How the public bets

[GameTheory]

I submit that the public overbets all positive factors, and underbets all negative factors. And by "all", I don't actually mean all, but the factors they actually consider, or suitable proxies for what they consider. (They may be looking at Beyer figs as a factor, but if in my research I look at BRIS figs instead and see how they correlate to the betting, the effect will be similar.)

In other words, if you isolate single factors (like "last race speed figure" or "trainer win%" or even your own composite factor that no one else has access to) and compare how that single factor performs in terms of predicting winners while at the same time comparing how that same factor predicts the final odds, what will you find? In most cases you will find that they overbet or underbet the factor (same thing really depending on which way you look at it) in a predictable fashion. A high trainer win% tends to pick winners, but when looking at the betting you will find it doesn't pick as many as the betting on that factor would indicate. Same with any other positive factor. With a negative factor, the actual performance of the horses is better than the betting would indicate. Again, the negative side is just the other side of the same coin -- low trainer win% is a negative factor.

So with any predictive factor, the more positive or negative it gets, the more the public will overdo it betting-wise in that direction. I would then conclude that the quite high accuracy the public achieves is not due to their great capacity to weigh each factor properly and get it just right, but rather due to the fact that they are considering many factors, both positive and negative, and it is their errors of overcompensation that cancel each other out. So if they see something positive about a horse they will tend to give it too much credit for that factor, but if they then see something negative about that same horse they will then deduct too much from his winning chances based on that. The overall effect is that the errors cancel each other out and they get it about right in the end.

What does all this mean? Maybe nothing, but maybe that good betting opportunities come when a horse has an unbalanced number of positives & negatives (or unbalanced amount of aggregate perceived magnitude of positives & negatives). If a horse has more apparent negatives than positives, it will probably be underbet (although it still won't win very often). And if it has a lot of positives and less apparent negatives, it will probably be overbet (although it will still win often).

This sort of thinking is probably more useful for finding overbet favorites than uncovering winners...

[Fastracehorse]

I always thought that a well bet favorite was easy to see: Dominant Beyer speed figure and top jockey.

Top trainers are important, however, I am always surprised by how many horse players see the jockey as the #1 factor that determines the outcome of a horse race.

fffastt

[Dave Schwartz]

Oh, this is going to be a great thread!

Good job for starting it, GT.


Dave Schwartz

[GameTheory]

Oh, this is going to be a great thread!Not if you don't actually add anything... :)

[shanta]

The public bets a good last race.They want a good finish and oh yes a recent one too.

Find horses who have a horrible last race. If you can work around that race for whatever reasons / method you use almost always VALUE is added to the price of the horse man.

Most times a horse who had a sharp out last race and is a price today raises big time red flags with me. Unless the horse is stepping up in company I gotta ask myself " why am I blessed" with such a big overlay. Usually something's very wrong and the race when run proves it out.

Horrible last race = VALUE

Rich

[stu]

I am on my 4th day as the morning line maker for Ruidoso Downs (and 4th day ever as a morning line maker). The question that I ask myself is:

Q. "Who is the betting public?"
A. It depends who is taking our signal.


The first two of our races on Thursday and Friday are on TVG. With that in mind I am adjusting the odds to favor any horse that has recently run in California.

I am assuming that the percentage of bettors from California will over bet running lines containing BM, DMR, FPX, GG, HOL, LA, and SA. I plan on making the same horses a few points higher on races not aired on TVG.

I am open to constructive criticism on my morning line

[Overlay]

Top trainers are important, however, I am always surprised by how many horse players see the jockey as the #1 factor that determines the outcome of a horse race.

fffastt

I don't discount the importance of the trainer, but I lean toward working with the jockey because of combining the influence and/or intentions of the trainer with being the one who has direct control of the horse during the running of the race (in addition to some jockeys being undeniably more talented than others), and because a smaller percentage of the local population accounts for a greater share of victories with jockeys than with trainers. (At the same time, though, I believe that the public can go overboard too often on hot riders, and bet their mounts down to odds that are lower they should be.)

[Jeff P]

GT - extremely interesting thread that you've started...

I've spent a great deal of free time the past few months focusing on live play in the a.m. and R&D in the p.m. A lot of my R&D efforts have to do with devising new compound factors and then running them against my databases to see how they perform. The factors that help my own live play the most all share one common thread: They are things the public generally doesn't react to. I submit that the public overbets all positive factors, and underbets all negative factors. And by "all", I don't actually mean all, but the factors they actually consider, or suitable proxies for what they consider. (They may be looking at Beyer figs as a factor, but if in my research I look at BRIS figs instead and see how they correlate to the betting, the effect will be similar.)I agree with the above statement in general. Positive factors, with very few exceptions, tend to attract money. Negative factors tend to scare money off. Like you, I've come to the conclusion that when looking at large chunks of races the public overbets the positives and underbets the negatives.

Interesting that both final time based speed figures and morning line are mentioned in the same thread.

My own R&D says that traditional final time based speed figures (no matter who makes them BEYER BRIS TSN DRF etc) while predictive in nature all tend to be overbet by the public. This is especially true when looking at speed figurres taken from the last running line of a horse's record. Mainstream paceline picking tricks such as best of last 3, best of last 5, best of last 10, best 2 of last 3, best 2 of last 5 etc. don't appear to make an effective difference maker when it comes to grinding out profits at the windows.

I think there are two very obvious reasons for this:

1. Final time based speed figures all attempt to measure the same thing: How fast a horse ran from Point A (the gate) to point B (the finish line.)

2. Because the very thing that final time based speed figures measure makes such intuitive sense to so many, horses with strong speed figures tend to attract more money than horses with weak speed figures. My own R&D shows that the public consistently makes fewer mistakes with high speed figure horses than weak speed figure horses.



My own R&D shows that morning lines fall into the same category of effectiveness that speed figures do. For example, I have the subset of ML favorites winning all races in my databases last year at a rate of 31% with a flat bet win roi well below the track take. Predictive? Yes - highly. Effective at grinding out profits? Hardly... unless you avoid your selections that happen to be ML favorites.

I think the job of making a good ML has to do with two areas:

1. Predicting how the public is likely to bet.

2. Predicting a race winner.

I could be wrong here but I'm guessing that a lot of ML oddsmakers look at speed figures - as well as a lot of other positive things that the public overreacts to - when they create their lines.

