to whether slow times are due to slow tracks vs slow horses, projections can help you. The downside is that using them forces you to keep up with a lot of data.

[PaceAdvantage]

Ron, great response. Thank you for taking the time with that reply.

I agree with you completely when it comes to track-to-track adjustments. It's the day-to-day variant that gave me that nagging feeling that something isn't quite 100% sound.

[JustRalph]

Wow ! good thread.....

[Tuffmug]

Ken and Ron,
I applaud the serious effort that variant makers expend in trying to make sense of this game. However, Ron Tiller's estimate that the adjusted figures may improve win % by a few points over raw figures validates an estimate I made long ago that variants and speed figures are relatively unimportant in the handicapping selection process. We are awash with speed figures and variants and yet the public still picks only 33% winners. Even Andy Beyer says that the top Beyer figure horse wins approximately ONLY 33% of the time. That statement tells me that at least 67 % of races are unpredictable by the use of speed figures. So what good are variants and speed figures? What have we gained -virtually nothing!

I do not hold that all tracks are the same or that variations in track surface do not have an effect on some races. I do believe that: 1) those variations are only a small piece of a much more complex puzzle: 2) that the methodology for creating variants is flawed because it does not take into account the effect of early pace on the final times of the races: 3) and that the public as a whole and even sophisticated handicappers have overemphasized the importance of speed figures as a means of handicapping and selecting winners.

[kenwoodallpromos]

I use variants to help me decide how much emphasis to give certain factors. Variants by themselves do not point to certain horses but to certain situations that may occur.
Even If a certain track or race speed did point to a certain horse you would still have to use subjective judgement as to the betting value and how strongly the variant will affect ther race and horse.
The main thing I like about pace over speed numbers is I do not think most speed number systems leave enough room for subjectivity, like certain adjustments.

[Tuffmug]

Ken,

Despite my opinions on variants and speed figures, I do use them in my handicapping but in a very limited way. I use them to point out extraordinary days and race performances in need of further study. Variants point out extremes and may lead me to upgrade or downgrade my opinion of a particular horse. I suspect you do the same and that is what you imply with your statement about subjective judgements.

[kenwoodallpromos]

IMO you use variants correctly.
Anything I use based on the track be it track profile, dynamics, biases, variants, or overall pars or tendencies is point of view with which to consider other factors.
I find some factors much more important than other in certain races so I continually learning to adjust even degree of influence of certain factors in particular races. That is why I told some personal contacts if they wanted to bet the 2005 Derby to bet all horses over 20-1 odds- because there were only 6 horses under 20-1 in a 20 horse field and a lot of 20-1+ who finished well in preps. Even if a low odds horse won the 20-1+ odds bet was of good value based on my value/risk assessment without a lot of other handicapping.

[nick777]

Let me throw my 2 cents in here

I think they take the time of the race and compare it to other times of horses from similar class and age, and their final time for the same distance, from previous days.

Looking at Giacomo, for example

SA Derby speed-87, variant-09, beaten 2 lengths, 95 beyer

S Felipe speed-98, variant-0, meaning best time for distance was equalled or bettered. 6 1/2 lengths behind, 93 beyer

Sham speed-83, variant-16, 1 1/4 lengths behind, 98 beyer

Something we used a long time ago, which is now called Beyer speed figures, goes like this.

Add speed+track variant, every 2 lengths is 1 point, ie. won by 6=+3, lost by 6=-3

you get this, using Giacomo

SA Derby, my #-94, beyer-95
S Fel, my #-95, beyer-93
Sham, my #-98, beyer-98

i haven't really used that for a while and probably forgot some factors and exceptions, but you get the rough idea

[Macdiarmadillo]

Back to PA’s doubts about a daily variant. Aside from every horse running that day you do have all the back history on those horses. A few are remarkably consistent, particularly turf runners, and keying those sometimes help determine a variant. It’s not the only thing you go by. But perhaps cj will agree with me, many times figuring the variant would just be a slam dunk and it would line up nicely with the day before and the day after. There are enough times you have to use some sensibility as to what the majority of horses (not just the winners) would run. This would include considering some normal improvement in horses. Sometimes identifying and throwing out races with pace extremes would give you the variant. Some days can be figured by interpolation between yesterday’s and tomorrow’s variant. It’s only certain days where I had to struggle.

