Question on the Carroll Speed Figure

[Shacopate]

Quinn states in his new book that, "Carroll refines his methodology further by providing "constants" and "variants" to enhance the ratings."

Does anyone have a handle on this?

Such as, how the "constants" and "variants" are applied?

[highnote]

It's been awhile since I read Carroll's book, but I think "Constants" are his track to track constants. Through a bit of record keeping you might find that 400 yard races at one track are "constantly" 1/2 second faster than at some other track. So when you compare past performances between the two you need to take the constant into account.

Carroll doesn't believe in track variants. He believes in single race variants (if you can measure them). He thinks daily variants are overrated. I agree with him somewhat. Variants may be less of a factor in short quaterhorse races, but I think daily variants applied to longer thoroughbred races are useful.

"Handicapping Speed" is an excellent book. I highly recommend it, even if you disagree with hom.

[Shacopate]

Thanks for the reply.

A local book store had had this on the shelf for over a year. I skimmed through it once, and quickly discarded it because it seemed to focus on quarter-horses.

When Quinn made his mention, I got interested in the theory and decided to go and buy it. The bookstore was recently bought up and the name changed to "Borders". I went to the gambling section and found only 3 (handicapping) books. I asked the clerk at the info counter "what happened to Handicapping Speed?, it had been here for over a year."

He said, "that's why we returned it."

Anyway,

I think Carroll has some interesting ideas that need to be explored. I love his theory on beaten lenghts. If the "constants" are track to track adjustments, then that answers my question.

Thanks again.

[karlskorner]

You can buy his book for $14.95 at his site

www.desertsea.com/index.htm

I have been using the program at the end of the book since 91

[Shacopate]

Karl,

Must be a pretty good program if you've used it that long.

Can you explain how it works?

Thanks.

[karlskorner]

Answered in your PM box

[Maxspa]

All,
I checked the Desert Sea web-site and noticed that Carroll's
Speed Handicapper has gone through a revision process. The
update prices are reasonable.
Maxspa

[Zaf]

Its been a while since I read the book. In calculating Carroll speed figures for thoroughbred races, does CC recommend using a variant or just the raw times ?

Has anyone out there used his figures with success ? Do you feel his estimations of horse lengths is an accurate way to make a figure ?

Thanks in Advance.

ZAFONIC

[cj]

zafonic,

I definitely agree with his assessment that the length of a horse is right around 8 feet, and definitely not 10. I have it programmed into everything I do. It might not seem like a big difference, but you would be surprised.

[Zaf]

Thats seems like a more reasonable way to calculate beaten lengths. Most programs over simplify it.

ZAFONIC

[GameTheory]

8 feet is closer to the truth of how big the horses actually are, but a length is what the chart-caller sees. The best value is also dependent on just how you go about calculating times for non-winning horses. The way I do it I seem to get the most accurate times using 9.75 feet for a length. But I also try to estimate decleration, etc in what is probably a needlessly complex process.

The other sticking point is that while I believe you will get more accurate times (closer to what actually happened) by using something less than 10 feet, you'll get more predictive times (when comparing times or speed figures derived from those times from different horses in future races) by using a larger value. I think that's because the longer you make a length, the more you are penalizing a horse for being behind X lengths, which causes horses who finished closer up to get higher speed ratings regardless of their actual time. And if you run a database query you'll find that beaten lengths by themselves are a tremendously predictive factor, so if you tune the beaten length value by looking at predictive performance of the resulting times, you will tend to overestimate the value of a length. Did anyone follow that?

But if you get better performance, why not go ahead and overestimate it? That might be preferable for some, but if you're making velocity-based pace ratings for the fractions, it is going to throw those off...

[yak merchant]

Okay I know this has probably been answered someplace before, and I seem to recall it being discussed, but I don't remember a definite answer. When looking at a results chart. For lengths beaten, what the chart caller sees is what is put into the charts at the calls. But at the finish don't they take the finishing times and calculate the lengths beaten. Is the formula
(Time Behind/.20). Because because 1/5 of a second is alot different at 5 furlongs versus 1.5 miles. Or do they just eyeball that too?