The real trick for the successful player, I think, is the ability to find and evaluate the records of horses in terms of both hidden positives as well as hidden negatives... things that affect race outcomes but that the public and/or ML oddsmaker generally aren't even aware of. At least my own R&D (and experience with live play) continually bears this out.

This post is starting to get long - and because I can see where I'm going with it next I realize it's going to get a LOT longer... it's also a really nice day out here and I told a few friends I'd meet them at the beach to drink a few beers and catch a wave or two... so I'm going to stop now and pick this up at a later point.


-jp

.

[classhandicapper]

"The real trick for the successful player, I think, is the ability to find and evaluate the records of horses in terms of both hidden positives as well as hidden negatives... things that affect race outcomes but that the public and/or ML oddsmaker generally aren't even aware of. At least my own R&D (and experience with live play) continually bears this out."

:ThmbUp:

Can I add one thing?

I think it can be very helpful to understand those factors that the public tends to overbet if they are in fact very positive. Otherwise, you will tend to create odds lines that don't reflect the true chances of the horse very well and make bets on "supposed overlays" that really aren't overlays at all.

For example, just because the public tends to bet certain trainers heavily, it doesn't mean those trainers aren't winning more races than the PPs of their horses suggest they should win relative to other trainers (all else being equal) .

Todd Pletcher, Bruce Levine, etc.. in NY are heavily bet, but if I see 2 horses that look identical in every way except that one of them is trained by the above and the other is trained by Joe Blow, the Levine/Pletcher horse will be lower on my odds line every single time and I'll be quite confident I am evaluating the chances properly.

[Dave Schwartz]

Not if you don't actually add anything

LOL - Well, let's see if I can raise the bar just a little. I'll see if I can come up with a little something to make it interesting.

I built a "Race Filter:"


Dirt
6fur
Fast Tracks
Claiming & Optional Claiming Races
Males
Older
8-horse fields


http://www.horsestreet.com/BBSImages/PubBets01.JPG

http://www.horsestreet.com/BBSImages/PubBets02.JPG

http://www.horsestreet.com/BBSImages/PubBets03.JPG


I ran that Race Filter against our database which currently contains 297,000 races from the last 5 years It found 4,368 races.


http://www.horsestreet.com/BBSImages/PubBets00.JPG


Next, I built some Impact Value Tables of the 13 different ways we look at FT:


FT01 = Last Race
FT02 = Best of Last 2
FT03 = Best of Last 3
FT04 = Best of Last 4
FT05 = Best 2 of Last 3
FT06 = Best 2 of Last 4
FT07 = Best 3 of Last 4
FT08 = Avg of Last 2
FT09 = Avg of Last 3
FT10 = Avg of Last 4
FT11 = Best Ever
FT12 = 2nd Best Ever
FT13 = Best 2 Ever


Here is the table for Last Race:

http://www.horsestreet.com/BBSImages/PubBets04.JPG

Notice the column labeled "PubCh." This is the average public choice number for each rank. Okay, so you get the idea.

More to come...

[GameTheory]

Okay, so you get the idea.Maybe I do, maybe I don't. "Public choice number" is what -- the average ranking the public gives it or the actual odds?

[dastar]

Great Topic GT:

Going back into the 1980's, a trainer from Phila Park, would come into the turf club in Upper Darby, and make bets:

I can't recall the exact horses name in particular, but Red Ball/Rubber Ball, is fairly close.
Any way, When Alan Iwinski went to the betting windows, we all would play what he was playing. Not sure about today, but when you were good friends with certain tellers, you got all kinds of useful information.
Of course, we were only 5/10 dollar betters, so we certainly were not going to influence the odds!

To be perfectly honest, I remember when Alan put 6 bills ($600) on the nose, and was a distant third. Guess Trainers don't win every bet either!

My point is this: If you can identify which trainers are more likely to bet, than those who are not bettors, you can justify a solid bet on that particular trainers horse: Providing, you know he is betting on the race he's in.

This is not easy information to come by, but can certainly be helpful, if you can find a few.

It is hard to argue who makes the choice whether to send the horse, or just give him a tune up for an upcoming race.

I only want to know which horse is in TOP shape today, not 2 months ago!


dastar

[Dave Schwartz]

This screen represents the top ranking in each of the 13 ways of looking at FT (FInal Time or "Speed Rating").

Let's see what we can see here...

(You have no idea how difficult this screen was to paste together.)

http://www.horsestreet.com/BBSImages/PubBets05.JPG

Let's look at them in groups...

First, we have the "Best race of Last 1,2,3 or 4."

http://www.horsestreet.com/BBSImages/PubBets06.JPG

See how the "Average Public Choice number flows so nicely?

This says that the public considers the last race only slightly more often than the 2nd race back, but they are so close as to be virtually the same. (Note: The "2.61" means that the average horse with this characteristic is between the 2nd and 3rd public choice.)


Next, let's look at the "multiple-best-of's;" namely B2L3, B2L4 and B3L4.

http://www.horsestreet.com/BBSImages/PubBets07.JPG

Look! The average public choice on B2L3 and B3L4 actually lower than the Last Race! Let's place this in the proper perspective a little, shall we?

What this says to me is that the public wants the horse that has at least two good races in it's last 4 but no more than one bad race.

[Dave Schwartz]

Next, let's move on to the "average" approach, looking at Avg2, Avg3 and Avg4.

http://www.horsestreet.com/BBSImages/PubBets08.JPG

Here we find the absolute lowest public choices! So, the public actually bets Avg3 and Avg2. The public prefers the "perfect" horse... the horse with the best three-in-a-row!


Finally, we move to the "past class" horses; the Best Ever, 2nd Best Ever and Best 2 Ever.

http://www.horsestreet.com/BBSImages/PubBets09.JPG

Look at the difference! In comparison, the public virtually ignores these horses!

[Dave Schwartz]

Finally, let's take a quick look at performance.

http://www.horsestreet.com/BBSImages/PubBets10.JPG

Outlined here are the highest $Nets (i.e. return per $2.00 wager) and PIVs ("Pool Impact Value").

There is no denying the impact of the last race on performance, but it would appear that a place to find the overlooked horse would be in the "Best Ever" area; the Past Class horse.


GT,

Finished now. Hope this is what you were looking for 'cause it was a lot of work. <G>


Dave Schwartz

[GameTheory]

Ahem...

Let me suggest a simpler metric than your database extravaganza to illustrate what I'm talking about.