You have to know your races/pars, too. A 2yo race is not necessarily a young 3yo race and neither is a 4yo up race. A 10k NW3L is not a 10k open race, nor a starter allowance for 10k horses. A 20k claimer with a purse of 16k is not a 20k with a 24k purse. I attribute the “tidal effect” at racetracks to the fact that certain types of these races were always run the same time of day and people didn’t make the distinctions. The racetracks that can’t fill enough races don’t show this effect anymore since they can’t schedule like that.

When I was calculating Beyers and doubted the daily variant I came up with, by far, it was an accurate measure, way better than what I could guess was proper. Pace was part of what was throwing off my perception. But also in calculating variants, I found oddball races that turned out to be race timing errors (later reported to be so, and not always corrected) and could largely cancel out the variation resulting from “about” distances without being at the track every day.

[cj]

I agree, there are some strange races, and some strange days. I don't think I've ever used different days, and wouldn't want to go that route. Too many things change from one day to the next, even if not apparent to us visually.

There is also the question of Days like the Wood and the Derby, where data is scant for route races. Wood day had two, run many hours apart, while the Derby only had two, one with a wicked fast pace. You have to look at all the horses in the race in these spots.

I was 90% sure I was way closer to the Wood figure than Beyer was, because I used the very consistent Scrappy T and Sort It Out to make my numbers, including the effect of pace. Now that Scrappy T and Bellamy Road have run back, and they both ran similar figures to what I gave them for the Wood, I'm about 99% sure.

One thing you can bet your hat on, Beyer will never go back and revise that 120, no matter how much evidence later points out it was bogus. I actually can't wait to post Scrappy T's PPs for the Preakness and show he ran virtually identical races in the Wood, the Withers, and several others, even though his Beyer's vary wildly.

[RonTiller]

I don't want to flog a dead horse but I'd like to clarify a few things.

Although I wrote rather vaguely of a couple of % points here and there, I don't want to give the impression that anything I wrote is just an estimate. We have precise performance data (an endless amount, it seems) for all the various components, including the part contributed by variants, that go into our speed ratings, by track, year, season, weather, etc. I am hesitant to publish exact data here because much is proprietary and it is Jim's work, not mine.

Fluctuating final times can be due to both daily variants (whatever THEY are) and pace / running style components; they are not mutually exclusive. In practice, they can be agonizingly difficult to deconstruct. Part of the job of good projections is to take into account as much as possible, including pace scenarios. And part of the job of assigning a variant to a race is to avoid the minefield of calling the track slow when the pace and running styles dictated the slow time, not the track. If it were easy, everybody would be doing it. It is frustrating beyond belief to meticulously analyze the data, come up with a way to improve the variant by factoring in some sophisticated looking pace scenarios, recalculate all the numbers for 5 years, and end up with a WORSE number. This is not for the faint hearted.

The options are 1) throw up your hands, for practical or theoretical reason, and use no variants, 2) do the research and analysis and take whatever improvements you can get, or 3) say f*#k it all - who cares about speed ratings anyway - trainer stats, track bias, tote analysis. etc. are where the action is. Speed ratings are for idiots (remind anybody of any characters on this board?) Option 3 is availble for everybody on this board except for Value Added Resellers like ourselves. Option 1 is not a real option either. That leaves option 2.

Finally, debating these fine points of projection methods and track variants in no way commits one to touting speed figures to be the single most important handicapping factor. If you've got to make a speed rating though, you might as well make the best one you're capable of making, or more accurately, the best one you have the time and patience to make, since there is an endless amount of tedious work involved.

Jim's own RsPos methodology relies secondarily on speed ratings - running style, position and pace are the focus of his own handicapping. The engineering problem of how to make the most predictive speed ratings is entirely seperate from what to do with them once they are made. Thise is where the real battle lines are drawn. Just don't shoot the piano player, guys.

Ron Tiller
HDW

[Show Me the Wire]

I don't want to flog a dead horse but I'd like to clarify a few things.


