Exceptions to the Wisdom of Crowds?

[podonne]

So I was reading the Wisdon of Crowds the other day and the guy (like many, many others) talked about how horse racing is pretty much a good example of how the aggregate decisions of many people do a pretty good job of correctly estimating a horse's chance in a race (Longshot bias being the exception).

I'll buy that, but it stands to reason that the crowds don't get every type of race equally right, there are probably races they get very wrong and some they get very right, but in aggregate it all evens out. So has anyone ever looked at, say, the favorite win% in different types of races.

How about at different tracks? It seems likely that each track is basically composed of the same players over time betting the same way, so certain tracks should be "better" and certain tracks should be "worse"? Are certain tracks more likely to over\under emphasize a certain trait?

Cheers,
Philip
-----------------------------------------------------------------------
Whenever I read something I think about horse racing... Is that an addiction?

[point given]

"So I was reading the Wisdon of Crowds the other day and the guy (like many, many others) talked about how horse racing is pretty much a good example of how the aggregate decisions of many people do a pretty good job of correctly estimating a horse's chance in a race (Longshot bias being the exception)."

I've not heard of a " long shot bias " Could you explain ? I've seen speed and closer bias, inside / ooutside bias , but Long shot bias ? How can this be ? I would think it is a coincidence when alot of chalk or longshots come in on a card. Then the numbers are crunched at meets end and you've got around 33%
chalk.

[podonne]

Sure point given,

Properly called the Favorite\Longshot Bias, as I understand it the betting public has a tendency to overestimate the probability of a long-shot winning and underestimate the favorite. There are alot of differnt thoeries as to why, but it's been proven time and time again.

[misscashalot]

So I was reading the Wisdon of Crowds the other day and the guy (like many, many others) talked about how horse racing is pretty much a good example of how the aggregate decisions of many people do a pretty good job of correctly estimating a horse's chance in a race (Longshot bias being the exception).

I'll buy that, but it stands to reason that the crowds don't get every type of race equally right, there are probably races they get very wrong and some they get very right, but in aggregate it all evens out. So has anyone ever looked at, say, the favorite win% in different types of races. ...

Cheers,
Philip
-----------------------------------------------------------------------
Whenever I read something I think about horse racing... Is that an addiction?

Your reasoning is correct. In my database I have the bet ranking of winners, as well as class, age, gender, distance etc etc. and sort the race eaach race type for the NYRA that day to see if any race favors well bet or the contrary. . As an example prior to the running of the Wood I posted here that in the last 16 running, 8 were won by favs and 7 by second bet choice. (94%) Females races are more predicatable than males. 4 yr up maiden races most formful as far as winning well bet horses. There's a whole lot of unlevel info you can draw from this study.

[podonne]

I've not heard of a " long shot bias " Could you explain ? I've seen speed and closer bias, inside / ooutside bias , but Long shot bias ? How can this be ? I would think it is a coincidence when alot of chalk or longshots come in on a card. Then the numbers are crunched at meets end and you've got around 33%
chalk.

To answer your question though, it's a trait of betting markets, not something that effects how a horse runs a race or who wins.

[Edward DeVere]

According to Ainslie, the most extensive studies in this area "show that the percentage of winning favorites remains essentially the same in all types of races."

[point given]

To answer your question though, it's a trait of betting markets, not something that effects how a horse runs a race or who wins.

Interesting topic and one which I have given nary a thought. As is evident !
I handicap this territory by my years of handicap experience and the type of bets I make. I stay away from some, tread lightly in others and dig in when I see my type of race. Never bet enough on the winners though ! If only I knew when my opinions were correct beforehand :D

Do they have an after race / pre race handicap pill invented yet ?