Let's take any simple factor, say LAST RACE SPEED RATING. Now obviously with this factor, higher = better. If the public likes a horse more, they bet more money on him than a horse they like less. So higher = better = more money. Or it should, all else being equal. Of course, when isolating a single factor in a single race, all else is never equal, so we isolate that factor over many races and look at the average.

So how do we determine how well a factor is performing? Well, one simple way is to look at how often a horse with a higher factor value finishes ahead of a horse with a lower value. But since our goal is to compare to the win odds (which supposedly represent the chances of each horse to win, not their projected order of finish, a subtle difference), a possibly more correct way is to just look at the winner and compare him with the losers in terms of his factor values.

I'm sure no one understood the above paragraph, so let's look at some examples:


Horse LRSR Public Odds Finish
Ranking Ranking Position
A 1 1 1
B 2 2 2
C 3 3 3
D 4 4 4


The above race is perfect. The order of finish is perfect according to the speed ratings, as well as according to the public ranking.

So to compute our performance metric for this race, we would compare each pair of horses (all permutations -- a total of six comparisons for this race) and look at their finish positions. If horse A finished ahead of horse B, we then ask: did our factor also rank horse A ahead of horse B? The metric is the percentage of time the factor agrees with the final result over all pairs. In this race it is obviously 100%. We'll call this P1 for Performance Metric #1.

The other way to do it is simply to compare the winner to each loser (only three comparisons now) and that metric is the percentage of the losers that the winner is ranked better than factor-wise. Again in this example it is obviously 100%. We'll call that P2.

Now we want to compare P1 or P2 to how the factor relates to the betting. To do that we do the same thing as in P1, except the public rankings are substituted for the finish position. So this metric is the percentage of the time a horse with lower odds than another horse also has a higher factor value. We'll call this OC1 for Odds Comparison. And we'll also do OC2 where instead of using all the public rankings we just compare the favorite against the non-faves just like we did with the winner vs. the losers in P2. Again in the above example we get 100% for both. Kind of a lame example. Here's another:


Horse LRSR Public Finish
Ranking Ranking Position
A 2 1 1
B 1 3 2
C 3 2 3
D 4 4 4


If you work it out, this race gives us:

P1 = .83
P2 = .66

OC1 = .66
OC2 = .66


Still not a very exciting example. Anyway, so then you'd look at a whole mess of races and take the averages. The idea is that like I said above, ALL ELSE BEING EQUAL, if horse A is better on factor X than horse B, than horse A should have more money bet on it. So that's what we look at, one pair at a time. And since we're doing pair comparisons, with each of these metrics a value of 0.5 indicates that the factor is not relevant. So the strength is indicated by how far away from 0.5 the value is.

So doing this with some simple sample factors over 1000s of races in the last year gives us results like these:


Factor P1 P2 OC1 OC2
ML 0.643 0.708 0.841 0.908
BREAK 0.511 0.546 0.556 0.582
POSQ 0.539 0.595 0.602 0.654
POSH 0.560 0.611 0.630 0.687
PPR 0.540 0.592 0.596 0.644
PSR 0.642 0.702 0.786 0.858
LRSR 0.617 0.655 0.709 0.776

ML = morning line
BREAK = avg break pos
POSQ = projected position rank quarter mile
POSH = projected position rank half mile
PPR = projected pace rating
PSR = projected speed rating
LRSR = last race speed rating

Take a look at ML. It has a huge affect on how the horses will be ranked odds-wise, but is much less predictive of how they will actually run. The other factors all show the same trend, but to a lesser degree. After ML, the speed rating factors, being the most obvious & most predictive of performance, also show the most overcompensating/overbetting by the public, which was just my original point was about.

The holy grail of a factor would be one with P metrics above .5 but with OC at .5 indicating they are predictive but that the public takes no notice. (An actual negative relationship with OC values below .5 and P values above is too much to hope for -- a true demi-god of a factor.) Now then, these metrics are fairly simplistic since they don't take any account of the amount of gap factor-wise between horses, but you get the point:

The effect of a factor on the way the public bets the horses is greater than the actual predictive performance of that factor.

[Dave Schwartz]

So to compute our performance metric for this race, we would compare each pair of horses (all permutations -- a total of six comparisons for this race) and look at their finish positions. If horse A finished ahead of horse B, we then ask: did our factor also rank horse A ahead of horse B? The metric is the percentage of time the factor agrees with the final result over all pairs. In this race it is obviously 100%. We'll call this P1 for Performance Metric #1.

The other way to do it is simply to compare the winner to each loser (only three comparisons now) and that metric is the percentage of the losers that the winner is ranked better than factor-wise. Again in this example it is obviously 100%. We'll call that P2.

Interesting approach... also the one I used in the original ThoroBrain neural network I wrote back in 1990.

In that, I compared each winner to the other contenders in the race through the use of what I called "fact-pairs."

Thus, in a race with 5 contenders one would get 4 fact-pairs.

In fact, on the first pass, the neural net would actually seed itself with the weights equal to the "correctness percentage" of each factor. That is, if a factor was successful in (say) 90% of the fact pairs it received a 90 weight; a factor that was always right received a 100 weight.

No factor could ever start with a weight below 50% because if it did, you would simply select the reverse of the selection. (Huh? If playing the horse that had the most "early pace" actually resulted in less wins then the netowrk would would prefer the horse with the least early speed.

Of course, after that initial pass the neural net would lapse into a more classic back-prop function of adjusting weights, but this initial starting point was far ahead of starting with a total "random" set of weights.



LOL - Perhaps I misunderstood the question. I thought you wanted to discuss how the public wagers as opposed to how to weight factors to take advantage of how the public wagers.

Now, argueably, this would be an even more-interesting thread, but it seems that what you had in mind was different than what the thread title indicated.


Dave


Dave

[classhandicapper]

I have no database to prove this, but I've always felt that the pubic looks at the horse's entire record and changes the weights of each race depending on the horse. That's what I do!

If a lightly raced 3YO ran these figures: 75, 79, 79, 83 (83 being the latest), the public is going to put a lot of weight on the 83 because it's quite likely that this is an improving horse and an 83 or better is reflective of the what the horse will run today.

If an 8yo claimer ran these figures: 75, 79, 76, 86, the public is likely to look back in his record to see if he's put 2 86s back to back in the recent past or run other figures like it or better. Depending on that answer, the weight of each figure will vary. If the rest of the horse's recent record is filled with mid 70s figures, they will rate the horse better than another mid 70s horse without any 80s (because there is SOME PROBABILITY he will run another 80+), but perhaps below a horse that typcially runs low 80s.