Fluctuating final times can be due to both daily variants (whatever THEY are) and pace / running style components; they are not mutually exclusive. In practice, they can be agonizingly difficult to deconstruct. Part of the job of good projections is to take into account as much as possible, including pace scenarios. And part of the job of assigning a variant to a race is to avoid the minefield of calling the track slow when the pace and running styles dictated the slow time, not the track. If it were easy, everybody would be doing it. It is frustrating beyond belief to meticulously analyze the data, come up with a way to improve the variant by factoring in some sophisticated looking pace scenarios, recalculate all the numbers for 5 years, and end up with a WORSE number. This is not for the faint hearted.


Ron Tiller
HDW

Basically, I understand the above to mean making daily variants is subjective just like trying to identify track bias.

Using judgments about pace and running style of the race is subjective opinion in determinig if a track surface is playing fast or slow.

[sjk]

I, for one, make variants in a thoroughly objective (and automated) way.

I see no reason why one could not make objective measurements about pace and running styles and incorporate them into the calculation if one's research showed that doing so would improve the predictiveness of the results.

[RonTiller]

Basically, I understand the above to mean making daily variants is subjective just like trying to identify track bias. Using judgments about pace and running style of the race is subjective opinion in determinig if a track surface is playing fast or slow.
If you mean a human reviewing each race and pronouncing judgement about the likely pace, you understand incorrectly. All variants we make are run on the computer without human intervention or "judgment". The sequence of programming steps that go into this process DID take a lot of human research and testing, so if you mean subjective in THAT sense, I can't argue.

If identifying track bias is fully programmable, using a series of algorithms that requires no human judgements or opinions (regardless of how complex), and if some performance parameter that track bias affects can be measured and quantified, then I guess they are similar.

The point is, when making projected variants, the better one can project the final times, the better will be the variant. HOW you choose to project a final time is up to you - roll the dice, hire Tom Brohammer to pace handicap every race every day, hire Len Friedman to sheets handicap every race every day, write a computer program based on years of research and testing. They all ultimately have to pay the piper - be measured and tested over lots of races.

Ron Tiller
HDW

[Show Me the Wire]

The sequence of programming steps that go into this process DID take a lot of human research and testing, so if you mean subjective in THAT sense, I can't argue.

Ron,

That is what I meant, someone made a subjective determination based on observation and interpretation of data. As you said the computer is processing a sequence of interpretaions of raw data.

Please, don't take my position as a knock on HDW or any of its products.

[KingChas]

If we all handicapped the same way.All had the same figures.All read the same publication.All bet the same horse.We should just chip in $50 apiece and play red or black on the roulette table all day!Moral of the story;
"Different strokes for different folks"

Always a Leader-Never a Follower!

[kenwoodallpromos]

The more objective the better. I feel I still am learning how to apply some subjectiviety. Simply clasifying a horse EP or S is fine but I have a feeling you do more than that. I guess the most important thing I am working on is applying variant and running style to possible race pace scenariio using my "figures".

[Tom]

When I do variants, I look at more than one day. I will print out a whole month at a time and look for days where the track condition remains constant. I then look at only the perceived consistatn races - older males, claimers and allowance, then same races for females. If I see a cluster of similar variants over 2-3-4 days, from these races, that is my variant for the week. this works great at som etracks, like FL, where we only have dirt races and basically four distances. A lot less meaningful at Belmont with more variaety.
I always go into the process assuming that is no variant unless I see convincing evience that there is.

Kreed - if you need good variants, look at CJ's program - he has variants - pace and final - for every race, every track, every day, and they are good. You can convert them from Beyer style to Quirin style easily.

[sjk]

Kenwood,

As you have surmised I do all the calculations objectively. That is the direction I chose many years ago. It eliminates all of the second-guessing and allows much more play since there is no time taken for looking at PPs etc.

On the other hand the development time is substantial and you don't know up front if you will be successful.

[Kreed]

Is there any way I could get your variants? btw, when you figure guys do
these variants, do you do (Dist Adjust's + Track-2-Track + TVs) all separate.
These issues get me panting. I would like to get a program that outputs these
figs.

[sjk]

Kreed,

I make variants only for my personal use and do not want to disseminate any racing data. At the same time, I have no reason to think that my variants are any better or worse than the other options available to you. I think there is more to how you use the variants than there is to how you make them anyway.