[GameTheory]

Accounts differ, but I don't think they calculate it like that. When looking at the video myself the timings never match up to a neat formula. But they do take the finish beaten lengths from the camera rather than just eyeballing it. They have to do this to make sure the placings are right, of course, because that is how the purse is paid out.

[BillW]

Originally posted by GameTheory
Accounts differ, but I don't think they calculate it like that. When looking at the video myself the timings never match up to a neat formula. But they do take the finish beaten lengths from the camera rather than just eyeballing it. They have to do this to make sure the placings are right, of course, because that is how the purse is paid out.

GT,

Isn't the chart caller different from the placing judges though?

[GameTheory]

You're never going to find a chart that disagrees with how the horses were placed for payment purposes -- they both work off the video for the finish. What is the chain of responsibility for figuring it all out? I have no idea. But DQ's and DH's are "official" designations that appear in the chart, but are not decided by the chartcaller. It is not like the chart is made in a vacuum...

[BillW]

Originally posted by GameTheory
You're never going to find a chart that disagrees with how the horses were placed for payment purposes -- they both work off the video for the finish. What is the chain of responsibility for figuring it all out? I have no idea. But DQ's and DH's are "official" designations that appear in the chart, but are not decided by the chartcaller. It is not like the chart is made in a vacuum...

I'm sure the chart caller makes sure he coincides with the placing judge, but beyond correct placing, the beaten lengths can still be determined in an independent manner. I'm probably going to the track tmw nite and will ask about their procedure.

Bill

[BillW]

Originally posted by BillW
I'm sure the chart caller makes sure he coincides with the placing judge, but beyond correct placing, the beaten lengths can still be determined in an independent manner. I'm probably going to the track tmw nite and will ask about their procedure.

Bill

I confirmed last night that the Equibase chart caller at HOU does not routinely access the placing judges film for final beaten lengths, but rather determines them just as at the intermediate calls. Other tracks may do it differently.

Bill

[Figman]

The "keyman" here is the photo finish operator. He or she provides the info for the placing judges to certify the order of finish and for the chartman to have both the order of finish and the beaten lengths at race end. Those beaten lengths are taken right off the film. Each horse's actual finishing time is also on that film but I know of no thoroughbred track that makes use of this (to us handicappers) this valuable piece of info. The only final race times transcribed are the winner's time.

Bill, the placing judges at Hou probably were telling you the truth as they probably don't interact with the chartman. Their sole job is to determine the correct order of finish and it is not in their duties to deal with times or beaten lengths.

Here is the Texas rule on photo finish:
Sec. 309.125. Photofinish Equipment.
(a) An association shall provide an electronic photofinish device with mirror image to photograph the finish of each race and record the time of each horse or greyhound in at least hundredths of a second.
(b) The association may use an additional, more exact time measurement device.
(c) The association shall maintain an auxiliary photofinish device in case of an emergency.
(d) The location and operation of the photofinish device must be approved by the executive secretary before its first use in a race.
(e) The association shall, upon request, make available for viewing the photofinish of each race for win, place, or show in a designated area accessible to the public.
(f) The association shall ensure that the photofinish device is calibrated before the first day of each race meeting and at other times as required by the executive secretary.
(Added eff. 10/21/1999; (f) amended eff. 01/08/2004)

[BillW]

Tks Figman,

As I suspected, there is no rule but rather probably based on the relationship between the chart caller and the track people.

Bill

[Figman]

We wouldn't have to argue that old 1 length equals a 1/5 of a second axiom if the data collectors (Equibase) would just transcribe the times that are contained on the photo finish disk. At least at the NYRA racetracks, every horse that finishes a race is timed to the one/one-hundredth of a second and the time appears right on the film with the beaten lengths. This system is completely computerized and requires a single human operator. So the lengths behind are totally accurate and the times are right on the printout next to the lengths behind and totally discarded. Only the beaten lengths are utilized. WHY? And we think GPS timing at various racepoints is on the horizon? I doubt it!