In general, the public weights the horse's most recent race heaviest, next to last second heaviest, etc.... but it's always in light of the type of horse, the overall record, and other factors.

It believe the odds are a result of a form of pattern reading, but not as specific as with many sheet players and much more comprehensive than the sheet players.

[DJofSD]

I have no database to prove this, but I've always felt that the pub[l]ic looks at the horse's entire record and changes the weights of each race depending on the horse. That's what I do!

This assumes the majority of the public uses (and understands) the PPs in the DRF.

I don't think that's true. Unless things have changed radically, only a minority of the betting public actually have the entire record (or at least the last ten races) in front of them.

[classhandicapper]

This assumes the majority of the public uses (and understands) the PPs in the DRF.

I don't think that's true. Unless things have changed radically, only a minority of the betting public actually have the entire record (or at least the last ten races) in front of them.

That's true, but I have since learned something.

I believe the odds aren't really set by my grandmother and her friends at the Maspeth OTB reading tip sheets and following trainers/jockeys. All that unsophisticated money comes through the windows, but more sophisticated handicappers looking for overlays tend to focus their money on whichever horses are underbet. That has a way of making the odds more efficient.

[classhandicapper]

Game_Theory,

Do favorites that were not also morning line favorites outperform favorites that were also morning line favorites (win percentage and ROI)?

I believe to check that you'll have to do it by odds categories because I suspect the ML favorite - post time favorite group would tend to isolate many more huge favorites.

[andicap]

Stu,
did I miss this news on some thread somewhere -- That's way cool.

How did you get to do this???? What's your background?

:cool:

I am on my 4th day as the morning line maker for Ruidoso Downs (and 4th day ever as a morning line maker). The question that I ask myself is:

Q. "Who is the betting public?"
A. It depends who is taking our signal.


The first two of our races on Thursday and Friday are on TVG. With that in mind I am adjusting the odds to favor any horse that has recently run in California.

I am assuming that the percentage of bettors from California will over bet running lines containing BM, DMR, FPX, GG, HOL, LA, and SA. I plan on making the same horses a few points higher on races not aired on TVG.

I am open to constructive criticism on my morning line

[GameTheory]

LOL - Perhaps I misunderstood the question. I thought you wanted to discuss how the public wagers as opposed to how to weight factors to take advantage of how the public wagers.

Now, argueably, this would be an even more-interesting thread, but it seems that what you had in mind was different than what the thread title indicated.I don't know what I wanted to discuss -- just throwing some stuff out there. Doing that kind of pair analysis is what got me thinking about it though. I've done other studies to see how the public changes their betting over time, or seasonally -- also quite interesting. Originally, I thought maybe I could someone correct for these biases automatically, and you can for a single isolated factor, but when I tried to layer more factors in the whole thing fell apart, which is where I came up with my "errors canceling each other out" theory. So I have found no way to make any particular use of this data in practical terms unless I find some factors the public is totally ignoring...

[GameTheory]

As others have talked about the public betting on this and this, keep in mind that when I say "the public" I just mean the final odds, however they are arrived at. There are probably a small number of "market-makers" rather than a vast army of little old ladies that have the most influence over the odds and so therefore you're not really analyzing the patterns of the majority of the people, but rather the composite psychology of the market itself, if that makes sense.

Also, when isolating a factor, you can make up any factor at all -- even something that literally no one considers in that same form. You still do the same analysis, and the results represent however the public treats the factor in the form they are looking at it. So you are really looking at side effects of how they treat the data they have that is correlated to the data you are analyzing.

You can look all day at different factors looking for what the public stresses and what they don't. The interesting thing I thought was that they seem to overcompensate at least slightly for pretty much all factors, which would indicate that overlays are actually caused by unbalanced factors and the public underbetting negative factors.

Or maybe this is all too superficial and doesn't mean a damn thing, since the only way to analyze this way is to look at averages. Possibly the errors they make race-to-race are of a much different character and this is just an interesting side effect...

[Tom]

Does it really matter how the public bets? Or does it matter how the whales bet?

I can't tell you how many times I am betting a 4-1 shot at Fort Erie and suddenly it is 2-1, 8-5....last second stuff.

With big money looking for inefficincies and making last second wagers, can we really say anything about the public? Do the pools really represent the public anymore?

Ah, anything to stay inside today! :p

[GameTheory]

Does it really matter how the public bets? Or does it matter how the whales bet?

I can't tell you how many times I am betting a 4-1 shot at Fort Erie and suddenly it is 2-1, 8-5....last second stuff.

With big money looking for inefficincies and making last second wagers, can we really say anything about the public? Do the pools really represent the public anymore?That's what I mean -- when I say public it just means the odds, which really means something like:

final odds = 50% whales, 25% non-whale serious handicappers, 25% average joes

And since it might be hard to predict last minute changes in real-time, knowing where the odds usually end up based on this or that factor might end up being useful...

[sjk]

There is more randomness in the running of the race than in the making of the odds. The crowd rationally decides that the horse that looks good on paper should be the favorite. Many times as the race unfolds some other horse will finish in front.

[stu]

Stu,
did I miss this news on some thread somewhere -- That's way cool.

How did you get to do this???? What's your background?

:cool:

As of last week I am the new asst. racing secretary at RUI and ZIA. Making the morning line is one of my responsibilities.

Prior to being empolyed by RUI/ZIA, I worked in the racing office of RET/SUN/PRM as a placing judge/entry clerk.

Before that I failed miserably as a jockey agent.

Before that I was Unix Systems admin in corporate America.

[Dave Schwartz]

Tom,

Does it really matter how the public bets? Or does it matter how the whales bet?

I can't tell you how many times I am betting a 4-1 shot at Fort Erie and suddenly it is 2-1, 8-5....last second stuff.

This is not "whale" activity.

Whales are typically dutch players, spreading across multiple horses. As such, they typically do not slam a single horse. What you are seeing is the impact of the last-minute bets feeding into the tote system after the gate has opened. This group of late bettors is far more astute than the people who wager well before actual post.

Also, depending upon your tote feed, you may be getting a feed that is as much as 90 seconds behind. Suddenly, well after the gate opens, it catches up.