[Overlay]

In Modern Impact Values, Mike Nunamaker found that, in his overall sample, favorites won 33.04% of the time. Odds-on favorites (below even money) won 51.13% of the time, and favorites at even money or higher won 29.73% of the time. Broken down by individual race categories, the percentages for those three types of horses were as follows:

All Dirt sprints: 32.89/51.46/29.53
All Dirt routes: 33.61/50.78/30.37
All Turf routes: 31.85/51.63/28.60
Dirt sprints for maidens, three years old and up: 33.63/53.39/29.91
Dirt sprints for two-year-olds: 32.58/50.56/28.78
Dirt sprints for non-maidens, three years old and up: 32.72/51.00/29.57
Dirt routes for maidens, three years old and up: 34.38/50.00/31.29
Dirt routes for two-year-olds: 44.91/59.32/39.49
Dirt routes for non-maidens, three years old and up: 32.68/49.89/29.65
Turf routes for maidens, three years old and up: 29.22/48.39/24.39
Turf routes for non-maidens, three years old and up: 32.29/53.19/29.16

Looks like, in the aggregate, dirt routes for two-year-olds were the easiest to figure, and turf routes for older maidens were the most difficult, with less variance among the other categories.

[BeatTheChalk]

Interesting topic and one which I have given nary a thought. As is evident !
I handicap this territory by my years of handicap experience and the type of bets I make. I stay away from some, tread lightly in others and dig in when I see my type of race. Never bet enough on the winners though ! If only I knew when my opinions were correct beforehand :D

Do they have an after race / pre race handicap pill invented yet ?


I can add nothing to the posting I quoted ABOVE! It is right on ! Of course..some purist will come along and say " Earl, you already added something ..by saying that you weren't going to add something..."

[misscashalot]

According to Ainslie, the most extensive studies in this area "show that the percentage of winning favorites remains essentially the same in all types of races."

wrong in the nyra circuit anyway. some races types may have 10% win favs, and other 60% just 2 examples

Fem 2 yr MSW 7furlong Bel Dirt Sep-Oct 22/39 win fav 56%

same type but 1 mile Bel Widener Turf Sep-Oct 3/33 win fav .09

and there's a lot more like this

[betovernetcapper]

Something I find interesting is the ability of some trainers to win with favorites. Larry Eilers started 20 favorites in the last year and won with 14 of them. That's seventy percent winners with a $1 ROI of $1.56.

[Overlay]

Something I find interesting is the ability of some trainers to win with favorites. Larry Eilers started 20 favorites in the last year and won with 14 of them. That's seventy percent winners with a $1 ROI of $1.56.

Sounds like a case of, "Which came first, the chicken or the egg?" I understand your point, but unless the trainer by himself is responsible for making the horse the favorite (and what individual (even a betting trainer) has that kind of capital?), I think it's the public that gets the credit. Maybe Mr. Eilers has a particular conditioning pattern that characterizes his well-meant horses, and the public has caught on to it. Or are you perhaps suggesting that his horses (for whatever reason or at whoever's prompting) "try harder" when they're favored?

[Edward DeVere]

MissCashALot - Your samples are much too small. Ainslie agrees that small samples can say anything, but says that the truly large studies shows that favorites vary little over class.

You need to look at thousands and thousands of races, and, I would imagine, correct for field size.

[misscashalot]

MissCashALot - Your samples are much too small. Ainslie agrees that small samples can say anything, but says that the truly large studies shows that favorites vary little over class.

You need to look at thousands and thousands of races, and, I would imagine, correct for field size.

Edward DeVere. This is from a datebase of 40,000 races. Are you saying that if you had 2 of the races as I described above you would not find those results are having no bearing on your thought process?

Here are 2 more samples:

F 3u G1 1-1/16 BD 17/28 win fav .607%

same 1-1/4 Bel Inner Turf Zero/16

These have no effect on you?

[betovernetcapper]

Overlay-your right that the public deserves the credit for making a horse the favorite. The credit for getting that favorite to win goes in some measure to the trainer. Glancing through today's trainer stats I found a couple guys that were 0/6 with favorites in the last year. If I were going to bet a favorite, because he's a favorite, I'd stick with guys that had competence in the favorite area.

[singunner]

This is a subject I'm quite curious about. Unless you have some measure by which to determine the individual probability of every horse in the race, you have no means of determining the accuracy of the crowd. In this case, saying that the crowd is good at its picks just means that the average odds for first place are close to the probability of first place actually coming in.