[delayjf]

The other sticking point is that while I believe you will get more accurate times (closer to what actually happened) by using something less than 10 feet, you'll get more predictive times (when comparing times or speed figures derived from those times from different horses in future races) by using a larger value. I think that's because the longer you make a length, the more you are penalizing a horse for being behind X lengths, which causes horses who finished closer up to get higher speed ratings regardless of their actual time. And if you run a database query you'll find that beaten lengths by themselves are a tremendously predictive factor, so if you tune the beaten length value by looking at predictive performance of the resulting times, you will tend to overestimate the value of a length. Did anyone follow that?

Sorry GT, not sure I follow your meaning. Would love to hear more.

[GameTheory]

Originally posted by delayjf
Sorry GT, not sure I follow your meaning. Would love to hear more.

Careful what you wish for. Here goes:

-- We want to calculate final times for all the runners-up in a race.

-- Instead of using something like 1 length = 1/5 second for all races, we want something a little more adaptive depending on the distance of the race (1 length is a shorter amount of time in shorter races because they are running faster at the end than they are in longer races). Now the question "How long is a length?" becomes important because a shorter length will lead to faster calculated times than a longer length.

-- The Charles Carroll method, which started off this thread, goes something like this. This is from memory off the top of my head, but I think he calculates like this:

1 length = 8 feet

A 6 furlong race is 3960 feet, and therefore 495 lengths long.

Let's say the winner's time is 72.0, and we want to calculate the time of a horse that finished 5 lengths back.

The average time per length = 72.0/495 = .14545

Therefore, we give a horse that finished 5 lengths back a time of 72.0 + (5 * .14545) = 72.73. (I think Carroll diverges at this point and simply uses his time per length values in relation to a baseline standard for the distance to create speed figures.)

(Yes, there are flaws with this method, but we aren't talking about that now.)

But let's say we had used the more common 1 length = 10 feet.

6 furlong lengths per race = 396.

Avg time per length = .1818

Time for horse 5 lengths back = 72.91

So increasing the value of a length by 2 feet in this race gave a us a time 72.91 - 72.73 = .18 seconds faster, which could be significant. If it was a longer race, the difference would be even greater. If this horse was farther back, the difference would also have been greater.

-- Even if you don't use Carroll's exact method, any similar method that attempts to adjust for the length of the race and actual time of the race will be sensitive to the value of 1 length.

-- So we've established that the longer 1 length is in terms of feet, the slower the calculated times are going to be.

-- Any times we calculate are estimates, and we have no measurements of the actual times of each horse to compare to, so we have no way of knowing whether our calculated times are actually accurate. However, we know the winner's time IS accurate, so we want to be as close as possible to reality for the runners-up in order to be able to compare speed figures derived from our times with confidence. If we calculate things like feet-per-second pace ratings we also want accurate times.

So, now we're left wondering if we made the right choice of how long a length is supposed to be. Carroll chose 8 feet because he claims that is the average length of a horse. I dispute him on that -- Davidowitz points out in his book a study of thoroughbreds that puts the average at closer to 9 1/2 feet (I think). I've never measured a horse, but 8 feet seems a little shy. Carroll might be talking about quarter horses, as that is what much of his book is about. Also, we really need to match up with what the average chartcaller sees, which in the action of a horse race I believe will tend to be stretched a bit. Once again, without scientifically measured actual times from each horse, we have no way to tell who's right.

With me so far?


Well, if we went to all the trouble of calculating those times, we're probably going to be making some sort of distance-equalized speed figures from them. We'll leave aside the question of how we might do distance equalization (Carroll divides his time per length for each horse by the record for the distance), but we do know that faster times will lead to higher speed figures, right? And therefore a "longer length", which leads to slower times as established above, will lead to lower speed figures.