A good test is to look at the total money wagered at 0 minutes to post and compare to the total after the race is completed. At some tracks you may see as much as 60% appearing to come down late when, in fact, it is the result of slow initial reporting.

A more realistic figure is probably in the 45% range.

Dave




Dave

[BillW]

Stu,
did I miss this news on some thread somewhere -- That's way cool.

How did you get to do this???? What's your background?

:cool:

Andy,

I think you got the wrong Stu - this is not Don S. (Stuball) :)

Bill

[Jeff P]

posted by Dave Schwartz - There is no denying the impact of the last race on performance, but it would appear that a place to find the overlooked horse would be in the "Best Ever" area; the Past Class horse.Running a query against my 2005 JCapper database, here's what I have for Dirt Sprints only (with first time starters removed) broken out by rank for last race Bris Speed Figure:

By: 62 rankForLastRaceBrisFig

Rank Gain Bet Roi Wins Plays Pct Impact
1 -4235.60 30516.00 0.8612 4154 15258 .2723 2.1172
2 -4374.70 28484.00 0.8464 2611 14242 .1833 1.4257
3 -5777.80 27328.00 0.7886 1848 13664 .1352 1.0517
4 -5957.40 25752.00 0.7687 1337 12876 .1038 0.8075
5 -7560.50 24346.00 0.6895 1066 12173 .0876 0.6810
6 -6331.20 21348.00 0.7034 779 10674 .0730 0.5675
7 -5067.60 16612.00 0.6949 539 8306 .0649 0.5046
8 -5173.50 11114.00 0.5345 250 5557 .0450 0.3499
9 -2118.50 7074.00 0.7005 167 3537 .0472 0.3672
10 -1709.10 4220.00 0.5950 67 2110 .0318 0.2469
11 -725.90 2238.00 0.6756 45 1119 .0402 0.3127
12 -552.60 1036.00 0.4666 14 518 .0270 0.2102
13 -134.00 134.00 0.0000 0 67 .0000 0.0000
14 -68.00 68.00 0.0000 0 34 .0000 0.0000
15 -4.00 4.00 0.0000 0 2 .0000 0.0000
16 -2.00 2.00 0.0000 0 1 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000
19 0.00 0.00 0.0000 0 0 .0000 0.0000


And here's a slightly different twist using the same set of horses - this time broken out by rank for Bris Speed Figure best of last 10:

By: Bris Speed Figure Rank (Best of Last 10)

Rank Gain Bet Roi Wins Plays Pct Impact
1 -3523.80 30986.00 0.8863 4034 15493 .2604 2.0248
2 -4533.50 29144.00 0.8444 2725 14572 .1870 1.4542
3 -6123.60 27852.00 0.7801 1951 13926 .1401 1.0895
4 -5720.90 26016.00 0.7801 1507 13008 .1159 0.9009
5 -5995.60 24356.00 0.7538 1126 12178 .0925 0.7190
6 -6980.20 20954.00 0.6669 707 10477 .0675 0.5248
7 -5840.10 16090.00 0.6370 418 8045 .0520 0.4040
8 -4963.70 10856.00 0.5428 207 5428 .0381 0.2966
9 -2563.80 6758.00 0.6206 118 3379 .0349 0.2716
10 -1847.70 3990.00 0.5369 57 1995 .0286 0.2222
11 -1101.60 2156.00 0.4891 19 1078 .0176 0.1371
12 -514.70 948.00 0.4571 6 474 .0127 0.0984
13 -78.00 122.00 0.3607 1 61 .0164 0.1275
14 -5.20 48.00 0.8917 1 24 .0417 0.3240
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000
19 0.00 0.00 0.0000 0 0 .0000 0.0000


Best of last 10 has a better flat bet win roi for the top spot than best last line... which supports what Ritchie said earlier in this thread about boosting roi by getting away from the last line... which is also supported by what Dave took the time to post.

Still, look at the roi numbers to 1.00 that I posted. And look at the 2.00 roi numbers that Dave posted... everything is in the 80's. This just confirms what I've come to believe about final time based speed figures. Yes, they are predictive - no matter who makes them. Unfortunately that's about as far as they get you.... Unless you make your own and keep them to yourself. For the most part they are seen and used by too many pairs of eyes each day to produce the kind of roi needed to consistently grind out profits. In short, they are one of the many positive factors that the public sees and overreacts to.

IMHO, players interested in profits need to look elsewhere.

What if, instead of Final Time based Speed Figures, we focus on pace figures? Nothing fancy ...not compound figures.. but just something simple like Bris E2... Looking at the same set of horses as above broken out by both last line Bris E2 and E2 Best of Last 10 here's what I have:

By: 74 rankForPaceFig_4F_InLast

Rank Gain Bet Roi Wins Plays Pct Impact
1 -3162.50 31936.00 0.9010 3619 15968 .2266 1.7625
2 -4660.60 28282.00 0.8352 2401 14141 .1698 1.3204
3 -7281.60 27266.00 0.7329 1815 13633 .1331 1.0353
4 -6492.00 26252.00 0.7527 1530 13126 .1166 0.9064
5 -7493.10 24410.00 0.6930 1221 12205 .1000 0.7780
6 -6529.10 21432.00 0.6954 948 10716 .0885 0.6880
7 -5914.60 16222.00 0.6354 602 8111 .0742 0.5772
8 -3446.00 10862.00 0.6827 380 5431 .0700 0.5441
9 -2102.70 6618.00 0.6823 205 3309 .0620 0.4818
10 -1480.00 3950.00 0.6253 93 1975 .0471 0.3662
11 -930.40 2060.00 0.5483 39 1030 .0379 0.2945
12 -157.80 844.00 0.8130 24 422 .0569 0.4423
13 -106.00 106.00 0.0000 0 53 .0000 0.0000
14 -36.00 36.00 0.0000 0 18 .0000 0.0000
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000
19 0.00 0.00 0.0000 0 0 .0000 0.0000