My question is this: Based on individual odds divisions, how accurate are the results? Say the odds dictate that the horse should win 50% of the time (after taking into account takeout), what percent of the time do those horses actually win, eh? On top of that, what's the standard deviation? That is, how much does it see-saw back and forth over ideal (assuming it is even ideal)?

[Robert Fischer]

When you turn your critical eye to a horse you should be doing all these things.

The odds do not influence your opinion of probability,- they influence your wagering strategy and play an important role in value.

I do use general factors of high value races as one of my decisions on whether or not to spend time handicapping a race.

The majority of factors follow logical reasoning. Nunmaker says 3up maiden turf routes? Why is that? Something in the grass?...well they do average 10horse fields and you have pedigree issues, you have first time turfers, late bloomers,some of these supertrainers arent turf specialists etc... Then you can move on and say "hey, im not betting favorites in those races!" or you can get into the race specific factors.

[podonne]

Overlay-your right that the public deserves the credit for making a horse the favorite. The credit for getting that favorite to win goes in some measure to the trainer. Glancing through today's trainer stats I found a couple guys that were 0/6 with favorites in the last year. If I were going to bet a favorite, because he's a favorite, I'd stick with guys that had competence in the favorite area.

I don't think you could find a cause\effect between the ability of a trainer and his ability to win when his horse is a favorite. These are two different things....

However, going back to my original post, this is about a particular segment of the betting public's tendency to over or under estimate a particular horse's chances in a particular type of race. I'd say the example above (0/6) just illustrates that whatever common traits this trainer's horses show, the betting public at that track tends to overestimate the effect of those traits.

Prices are are function of the markets. Trainer's can't set the odds on thier horses, but they can influence them. How then does the market respond to those influences?

You also don't have to use favorites, it's just a statistical shortcut. You could use the deviation from a horse's expected chance (post time odds) with the actual probability (given a range of post time odds, what are the actual chances?). This is what they are doing when they divide a "favorite" into "favorite < 1/1" and "favorite > 1/1" groups.

[Overlay]

My question is this: Based on individual odds divisions, how accurate are the results? Say the odds dictate that the horse should win 50% of the time (after taking into account takeout), what percent of the time do those horses actually win, eh? On top of that, what's the standard deviation? That is, how much does it see-saw back and forth over ideal (assuming it is even ideal)?

In Winning at the Races (1979), William Quirin found that, as a general rule, horses win in accordance with the formula (1 - Mutuel take %) / (Tote Odds-to-$1.00) + 1.05 (assuming dime breakage). For example, with a 17% take, a horse that was 4-1 on the toteboard would be expected to have a winning probability of (1-.17)/(4 + 1.05), or .164 (16.4%). The only odds ranges that showed a statistically significant variation from the expected number of winners (1-5, 1-1, 5-2, and 6-1 on the positive side; and 3-1 on the negative side) were too isolated or dispersed to indicate any overall pattern of deviation, so Quirin concluded that (in general) horses do win in accordance with their odds. (Of course, that says nothing about the accuracy of the odds on any specific horse in any given race. The horse's odds may be higher or lower than its true chance of winning.)

[singunner]

Thanks Overlay. The next question would be whether the distribution is normal and what the standard deviation is.

[podonne]

Yes, that is very good, but that point is pretty well understood, the horse racing market is pretty "efficient" (for the economists out there). My question is that in any "efficient" market there are bound to be segments (subsections) that are more or less efficient than others, so has anyone done this analysis for the horse racing market?

I've seen some previous on this thread by race type, does anyone have it by track? Each track should represent a distinct and repetitive group of participants, different from any other track, and may show certain tendencies that make those market more or less efficient.

[betovernetcapper]

Brisbet and DRF each publish the percentage of winning favorites by meets.

The Capper (Netcapper) lists the favorite winning percentage-roi-and impact value of each race by grouping. The groupings are track-age-sex-conditions/class-distance and surface. Here are the impact values for today at OP

Race # IV
___________
1 --- 1.12
2 --- - .97
3 --- 1.08
4 --- - .94
5 --- 1.26
6 --- 1.09
7 --- 1.08
8 --- 2.02
9 - --- .99
10 --- .89

[ALL CIRCUITS GO]

The Wisdom of the Crowds is not applicable for horse racing because.....The main theory is that if each persons vote is given equal weight, then the crowds determination is found to be pretty accurate. However, in pari-mutuel betting, the odds are determined not by the number of people who placed a bet, but by the total amount bet.