But we still don't know if 8 feet is better or 10 feet is better. We might be tempted to figure them both ways, and then compare the performance of the resulting speed figures to find out which works better. That's what my original quote was about. When we talk about measuring "performance" of speed figures as a substitute for "accuracy" of times, we've actually made quite a leap. Because to measure the performance of a speed figure you're going to be looking at speed figures from PAST races to predict who will run faster today, and it is generally assumed that the most accurate speed figures (most representative of the true time it took a horse to run a race) will be the most predictive figures of future performance. (Will predict the most winners.) I think that assumption is FALSE!

At least in some scenarios. I can't make a hard case because there are an infinite number of ways to take past speed figures to predict future ones. Do we use the last race figure? Do we use the average of the last 3? Best lifetime? Best in the last 3? Best in the last 90 days? There are a million ways to do it.

Let's choose the most common -- the last race speed figure. (I'm closing in on my point, I promise.) We'll look at the last race figure for each horse in a race, and predict the highest figure to be the winner. But let's calculate our speed figures in two ways: one way with 8 feet for a length and one way with 10 feet. We do this for a large sample, and then we predict a bunch of races each way. We find that our winning percentage is higher with the figures that used 10 feet as a length. That means 10 feet must be more accurate than 8 feet, right?

NO!

Let's try 11 feet. We find our win percentage goes up a little more. Now 12 feet -- it goes up again very slightly. Ever see a 12 foot horse? I doubt it. Now 13 feet -- now it's leveling off -- no real gain in win percentage.

In my tests I've gone up to 14 feet for a length before seeing degradation in predictive performance, and often the best value is around 12-13 feet, which is clearly an outrageous value for a length. The times it produces are crazy.

Why does this happen? Because "beaten lengths" from recent races are very predictive by themselves, and by increasing the value assigned to a length you are magnifying the effect of beaten lengths. Horses that won don't have their times affected at all, but horses that are more than a few lengths back are severely punished figure-wise and what you're really making is a times/beaten length power rating -- not a speed figure. (Horses that finished close up in the past will tend to finish close up in the future -- at least more so than horses who tended to finish farther back in the past.)

So how do we figure out how long a length is? Well, the way I did it is to hand-time a bunch of runner-up horses and then adjust my time calculating procedure and beaten length value until I was fairly close most of the time. You'll find there really is no good value because the horses are all slowing down at different rates. But I find something between 9 - 10 feet works best for me. Without hand-timing races, what you need to do is look for your past figures to predict the VALUE of the figures for the same horse in the future, which relates time to time instead of time to winning (which is what most people do). In other words, you want to reduce the variance between figures in successive races. But there are lots of problems there too which I won't get into now.

With all these problems, it is easy to see why most people just stick with 1 length = 1/5 second and be done with it. And once you use beaten lengths or position as a factor and discover they are nearly as predictive by themselves as real speed figures, you begin to wonder if it is worth it to calculate times at all.

Did I get the record for the longest explanation ever?

[Niko]

Great Explanation but I might have missed something simple at the end.
How can position or beaten lengths be as predictive as speed ratings. Are you talking in the context of class of race or in relation to the winners time?
I read it twice and maybe a hard day of work has gotten the best of my mind.
If it's not in relation to speed or class I don't get it? I get the variation of a length concept though.
As I say "there are no stupid questions, only stupid people" but I'll ask this one.