By: Best E2 Rank

Rank Gain Bet Roi Wins Plays Pct Impact
1 -3126.90 32422.00 0.9036 3655 16211 .2255 1.7533
2 -4480.00 28494.00 0.8428 2505 14247 .1758 1.3673
3 -5652.70 27244.00 0.7925 1903 13622 .1397 1.0864
4 -5775.80 25876.00 0.7768 1542 12938 .1192 0.9268
5 -6715.00 24034.00 0.7206 1227 12017 .1021 0.7940
6 -7170.80 20842.00 0.6559 831 10421 .0797 0.6201
7 -6108.20 16292.00 0.6251 568 8146 .0697 0.5422
8 -4143.10 10956.00 0.6218 314 5478 .0573 0.4458
9 -2617.00 6688.00 0.6087 193 3344 .0577 0.4488
10 -2127.50 4126.00 0.4844 83 2063 .0402 0.3129
11 -1274.50 2184.00 0.4164 38 1092 .0348 0.2706
12 -476.70 950.00 0.4982 16 475 .0337 0.2619
13 -74.20 118.00 0.3712 2 59 .0339 0.2636
14 -50.00 50.00 0.0000 0 25 .0000 0.0000
15 0.00 0.00 0.0000 0 0 .0000 0.0000
16 0.00 0.00 0.0000 0 0 .0000 0.0000
17 0.00 0.00 0.0000 0 0 .0000 0.0000
18 0.00 0.00 0.0000 0 0 .0000 0.0000
19 0.00 0.00 0.0000 0 0 .0000 0.0000

Comparing E2 pace figures to Final Time Speed Figures, notice how the win percent (predictive value) of the top slot goes down but how the flat bet win roi actually goes UP?

Taking things a step further, look at what happens if we further restrict our analysis to just those horses with at least a 2 point advantage over any horse in the field as measured by E2 Best of Last 10:

UDM Definition:
Profile: OnThePath
UDM Divisor: 999
Surface Req: D* Distance Req: S
Running Style: ALL
BestE2: MinRank= 1 MaxRank= 1 MinVal= -999 MaxVal= 999 MinGap= 2 MaxGap= 999


Data Window Settings:
Divisor = 999
Filters Applied: LT0-
Dirt (All*) SPRINTS (From Index File: D:\2006\Q2_2006\pl_Complete_History.txt)


Data Summary Win Place Show
Mutuel Totals 17948.60 17450.20 16938.90
Bet -19190.00 -19190.00 -19190.00
Gain -1241.40 -1739.80 -2251.10

Wins 2384 4121 5328
Plays 9595 9595 9595
PCT .2485 .4295 .5553

ROI 0.9353 0.9093 0.8827
Avg Mut 7.53 4.23 3.18


Suddenly the picture brightens. The flat bet win roi is almost high enough as is to churn a profit if we add in a rebate. I have very little doubt that further research into horses with a true Pace Advantage over their fields can turn up any number of mechanical spot plays that will prove to be profitable going forward. Why? Because PACE is something that the public (which includes whales as well as Aunt Erma and Uncle Joe) collectively underbets.

-jp

.

[formula_2002]

Game_Theory,

Do favorites that were not also morning line favorites outperform favorites that were also morning line favorites (win percentage and ROI)?

Based on the following, the answer is yes.



In about 1153 RECENT RACES;

1.
Ml favorites won 29% of them, expect wins were 37.6%
Roi= .77

Post time favorites won 33% of them, expected wins were 41.5%.
Roi=.80

My QUANTUM favorite won 30% of them, expected wins were 35%
Roi= .86


2.
When post time favorites were not the ML favorite;
In 418 races, post time fav won 30% of them, expected wins were 35%
Roi= .86

In 201 race Quantum favorite won 25% of them, expected wins were 28%
Roi=.89.

3.
When post time favorites were the ML favorite;
In 731 races, post time fav won 45% of them, expected wins were 35%
Roi= .78

In 909 race Quantum favorite won 31% of them, expected wins were 31%
Roi=.77

Note:
Expected win % is that % required to break even. It’s the sum 1/(odds+1)

[classhandicapper]

Thanks formula_2002. I appreciate those stats a lot. They were along the lines of what I was expecting, but I don't think I've given this kind of thing enough thought. THere may be a way to incorporate some of this into some of what I am already doing.

[formula_2002]

In about 1153 RECENT RACES;


In 909 race Quantum favorite won 31% of them, expected wins were 31%
Roi=.77


one of those numbers is obvioulsy incorrect..I'm away from the figures so I can't tell which.

[Fastracehorse]

I don't discount the importance of the trainer, but I lean toward working with the jockey because of combining the influence and/or intentions of the trainer with being the one who has direct control of the horse during the running of the race (in addition to some jockeys being undeniably more talented than others), and because a smaller percentage of the local population accounts for a greater share of victories with jockeys than with trainers. (At the same time, though, I believe that the public can go overboard too often on hot riders, and bet their mounts down to odds that are lower they should be.)

My point was that a good jockey is over-bet because he is a good jockey. If the horse has no engine underneath it doesn't matter who is on him.

Of course, a sharp charge with a talented jock is less likely to find trouble than with a mid-pack mount. But I don't care who is riding my horses if I like them.

fffastt

[stu]

Andy,

I think you got the wrong Stu - this is not Don S. (Stuball) :)

Bill


I have been called Stuart, Stu, Stu Sausage, and Studebaker but never Stuball.

[Pgh. Gere]

GameTheory,

Seems to me the public overbets the last race, specifically the finish position.Often the horse that finished 2nd last time out may be the top speed figure, but not always. Cardello in "Speed To Spare" has an example like this, where the top Beyer wasnt the favorite, the public bet a different horse that had finished second in its last race, but had a lower figure.

I don't have the database that you or Dave have, but IMO, the public probably emphasizes finish position more than speed figures. Finishing 2nd or 3rd in good time or against a higher class, probably a positive factor, but what about the plodder or closer that always seems to finish 2nd or 3rd. The public may fail to make the distinction between the two, thus creating value.

[mcikey01]

Because PACE is something that the public (which includes whales as well as Aunt Erma and Uncle Joe) collectively underbets.

-jp

.

Does the public collectively underbet pace in all types of races? Are there differences depending on class or distance of race, for example? How about field size?

Consider some data from the Thorostats site covering all tracks for nearly a 2 year period:

STKS-DIRT-SPRINTS

Rating --- Races --- Winners --- Win% --- Avg.Win --- Return --- ROI ---



Pace --- 1775 --- 518 --- 29% --- $5.48 --- $2837 --- -20% ---





STKS-DIRT-ROUTES

Rating --- Races --- Winners --- Win% --- Avg.Win --- Return--- ROI ---



Pace --- 1467 --- 441 --- 30% --- $6.43 --- $2834 --- -3% ---




Seems to indicate that pace may be overbet in Stks Dirt Sprints and underbet in Stks Dirt Routes.