Therefore, a favorite can be born out of 100 wagers(separate distinct people) at $2 and 1 wager at $1000, for a total vote count of 101 and total wagered of $1200;; while another entrant could have 500 $2 wagers (separate distinct people) for a total vote count of 500 and total wagered of $1000.

Does having a fat bankroll and the ability to wager $1000 make your wager (vote) more likely to succeed? I think not. I would rather be on the side of the 500 voters.

I can't wrap my head around it, but I would think that this is why only 33% of favorites win, and the 'crowd' is 'wrong' 66% of the time.

If only I could have the statistics that show how many people wagered (voted) on a horse, and then follow that Wisdom of the Crowds!

Am I making any sense??

[Robert Fischer]

The Wisdom of the Crowds is not applicable for horse racing because.....The main theory is that if each persons vote is given equal weight, then the crowds determination is found to be pretty accurate. However, in pari-mutuel betting, the odds are determined not by the number of people who placed a bet, but by the total amount bet.

Therefore, a favorite can be born out of 100 wagers(separate distinct people) at $2 and 1 wager at $1000, for a total vote count of 101 and total wagered of $1200;; while another entrant could have 500 $2 wagers (separate distinct people) for a total vote count of 500 and total wagered of $1000.

Does having a fat bankroll and the ability to wager $1000 make your wager (vote) more likely to succeed? I think not. I would rather be on the side of the 500 voters.

I can't wrap my head around it, but I would think that this is why only 33% of favorites win, and the 'crowd' is 'wrong' 66% of the time.

If only I could have the statistics that show how many people wagered (voted) on a horse, and then follow that Wisdom of the Crowds!

Am I making any sense??


those are great points.

I think you are correct in that a pari-mutual system is at least a different variation of the the "Wisdom of Crowds" theory if not something totally different. Some of the numbers in this thread show a correlation between win-odds and winning percentage. So there is some phenomenon present.

The "high roller" aspect has a lot of variations in and of itself. Some scenarios to throw out there - A wealthy owner makes a sizeable wager every time his slow horse runs, or is it simply a big bet in a small pool made by a mediocre gambler, or is it some expert handicapper who figures there to be such a high probability of outcome for this particular horse that a large wager is a wise investment?

The more specific we get the more the generalities break down.

Is the Wisdom of Crowds theory applicable to the pari-mutal system?

[JPinMaryland]

Two interesting ideas:

. However, in pari-mutuel betting, the odds are determined not by the number of people who placed a bet, but by the total amount bet. ...I can't wrap my head around it, but I would think that this is why only 33% of favorites win, and the 'crowd' is 'wrong' 66% of the time.



Preceded by:

"The only odds ranges that showed a statistically significant variation from the expected number of winners (1-5, 1-1, 5-2, and 6-1 on the positive side; and 3-1 on the negative side) were too isolated or dispersed to indicate any overall pattern of deviation.."

I've seen these sort of spikes mentioned before. (Occuring at 5/2 and 6-1; nots ure of the other pts). Is it possible that these spikes have some relation to the problem that All Circuits Go mentioned, the whales the upset the natural proclivity of crowds to predict the outcome?

Perhaps there is more "whale money" that occurs at 3-1? One might speculate that this might be a pschogical pt. that whales are unwilling to go beyond in risking a large bet.

And perhaps at 5-2 and 6-1 there is less whale money so the horses are outperforming the odds. Maybe there is soem psychological barrier at these odds lines as well?