[raybo]

Goodness gracious, GT! Whew. This is a touchy subject, for sure. I personally use ".2" as a multiplier but I tend to think like you, GT, (I think), that it doesn't really matter what factor you use as long as you apply it consistently. I have used varying factors ,according to distance, and don't see much difference in the bottom line. I think I speak for everyone knowledgeable in handicapping when I say that "raw times" mean very little when comparing one PP to another, even at the same track. I don't think you can get too technical in this area but should use these times or figures, or whatever you use, as a general indication of performance and as an indication of a horse's condition at a particular place in time. After all, a horse's condition determines what he is capable of at the time. I personally try not to get too caught up in the exactness of numbers but rather in the direction and volume of the changes in those numbers and how the race proceeded. Numbers will only take you so far, then it's up to how well your mind analyzes what you see. If you get too dependent on the numbers themselves you forget how to think. Don't get me wrong, I use the numbers and calculate my own "figures" from them in my handicapping analysis program but all those numbers and calculations are really only a guide. The real "numbers" are determined by the human brain shortly before post time, at the point of deciding whether to bet or not. I will bet an improving animal over an animal with a better time or figure more times than not. It just makes sense to me to bet on a horse that is moving towards his capability rather than one that you aren't sure will perform at those better numbers.
For what it's worth :rolleyes:

[GameTheory]

Originally posted by Niko
Great Explanation but I might have missed something simple at the end.
How can position or beaten lengths be as predictive as speed ratings. Are you talking in the context of class of race or in relation to the winners time?
I read it twice and maybe a hard day of work has gotten the best of my mind.
If it's not in relation to speed or class I don't get it? I get the variation of a length concept though.
As I say "there are no stupid questions, only stupid people" but I'll ask this one.

I mean totally raw beaten lengths or position taken as a rating (in both cases lower = better).

If a horse finished 5 back, then he gets a 5. If he finished 10 back, he gets a 10.

What I'm saying is that if you bet on "lowest beaten lengths last out" and "highest speed figure last out"; the beaten lengths wouldn't do badly at all in comparison. I'm not suggesting it actually be used raw as a rating, just saying it has a lot of predictive power on its own -- beaten lengths in the past has a high correlation with beaten lengths in the future, and horses who finished up close last time are likely to be the ones finishing up close today....

[GameTheory]

Originally posted by raybo
Goodness gracious, GT! Whew. This is a touchy subject, for sure. I personally use ".2" as a multiplier but I tend to think like you, GT, (I think), that it doesn't really matter what factor you use as long as you apply it consistently. I have used varying factors ,according to distance, and don't see much difference in the bottom line. I think I speak for everyone knowledgeable in handicapping when I say that "raw times" mean very little when comparing one PP to another, even at the same track. I don't think you can get too technical in this area but should use these times or figures, or whatever you use, as a general indication of performance and as an indication of a horse's condition at a particular place in time. After all, a horse's condition determines what he is capable of at the time. I personally try not to get too caught up in the exactness of numbers but rather in the direction and volume of the changes in those numbers and how the race proceeded. Numbers will only take you so far, then it's up to how well your mind analyzes what you see. If you get too dependent on the numbers themselves you forget how to think. Don't get me wrong, I use the numbers and calculate my own "figures" from them in my handicapping analysis program but all those numbers and calculations are really only a guide. The real "numbers" are determined by the human brain shortly before post time, at the point of deciding whether to bet or not. I will bet an improving animal over an animal with a better time or figure more times than not. It just makes sense to me to bet on a horse that is moving towards his capability rather than one that you aren't sure will perform at those better numbers.
For what it's worth :rolleyes:

If I'm going to be making a numerical rating or calculating times for every horse, then I will take the time to make it as complex as I think it needs to be. The reason? I program it all into the computer -- I only have to figure it out once, and then it is done automatically in a microsecond for every race. So it doesn't really matter if it is complex or not. But most of the time, the extra complexity doesn't really get me much; but when I add up the little bit here and the little bit there I gain by using more sophistication for all parts of the process, it adds up...

[raybo]

(From GT: If I'm going to be making a numerical rating or calculating times for every horse, then I will take the time to make it as complex as I think it needs to be. The reason? I program it all into the computer -- I only have to figure it out once, and then it is done automatically in a microsecond for every race. So it doesn't really matter if it is complex or not. But most of the time, the extra complexity doesn't really get me much; but when I add up the little bit here and the little bit there I gain by using more sophistication for all parts of the process, it adds up...)