Jeff,
what do your studies show by class, distance and field size?

[classhandicapper]

Those stakes stats are extremely interesting.

[RXB]

Does it really matter how the public bets? Or does it matter how the whales bet?

I can't tell you how many times I am betting a 4-1 shot at Fort Erie and suddenly it is 2-1, 8-5....last second stuff.

With big money looking for inefficincies and making last second wagers, can we really say anything about the public? Do the pools really represent the public anymore?



I bet a horse at Canterbury tonight that was 9/1 as they loaded into the gate. The odds flash when the horses raced into the backstretch: 4/1. I almost vomited. I was actually kind of pleased when another horse got up in the final 50 yards to beat her.

[Niko]

Canterbury is terrible....I had a 19-1 that went to 11-1 1/2 way through the race when the pools came in? Many 5-2's go to 7-5 etc....Must be smart handicappers because it's always the lead horses that run well that come down.....?? I like the track though, the full fields and competitive racing.

I had one or two winners that came from off the pace and went off a tick higher but very, very rare

[mcikey01]

Those stakes stats are extremely interesting.

Here's more from Thorostats database for last 2 years that highlights the public's underebetting/overbetting certain factors:


2 Y/O__ STKS, HCP, ALW__ DIRT SPRINTS



Rating Type_ Races__Winners__Win%__Avg Win $__ROI

Connections -- 894-------236-------26%-------$7.49------(-1%)

Top Rated Horse







2 Y/O__ STKS, HCP, ALW__ DIRT ROUTES



Rating Type_ Races__Winners__Win%__Avg Win $__ROI

Connections -- 267--------64-------23%-------$6.31------(-24%)

Top Rated Horse

[Fastracehorse]

GameTheory,

Seems to me the public overbets the last race, specifically the finish position.Often the horse that finished 2nd last time out may be the top speed figure, but not always. Cardello in "Speed To Spare" has an example like this, where the top Beyer wasnt the favorite, the public bet a different horse that had finished second in its last race, but had a lower figure.

I don't have the database that you or Dave have, but IMO, the public probably emphasizes finish position more than speed figures. Finishing 2nd or 3rd in good time or against a higher class, probably a positive factor, but what about the plodder or closer that always seems to finish 2nd or 3rd. The public may fail to make the distinction between the two, thus creating value.

Also,

Trainers can set up a nice price by the public being thrown off by one bad race: A form darkener - ability doubter - an ugly one.

fffastt

[dastar]

Please explain EXACTLY what you mean by "Public Betting"

Do you infer: Everyone in the Horse Racing industry & everyone else is the public?

This is one issue that has never been fully explained to my total satisfaction, and may never be.

Here is the big reason why.

No matter who wins, the racetrack will get their percentages.

Are the race tracks betting? :cool:

Obviously some stables bet, so, they definitely can be included in "Non Public."

Jockeys
Agents
Have to be "Non Public"

The owners of the horses, may fall into both categories:
They can bet on any horse they choose to.
I imagine, if an owner is a bettor, and his horse doesn't look like he's ready, or the trainer informed him of such, the owner may bet on another horse.

But, if any one is not connected to anypart of the Racing Industry, does that automatically make them the public?

"Smart Money" is suppose to mean "Inside" money, but that is certainly no guarantee of a win.

If betting Syndicates exist, are they connected to the Racing Industry?
They certainly don't always win.

For what it is worth, I have always believed, you can keep a horse from winning, but it is much harder to get a horse to win.

This brings me to my last observation:
2 seperate trainers, in the same race are making bets on their respective horses.

The race is deceided by inches, and one of the trainers just lost a few thousand bucks or so: That trainer who just lost: Is he considered "Public" for that bet? :confused:

In closing, when you infer studying what the public looks for, you Must determine who is really the public.



dastar

[Overlay]

To me, the "public" is everyone who bets on the race, no matter who they are. It is their aggregate final opinion (the post-time odds) that provides the basis for judging whether a horse is offering value or not, depending on the bettor's own assessment of its winning chances. And I agree with your statement that it's easier to keep a horse from winning than to guarantee that a particular horse will win. That's why I tend to discount the idea or significance of "smart money". Yes, a horse with darkened form may be "trying" today, but is every other trainer in the race (especially those with horses that are fit and ready to win based on their published form) going to just sit back and let it win (in the absence of massive collusion and payoffs)? And if you believe that that's how the outcomes of races are determined, what's the point of trying to beat the game by handicapping?

[GameTheory]

Public = everyone that wagers into the pari-mutuel pools. If it affects the odds, that's the public. Public = toteboard odds.

[rrbauer]

Dave Schwartz wrote:
"Whales are typically dutch players, spreading across multiple horses. As such, they typically do not slam a single horse. What you are seeing is the impact of the last-minute bets feeding into the tote system after the gate has opened. This group of late bettors is far more astute than the people who wager well before actual post."

Comment:
Depends upon what you mean by "astute". Play Tampa Bay or Monmouth. Track the odds from the first flash. You will learn about a lot of "hidden" factors that will not be publicly evident from any other source about the race. At these two tracks (and probably more that I don't follow) disregard this information at your own peril. These are the backstretch "minnows" talking to you. Pay attention!

BTW Good thread!

[twindouble]

Dave Schwartz wrote:
"Whales are typically dutch players, spreading across multiple horses. As such, they typically do not slam a single horse. What you are seeing is the impact of the last-minute bets feeding into the tote system after the gate has opened. This group of late bettors is far more astute than the people who wager well before actual post."

Comment:
Depends upon what you mean by "astute". Play Tampa Bay or Monmouth. Track the odds from the first flash. You will learn about a lot of "hidden" factors that will not be publicly evident from any other source about the race. At these two tracks (and probably more that I don't follow) disregard this information at your own peril. These are the backstretch "minnows" talking to you. Pay attention!

BTW Good thread! rrbaurr.

Please explain what these "hidden factors" are.

Personally I never cared what other people are doing with there money, betting late, betting early or betting a lot. Most of all I never paid any attention to what comes from the barns or how the owners are dressed.

Go to any track, there's on the average of 90 to 100 horses that will be running on a given day, one would be a fool to listen many different people that have "inside information" on more than one horse in any given race that's for sure. Then to top it off be following all the money the so-called wise guys run through the tote or from teller that informs you so and so just bet a few grand on a horse.