[1st time lasix]

excellent points and interesting thread. We all have our own set of experiences in this game. As a handicapper that uses the win pool or the same race exotics..... I try to isolate the contenders based on sound handicapping principals {class, distance, speed, trainer, pedigree, pace, post position, form cycle, etc} I then prefer to seek {and find} some true vunerability in the favorite or in the second choice before I play..... What I believe clearly differentiates the long term winner's "edge" from the rest of the crowd is the true structure of the wagers to take advantage of their opinion. I think this ticket creation process at the windows or at the keyboard is paramount and extremely under appreciated by too many players. Whatever your bankroll is.....allocate your real wagers to take advantage of your percieved overlays when the conditions are right in any race or in a particular specialty of yours. It may be a win bet. It may be an exacta without the favorite on top. It may be a trifecta or a superfecta with a key or with separation of the top choices. Maximize when you are confident.....$1 exacta boxes when you just want some some rooting "action"

[robert99]

ROI (corrected for take out) and win percentage is closely correlated over a LARGE number of races - things average out. Each large set of races will give different spikes where the ROI is slightly out of line. Data for this used to be published in UK for each racing season.

The crowd is very poor at ranking horses within a SINGLE race. They are poor because 99 out of 100 never attempt to do this - they just back their fancy, or tip, or the jockey, or the trainer, or what seems to be shortening in the market - they never learn nor progress so have no wisdom. That is why 99% consistently lose long term and their average loss on investment is 16%.

Just because someone writes yet another book on some hairbrained thoughts does not mean the words have any truth or merit - without substantiated facts to back the theories up. In the 30's, the German crowd followed the rantings of Hitler - hardly a crowd showing wisdom. Horseracing is far more complex than politics.

[JPinMaryland]

The Hitler analogy is always a pursuasive argument..

[betovernetcapper]

If the argument had brought out 3 days ago we could have said it was hare-brained (Easter).

[Overlay]

Yes, that is very good, but that point is pretty well understood, the horse racing market is pretty "efficient" (for the economists out there). My question is that in any "efficient" market there are bound to be segments (subsections) that are more or less efficient than others, so has anyone done this analysis for the horse racing market?

Quirin (in Winning at the Races), and Nunamaker (in Modern Impact Values), each identified data subsets involving factors or combinations of factors where horses won or lost significantly more often than their odds said they should have, to a degree too large to be accounted for by random variation. Quirin also discussed multiple regression techniques using balanced, properly weighted combinations of factors, and how fair odds could be derived from the numerical ratings the regression formulas produced.

[thelyingthief]

i know i take exception to the "wisdom of the crowd".

[podonne]

Just because someone writes yet another book on some hairbrained thoughts does not mean the words have any truth or merit - without substantiated facts to back the theories up. In the 30's, the German crowd followed the rantings of Hitler - hardly a crowd showing wisdom. Horseracing is far more complex than politics.

I'd read the book before making that kind of statement. It actually makes a great deal of sense (although arguably alot of common sense). He even explains the "hitler argument", or "when good crowds go bad". GO and pick it up, its actually an entertaining read.

I would argue with the previous posters who deny that the premise can apply to a parimutuel system (ALL CIRCUITS GO, Robert Fischer). I'd have to read it again to be sure, but the author uses the stock market several times to make his point, and the stock market exibits some of the same characteristics of a parimutuel system (multiple votes, single individuals can tilt the results)

The authors point was that the collective judgement of a (very diversified) group is often better than any particular subset or individual. If the parimutuel system was structured so that some group of experts was allowed more "votes" than everyone else, it might throw things off. As it is "rich guys" is the subset with more "votes", so as long as that set is diversified (plenty of stupid rich guys and smart rich guys) then the efficiency of the system is sustained.

Another subset might be owners, but again, given a good mix of smart/stupid and weahtly/poor, everything works out.

[robert99]

The UK Observer Newspaper is running a sports tipping competition over 6 months in conjunction with Betfair. The contestants are sports professionals, readers and a collection of animals from Whipsnade Zoo. The crowd of reader tipsters managed a final score of one big fat zero.

Needless to say the wise animals have taken the leading first three places in terms of win predictions.

First: A 2 year old hippopotamus - won 5 lost 5, £2626 ahead - selections by which cabbage in the row it picks to eat.

Second: A 5 year old kangaroo - won 1 lost 6, £3250 ahead - selections from which item in a row the roo chooses to bounce on.

Third: A 5 year old giraffe - won 4 lost 6, £2369 ahead - selections from how many times it falls asleep, standing up, over 2 days.

If only they could write books!