As to your other question, I have par times for each track and distance. The par times accomplish the comparision between races at different tracks and differenct distances. The role of the variant is to adjust the time based on how fast or slow the track was for the day. Pretty much the way we have all read about it from the people who wrote about these things years ago.

[kenwoodallpromos]

I do not consider myself a figure guy but I like to adjust for a variant and how wide around each turn.
I just began working on something so it is not properly tested yet. Right now I mainly use variants as one of several major factors and various figures as to how a horse handles various paces.
I have used speed figures to indicate a chronic physical problem occasionally.

[JackS]

I think the use of a variant is valid but I also think that it is used as an excuse for very slow horses when the Var. is exceptionally high and conversley for very fast horses when the Var. is exceptionally low.
It seems to me that a happy medium could be applied in these extreme situations.
A normal variant of say 15par, used as a standard to adjust arbitrarly a slow variant of 30. The difference is 15. In an attempt to lessen the effect that such a large variant may have on a final number, decrease the difference by 50%, thus the adjusted Var becomes about 23-24.
The opposite is an attempt to explain the final time of very fast horses, i.e, a Var of 05. The same difference subtracted from 15par leaves this horse with a variant of 10.
Sorry ,I can't vouch for a method such as this ,or even the 50% that I use(perhaps 30, 40 or 60% would be better) but I still feel using something rather than nothing makes sence.
Weather useing a method such as this to predict more winners or even fewer winners at better prices I can't say.
For those who are having success using an unadjusted variant, there is no reason to change. For everyone else, a change might be needed.

[kenwoodallpromos]

You are not supposed to use a variant or any other factor as an excuse, but as a learning tool if a wrong pick is made.
I always thought a variant Was a kind of adjustment. I do not see any sense in using a variable called a variant, then needing to adjust it. Just don't use it.
I do not adjust mine.
When I do make numbers, it is actually velocity. I use my median workout number in combination with the chart. If you use DRF variant numbers you would use these points: 0-5=0; 6-10=.8; 11-15=1.5; 16-20=2.3; 21-up=3.0.
I then use finish for sprints and 2nd call for routes- subtract 8' per length from leader/winner's time; add 8' for each path wide on each turn (2w=8'); divide by leader/winner's time in seconds into your horse's adjusted distance; then add points 0 to 3.0 as above.
Number is in feet per second and then you use it to compare horses. You still have to subjectively look at track, pace and trip as this number just gives you individual horse's pace probability from its best races.
It is basically early pace with adjustment for track speed and running wide.

[jotb]

add 8' for each path wide on each turn (2w=8'); divide by leader/winner's time in seconds into your horse's adjusted distance; then add points 0 to 3.0 as above.

How do you know what "path" each horse is in during the race?

[kenwoodallpromos]

All I have to go by is the videos and/or chart callers. I am open to advice and suggestions as to crediting for a wide trip. That is just what I am tryig at this time.

[JustMissed]

CJ, When comparing your KD pps with the Bris pps, I notice they have figures different from yours.

For example, with Afleet Alex's last OP race Bris shows:

E2=88
LP=121
Spd=107

Your printout shows:

Pace=91
LP=118
Spd=110

My question is, is the track variant part of the data that is transmitted from Bris to your computer program? If not, how does your program know what the par time and track variant is for each track you download?

Inquiring minds want to know.

Thanks,

JM

[cj]

Hey JM,

My program uses my own pars and variants. The numbers will usually not even be close to each other.

Also, they are on a different scale. Mine are on a scale similar to Beyer's scale, BRIS use their own, of which I don't know the specifics.

[BillW]

BRIS use their own, of which I don't know the specifics.

I believe they are as Quirin described (or close to it).

[JustMissed]

Hey JM,

My program uses my own pars and variants. The numbers will usually not even be close to each other.

Also, they are on a different scale. Mine are on a scale similar to Beyer's scale, BRIS use their own, of which I don't know the specifics.

I guess I was thinking that your program was a stand alone program but it must reside on the internet so you can change the pars and variants as need be.

I have never used any handicapping software except for the free stuff from TSN to use with their comma delimited pp files.

Thanks for continuing to have the free stuff. Very interesting.

JM

[cj]

... it must reside on the internet so you can change the pars and variants as need be.



I send out weekly updates.