I agree that if you are crunching numbers anyway then, by all means, be as accurate as possible. My point was that no matter what numbers you use, whether it is times or speed figures or derivations of either, they should be used judiciously. A speed figure of 88 is not necessarily better than one of 85. It depends on the surrounding circumstances. They can be used as direct comparisons only to a certain point then we have to determine their validity ourselves.

[GameTheory]

Agree. I have two kinds of numbers: those which are data (inputs) and those which are predictions/probabilities (outputs). For the data side, I decide how I'm going to calculate them pretty much independently of how I think I might use them. I'm just trying to measure something that might not be measured elsewhere, and do it consistently. I want similar values to "mean" the same thing, more or less, all else being equal.

Then I apply my various prediction techniques and let them figure out if the data number is good for anything useful...

[raybo]

GT, you hit it right on the head! I take the "hard" data from Bris and massage it for my own purposes, then I run scenarios through a database of past races to check the validity of new ideas.

[Tom]

In his energy base programs, Energy and Thoromation, Kgen, and maybe even later ones (I stopped at T-Mation) Sartin used varying feet per length for different distances of race segments.
The result was that closers were more disadvantaged. The pace ratings that resulted were more predictive. I don't remeber exactly what the values were, but I will look for them and post.
Maybe some of the current Sartin users here will know this.
If I remeber correctly, if two horses were rated off the same race-one wire to wire, the other 5 back at the second call and then dead heats the first one, they will not re rated the same total energy-the closer will get a lesser rating becasue of his pace deficit.

[raybo]

I would be interested in seeing the actual multipliers he used at different points in a race. I have toyed with that idea in the past but those days were my "pre-computer" days and I was in over my head at the time, so I accomplished nothing concerning that. The question of whether or not a closer should be punished for his seeming lack of effort early is controversial, to say the least. I guess I agree that a horse that is able to "wire it" might be more predictable. However, don't you agree that such a horse will be lower priced as a result? Picking lots of winners is great, but if they're less than 2 to 1 what's the point? I don't know many handicappers with a 50% win percentage, do you? I personally prefer closers to front runners or pressers because I can usually depend on them to be moving towards the front of the field before the wire rather than the reverse and at a higher price much of the time. I guess you'd call me a racing "contrarian" of sorts. I don't believe in basing my wagers on the same criteria that the majority of race bettors rely on. High speed numbers are just too easy to see. I look for less visible indicators like combinations of pace figures put in context by the class of competition a horse faced. I concentrate on condition in the end because , in my opinion, that is the factor that wins most races. And, as you know, that is the most difficult factor to figure.

[delayjf]

Whenever I read something on this page that I find revolutionary or interesting, I copy it into a word document for future reference.
Congrats you've made another entry into my "The Best Handicapping theories of the PaceAdvantage Board". I appreciate your answer.

I don't know if this came up earlier or not, but Davidowitz also commented in his book that horses are also able to make up the beaten lengths in shorter races quicker, due to their higher velocity. So in this case, a beaten length is worth more at the longer/slower distances, which is the opposite logic of what previously had been widely accepted in Speed figure world.

Ken Massa at HTR calcualtes his beaten length that way and I suspect that Beyer does to. In the small print of Beyers newest book, Beyer mentioned that his beaten lengths are now calculated based on the final time of the race, so I assume that is what he is doing.

GT
Now how about your theory on distance equalization??