Common sense tells me no one is going to come up to you and say here's a chance for you to make a few hundred on my horse or my connection unless there's strings attached.

If your purely handicapping the thieves when you say, "hidden factors" I'll give merit to what your saying.




T.D.

[Tom]

What good is "inside information?"
They run outside!

[twindouble]

What good is "inside information?"
They run outside!

I'm not saying knowing if a horse is fit, sound and fits the race well isn't good information if it comes from a reliable source. Note I said "fits the race", that don't mean he's a sure thing. In most cases there's others in the race, that are geared up as well.





T.D.

[rrbauer]

rrbaurr.

Please explain what these "hidden factors" are.

Personally I never cared what other people are doing with there money, betting late, betting early or betting a lot. Most of all I never paid any attention to what comes from the barns or how the owners are dressed.

Go to any track, there's on the average of 90 to 100 horses that will be running on a given day, one would be a fool to listen many different people that have "inside information" on more than one horse in any given race that's for sure. Then to top it off be following all the money the so-called wise guys run through the tote or from teller that informs you so and so just bet a few grand on a horse.

Common sense tells me no one is going to come up to you and say here's a chance for you to make a few hundred on my horse or my connection unless there's strings attached.

If your purely handicapping the thieves when you say, "hidden factors" I'll give merit to what your saying.




T.D.

The first flash on the odds board for each race comes from the host-track betting pool only. That flash is concurrent with the host track notifying the pari-mutuel network that the race is open to receive bets via the network. Whereas, bets made prior to a race being opened to the network at non-host-track sites are captured and held at the location that serves as a network "hub" for each non-host-track betting site. Those early bets are transmitted to the host-track pool only after the network has been notified that the race is current and open to accept wagers.

I've found that the first-flash odds information means more at some tracks than others. Tampa and Monmouth are two tracks where it is relevant when it represents a large betdown from the morning line and seems to overstate a horses chances vis-a-vis the generally available handicapping information. It's hidden in the sense that it's only available from one place and at one time. It's not in the pp's. It's not coming from the owners in the walking ring or at the bar. It's not being influenced by what the other bettors are doing at race time. As information, it's there and then it's gone.

You can pooh-pooh it all you want. I know what works for me.

[twindouble]

The first flash on the odds board for each race comes from the host-track betting pool only. That flash is concurrent with the host track notifying the pari-mutuel network that the race is open to receive bets via the network. Whereas, bets made prior to a race being opened to the network at non-host-track sites are captured and held at the location that serves as a network "hub" for each non-host-track betting site. Those early bets are transmitted to the host-track pool only after the network has been notified that the race is current and open to accept wagers.

I've found that the first-flash odds information means more at some tracks than others. Tampa and Monmouth are two tracks where it is relevant when it represents a large betdown from the morning line and seems to overstate a horses chances vis-a-vis the generally available handicapping information. It's hidden in the sense that it's only available from one place and at one time. It's not in the pp's. It's not coming from the owners in the walking ring or at the bar. It's not being influenced by what the other bettors are doing at race time. As information, it's there and then it's gone.

You can pooh-pooh it all you want. I know what works for me.

I wasn't pooh-pooh-ing anything. Your first post sounded like there's big mystery going on that no one is aware of when picking winners. Tote conspiracies is what I call it, in other words the wise guys are jerking your chain and you are beat before you read the form.

I will say this, if I jumped onto late money or early money nowadays, I wouldn't have lasted as long as I have in this game.


Good long term luck,


T.D.

[prank]

I submit that the public overbets all positive factors, and underbets all negative factors. And by "all", I don't actually mean all, but the factors they actually consider, or suitable proxies for what they consider. (They may be looking at Beyer figs as a factor, but if in my research I look at BRIS figs instead and see how they correlate to the betting, the effect will be similar.)

In other words, if you isolate single factors (like "last race speed figure" or "trainer win%" or even your own composite factor that no one else has access to) and compare how that single factor performs in terms of predicting winners while at the same time comparing how that same factor predicts the final odds, what will you find? In most cases you will find that they overbet or underbet the factor (same thing really depending on which way you look at it) in a predictable fashion. A high trainer win% tends to pick winners, but when looking at the betting you will find it doesn't pick as many as the betting on that factor would indicate. Same with any other positive factor. With a negative factor, the actual performance of the horses is better than the betting would indicate. Again, the negative side is just the other side of the same coin -- low trainer win% is a negative factor.

So with any predictive factor, the more positive or negative it gets, the more the public will overdo it betting-wise in that direction. I would then conclude that the quite high accuracy the public achieves is not due to their great capacity to weigh each factor properly and get it just right, but rather due to the fact that they are considering many factors, both positive and negative, and it is their errors of overcompensation that cancel each other out. So if they see something positive about a horse they will tend to give it too much credit for that factor, but if they then see something negative about that same horse they will then deduct too much from his winning chances based on that. The overall effect is that the errors cancel each other out and they get it about right in the end.

What does all this mean? Maybe nothing, but maybe that good betting opportunities come when a horse has an unbalanced number of positives & negatives (or unbalanced amount of aggregate perceived magnitude of positives & negatives). If a horse has more apparent negatives than positives, it will probably be underbet (although it still won't win very often). And if it has a lot of positives and less apparent negatives, it will probably be overbet (although it will still win often).

This sort of thinking is probably more useful for finding overbet favorites than uncovering winners...

This is a very interesting idea. Indeed, suppose we're not interested in predicting the outcome of the race, but the outcome of the betting. I wonder how much more accurate one can be at that, than in predicting the race itself.

I think that this kind of thing can be done, and done fairly well. In fact, over the long run, one could look at predicting which quantiles the horses fall into based on their own features and the features of the race.

It reminds me of some recent work on predicting eBay sale prices, based on things like the category of the item being sold (analogous to, say, the type of the race), information on the seller (jockey, trainer, owner...), information on the number of comparable items being sold at the same time (competing horses), the date of the sale (weekend/weekday, seasonal stuff), etc. It's a pretty active area of economic research, actually: eBay data makes for a huge "lab" for economists' ideas to be validated, and for quantification of which factors most affect auction prices and behavior.

If anyone knows of this analysis with regard to races, I'd really love to know about it. I guess it somewhat enters into questions of market efficiency, since inaccuracy in gambling are indications of inefficiences, and the larger the inefficiency the more questionable is the behavior. (And, the more money foolish money is available for the taking by those of you doing the smart things. :))