[JackS]

If using a pencil and paper method, making adjustments for actual time a horse runs behind a leader has to be only reasonable. Short of having stopwatches stregically placed at the vaious calls, no one can know for sure how fast a horse behind the leader was actually running. Someone has stated in this thread that horses are really closer to 8ft long rather than 10ft. MY assumption has always been 9-11ft and the average 10ft. Due to the inherant errors of chart writers and their estimations of lengths behind (understandable) an attempt should be made to adjust times and not lengths. The common 1/5sec per length can be used as an average but does not truly represent how fast a horse ran early or late. A typical race run in 22 45 110 and a typical pace line in the same race with a horse running behind 4th x 5, 4th x5, 4th x 5 lengths, can be adjusted using a percentage factor. The factor can be a personal opinion or the factor could be derived fom a computer program. Using pen a paper, I chose 10% to be applied to the 1st call (decrease) and 10% to the final time (increase). Final pace line adjustment for this horse would be 22.8 ,46 flat, and 110.1. All times here are in tenths. Accuracy is not absolute and will never be. True adjustament which might place you closer to actual time might be possible but would probably be impracticle. Choose a % that you beleive is close and one you can work with accuratley and fast.

[GameTheory]

Originally posted by delayjf
Whenever I read something on this page that I find revolutionary or interesting, I copy it into a word document for future reference.
Congrats you've made another entry into my "The Best Handicapping theories of the PaceAdvantage Board". I appreciate your answer.Thank you. I have often thought someone should put together a "PA's greatest hits" -- maybe in an FAQ format that would address all the major areas of handicapping according to PA members -- highlighting the important posts & threads we've had on each topic.

I don't know if this came up earlier or not, but Davidowitz also commented in his book that horses are also able to make up the beaten lengths in shorter races quicker, due to their higher velocity. So in this case, a beaten length is worth more at the longer/slower distances, which is the opposite logic of what previously had been widely accepted in Speed figure world. I thought that was one of Beyer's supposed "revolutionary" ideas -- lengths are worth more time at longer distances. Of course Davidowitz' original edition predates Beyers stuff, I think.

Now how about your theory on distance equalization?? My official position is that I don't like it. I do make distance-equalized speed figures, but I don't really like comparing unlike distances too much because I think the distance preferences/abilities of the horse is a major handicapping factor.

We all know that if a horse runs an 80 speed figure going 6 furlongs, that sure as hell doesn't mean he's going to run an 80 going 1 & 1/8 miles. The two 80's are equivalent only mathematically and usually class-wise -- i.e. a horse that can run an 80 in a 6 furlong race is equivalent in class to another horse that can run an 80 in a 1 & 1/8 race. But who will win when they meet head-to-head is largely dependent on the distance over which they run. So distance equalization is really a class index (I think Cratos pointed this out about Beyers recently), which has its uses, but not for determining who is faster in a particular race going a particular distance.

A true distance-to-distance adjustment that purported to measure speed and not class would have to take into account what the individual horse's best distance was and add some some of penalty when converting to other distances, and vice-versa. But then you're getting into making individual adjustments for each horse, which of course is what handicapping is all about.

But even given that, I do have a distance-equalization in my speed figure procedure. But I also have a track-equalization that converts times back to their original distance, but adjusted for a universal baseline track/surface. There is nothing real special about the way I arrive at these adjustments -- I use class-based pars (except the pars are actually predictions created on the fly, they are not stored in a table) and then I calculate an average difference between equivalent classes at different distances to get the adjustment. All my adjustments -- distance to distance, track to track, even daily variants -- are expressed as multipliers (ratios). So after the adjustments are calculated, I have 2 or 3 mulitpliers which I multiply by the original time to get the adjusted time.

[JackS]

Edit- My post above to 22.9 first call and 111.1 final. 10% is a very difficult number to work with I guess.

[PaceAdvantage]

I think Larry Hamilton was compiling an archive of "best of" threads as well....

[Tom]

I have copied and pasted a lot of items from PA and from HTR and HSH boards ( in fact, even the old Yahoo board). There is stuff here that you just aren't going to find anywhere else. I think there is probably a cutting edge handicapping book nestled in here.
And free PPs to boot!

In fact, there is even a joke book in here somewhere! hehehe

General Handciapping: "Are you a handicapping thread?"
Off Topic: "I'm afraid not." (a frayed knot!)

:rolleyes: