Final odds to index race competiveness

[TrifectaMike]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race. Threfore, would it not be an unreasonable assumption that the level of race competiveness can be inferred by the final tote odds?

For our purpose, let's view the pari-mutuel system as one "grand" bookmaker and odds-setter. Let me give this entity a human characteristic and call it, "The Man".

"The Man" is very sharp indeed and quite good at his job. He reasons that if he had a race with with an infinite number of horses with an equal amount of money bet on each horse, he would define this as a perfectly competitive race. If on the other hand all the money were bet on a single horse, he would define this as the "most" uncompetitive race. Obviously theses are two extremes and highy unlikely conditions, but this will give him some bounds to work from.

"The Man" fiddles somewhat with the odds and draws some conclusions.

A race with a heavy favorite should be classified as less competitive.
A race with more balanced odds should be classified as more competitive.
Maybe I can use this classification for future reference.
The more "The Man" thinks about it, the more he likes the concept. It is not based on actual race outcomes (not race class, not pace of race nor time of race). It is based on the final odds distributions.

After a bit of pencil scratching he settles on the following formula for the classification:

sum {1/[O(i) +1]^2}/N

O(i) Odds of the ith horse
N Number of entries

The larger the value, the more uncompetitive the race.

Mike (Dr Beav)

[DeltaLover]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race.
I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race.

I totally disagree.

'All known information' about a race is a composite of the public and insider domain.

The proportion of these two sources of information is impossible to be known before or after the race.

In other words a betting stable will have a totally different impact to the final odds than a non betting one.

Even more than this someone can make a point about the various derivatives of the 'public' info .

A good handicapper can detect a betting angle that generates an huge overlay while his bankroll does not allow him to bet enough to correct the value leaving some of it unexploited....

[Saratoga_Mike]

DL - excellent post.

[Some_One]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race. Threfore, would it not be an unreasonable assumption that the level of race competiveness can be inferred by the final tote odds?

For our purpose, let's view the pari-mutuel system as one "grand" bookmaker and odds-setter. Let me give this entity a human characteristic and call it, "The Man".

"The Man" is very sharp indeed and quite good at his job. He reasons that if he had a race with with an infinite number of horses with an equal amount of money bet on each horse, he would define this as a perfectly competitive race. If on the other hand all the money were bet on a single horse, he would define this as the "most" uncompetitive race. Obviously theses are two extremes and highy unlikely conditions, but this will give him some bounds to work from.

"The Man" fiddles somewhat with the odds and draws some conclusions.

A race with a heavy favorite should be classified as less competitive.
A race with more balanced odds should be classified as more competitive.
Maybe I can use this classification for future reference.
The more "The Man" thinks about it, the more he likes the concept. It is not based on actual race outcomes (not race class, not pace of race nor time of race). It is based on the final odds distributions.

After a bit of pencil scratching he settles on the following formula for the classification:

sum {1/[O(i) +1]^2}/N

O(i) Odds of the ith horse
N Number of entries

The larger the value, the more uncompetitive the race.

Mike (Dr Beav)

I think the problem with the assumption is that different people have different infomation, a morning line based on Beyer figs will be different then those using HDW, Bris or Thorograph. Then you have the pedigree, jockey, trainer info. It's what makes this game so great, so many ways to play.

[TrifectaMike]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race.

I totally disagree.

'All known information' about a race is a composite of the public and insider domain. This is true, but irrelevant to the metric.

The proportion of these two sources of information is impossible to be known before or after the race. This is not true, but also irrelevant to the metric.

In other words a betting stable will have a totally different impact to the final odds than a non betting one. Irrelevant.. The final odds are simply just that final odds by definition. I don't get the point.

Even more than this someone can make a point about the various derivatives of the 'public' info . This is in conflict with your first point.

A good handicapper can detect a betting angle that generates an huge overlay while his bankroll does not allow him to bet enough to correct the value leaving some of it unexploited.... This is true, but also irrelevant to the metric.

Help me out, becuase I seriously don't get what your are attempting to say in relation to this thread.

Mike (Dr Beav)

[TrifectaMike]

I think the problem with the assumption is that different people have different infomation, a morning line based on Beyer figs will be different then those using HDW, Bris or Thorograph. Then you have the pedigree, jockey, trainer info. It's what makes this game so great, so many ways to play.

FINAL Tote-Odds are FINAL tote-Odds!

Mike (Dr Beav)

[GameTheory]

Help me out, becuase I seriously don't get what your are attempting to say in relation to this thread.
I think something along the lines of "the final odds mean something different in each race depending on who bet that race and what info they were using and how much money they had, and since you can't tell what the mix of factors and money was that contributed to the final odds for each race, it is basically a garbage statistic, and so therefore will any metric be that is based on it, or at least the one you are proposing." That's my guess, anyway...

[ALL CIRCUITS GO]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race. Threfore, would it not be an unreasonable assumption that the level of race competiveness can be inferred by the final tote odds?

For our purpose, let's view the pari-mutuel system as one "grand" bookmaker and odds-setter. Let me give this entity a human characteristic and call it, "The Man".

"The Man" is very sharp indeed and quite good at his job. He reasons that if he had a race with with an infinite number of horses with an equal amount of money bet on each horse, he would define this as a perfectly competitive race. If on the other hand all the money were bet on a single horse, he would define this as the "most" uncompetitive race. Obviously theses are two extremes and highy unlikely conditions, but this will give him some bounds to work from.

"The Man" fiddles somewhat with the odds and draws some conclusions.


A race with a heavy favorite should be classified as less competitive.
A race with more balanced odds should be classified as more competitive.
Maybe I can use this classification for future reference.
The more "The Man" thinks about it, the more he likes the concept. It is not based on actual race outcomes (not race class, not pace of race nor time of race). It is based on the final odds distributions.

After a bit of pencil scratching he settles on the following formula for the classification:

sum {1/[O(i) +1]^2}/N

O(i) Odds of the ith horse
N Number of entries

The larger the value, the more uncompetitive the race.

Mike (Dr Beav)




----


My math is not great, so can you give an example of how the formula you settled on looks when there is a heavy favorite versus when there is not a heavy favorite? I can't figure out where that info is taken into account. thanks.
:cool:

[TrifectaMike]

I think something along the lines of "the final odds mean something different in each race depending on who bet that race and what info they were using and how much money they had, and since you can't tell what the mix of factors and money was that contributed to the final odds for each race, it is basically a garbage statistic, and so therefore will any metric be that is based on it, or at least the one you are proposing." That's my guess, anyway...

The final odds are a garbage statistic. REALLY?

Mike (Dr Beav)

[GameTheory]

The final odds are a garbage statistic. REALLY?

Mike (Dr Beav)
A) I didn't say I agreed with it, I was just trying to put out there what I think others are likely thinking.

B) I followed up that bit with something about "at least in regard to the metric you are proposing", i.e. your metric seems to make the assumption that a horse that (for example) is 2-1 in a five horse race is "equal" to any other 2-1 horse in some other five horse race, but of course that 2-1 is determined by the betting, and the betting in those two races was done by a totally different (more or less) group of people for totally different (more or less) reasons, and those people had totally different (more or less) amounts of money to play with, so that equality assumption needs some justification, doesn't it?

[Dave Schwartz]

Mike,

So, if I am understanding what you are saying, you are building a metric which demonstrates (theoretically) the competitiveness of a race as viewed by the current state of the tote board.

Yes?


Dave

[ALL CIRCUITS GO]

thats what I am trying to understand..

given equal fields of 5 horses, how do you adjust the metric to show that in one race the 2-1 favorite faced horses with odds of 4-1, 5-1, 6-1, and 7-1 while in the other race the 2-1 faced horses with odds of 2.5-1, 3-1, 3.5-1, and 3.75-1; and how is that shown?

maybe its there and I don't see/understand the maths.

also, how do you account for the fact that the two races come up 'competitive', yet in one race the favorite finished 1st and in the other the favorite finished 5th. Was the public wrong? Or was it the horse?

looks like a starting point, but where can it lead?

[GameTheory]

Of course the basic idea is nothing new -- I think a few of the programs out there have an entropy stat that is similar (not sure if they actually use Shannon's entropy formula or not, but basic idea is the same). Doesn't HTR have a volatility index based on the ML which is basically entropy? (Pre-race stat).

[TrifectaMike]

A) I didn't say I agreed with it, I was just trying to put out there what I think others are likely thinking.

B) I followed up that bit with something about "at least in regard to the metric you are proposing", i.e. your metric seems to make the assumption that a horse that (for example) is 2-1 in a five horse race is "equal" to any other 2-1 horse in some other five horse race, but of course that 2-1 is determined by the betting, and the betting in those two races was done by a totally different (more or less) group of people for totally different (more or less) reasons, and those people had totally different (more or less) amounts of money to play with, so that equality assumption needs some justification, doesn't it?

That is NOT what the metric is saying!

Mike (Dr Beav)

[TrifectaMike]

thats what I am trying to understand..

given equal fields of 5 horses, how do you adjust the metric to show that in one race the 2-1 favorite faced horses with odds of 4-1, 5-1, 6-1, and 7-1 while in the other race the 2-1 faced horses with odds of 2.5-1, 3-1, 3.5-1, and 3.75-1; and how is that shown?

maybe its there and I don't see/understand the maths.

also, how do you account for the fact that the two races come up 'competitive', yet in one race the favorite finished 1st and in the other the favorite finished 5th. Was the public wrong? Or was it the horse?

looks like a starting point, but where can it lead?

The metric is measuring what you said and I bolded it.

Mike (Dr Beav)

[TrifectaMike]

Mike,

So, if I am understanding what you are saying, you are building a metric which demonstrates (theoretically) the competitiveness of a race as viewed by the current state of the tote board.

Yes?


Dave

Dave,

It can used to determine the state of the tote board as well as a metric to index the races after the race has been run.

Mike (Dr Beav)

[TrifectaMike]

A) I didn't say I agreed with it, I was just trying to put out there what I think others are likely thinking.


Since you know what everyone is likely thinking, why not provide some good solid answers to their questions. Something useful would be nice.

Mike (Dr Beav)

[TrifectaMike]

Before you sling the arrows, understand the metric. You can agree, disagree, call it bullshit, but don't mischaractierize it.

sum {1/[O(i) +1]^2}/N

O(i) Odds of the ith horse, i =1,2,...,N
N Number of entries


Mike (Dr Beav)

[TexasDolly]

I was glad to see your work. Previous race odds have long been an important factor in my handicapping ,so I am anxious to incorporate the " competitive number " to see if it will help narrow the contender list down any.
Thank you,
TD

[GameTheory]

That is NOT what the metric is saying!No, it is not the final point since it is a race metric and not a horse metric, but nevertheless in your metric 2-1 is 2-1 is 2-1 without regard to why or how it got to be that way in a particular race. I'm not arguing, just saying.

[GameTheory]

Since you know what everyone is likely thinking, why not provide some good solid answers to their questions. Something useful would be nice.I never said I knew what "everyone" is likely thinking, but some, sure. Just trying to help you get on their wavelength since you have trouble with that, and you did ask. Why is it always combat with you?

[TexasDolly]

Mike,
Do I sum the odds of all the horses and then do the remaining steps or do I solve the equation for each horse and then sum the results ?
Thank you ,
TD

[Saratoga_Mike]

Help me out, becuase I seriously don't get what your are attempting to say in relation to this thread.

Mike (Dr Beav)

Simple, you're essentially making an efficient market theory claim as it pertains to pari-mutuel wagering, claiming all known info is reflected in the mkt. DL showed - very clearly - that claim isn't true b/c information from barns that don't bet is not reflected in the betting pools.

Anyway, this looks like another one of your foolish threads that accomplishes nothing.

[ALL CIRCUITS GO]

ok, so I appealed to the o.p. for help in understanding, and I still request help. could the o.p. show a workout for a past race so that I can understand how the numbers work into the formula? I am not a math major, and symbols like ^ mean nothing to me. help please. thanks.

[Dave Schwartz]

It can used to determine the state of the tote board as well as a metric to index the races after the race has been run.

I am going to add your formula into H8 and we'll see how it looks. I think it has very interesting possibilities.

Thanks for sharing.


Dave

[TexasDolly]

Dave, maybe you can help me here. Do you sum all the odds for the horses and then compute the result or do you solve the statement for each horse and then sum the results ?
Thank you,
TD

[ALL CIRCUITS GO]

so, if you take the horses last race, and compute the races efficiency, and then compare the actual finish of the horse that race to the expected finish based upon that races efficiency, you can determine whether the horse was judged properly or improperly in that race, and then apply that understanding to todays race by correlating that statistical abstract to the odds of the horse in todays race, and in doing so for all of the horses, look for the horses you have determined were properly judged last race and finished accordingly and are also judged to be able to finish well today.

but some how do it with math?

is that right?

so this doctor belonged to a group of doctors that would meet once a month over drinks to talk about the tough cases they had in the last month and the diagnosis they had made. this one doctor walked into the meeting and described a case. when the other doctors had reached a consensus on the diagnosis, they asked him what the outcome actually had been. well, the doctor stood up, and rushing out of the room, said, I'll let you know tomorrow when I apply what you have told me tonight. so the doctor had used up his own database of knowledge and was looking to mine the depths of his associates, which he did successfully.

[dkithore]

Before you sling the arrows, understand the metric. You can agree, disagree, call it bullshit, but don't mischaractierize it.

sum {1/[O(i) +1]^2}/N

O(i) Odds of the ith horse, i =1,2,...,N
N Number of entries


Mike (Dr Beav)

Doesn't Brisnet's Race Rating trying to do the same?

DK

[Dave Schwartz]

Dave, maybe you can help me here. Do you sum all the odds for the horses and then compute the result or do you solve the statement for each horse and then sum the results ?

No, sum the SQUARES of the odds then divide 1 by the resulting sum.

Actually, Mike has ideas turning in my head. For example...

I'd like to use the top 3 horses for public choice only rather than the whole field. This becomes a metric of the "energy" (if you will) at the top of the food chain.

In other words, a race with 2/1, 2/1, 3/1 is far different that 6/5, 8/5, 5/2 or 3/5, 4/1, 9/2.



I find Mike's ideas very stimulating. So stimulating that I hope to get the chance to discuss a problem I am working on with him.


Dave

[DeltaLover]

Help me out, becuase I seriously don't get what your are attempting to say in relation to this thread.

Mike (Dr Beav)

The creation of the odds follows a non deterministic model that is influenced by hundreds or even thousands of factors, opinions, information and misinformation that constitute a very volatile process.

A single random bettor can easily change the 'board' not to mention organized attempts to deceive the public.

A Dutrow horse will never be bet the same way an Angel Pena's for example. The Dutrow gang, consisting of really big bettors, will bet their horse down to 7-5 while the same Senior's horse would had been 5-2 having very little insider action... As simple as that...

The truth is that exactly because the betting markers do not represent consistently and efficiently the available information this game can (at least theoretically) be beaten.

It is the difficulty to decipher handicapping factors that tends to mislead the public or even the insiders creating a subjective views of odds that would have been totally different had they had been created by a totally different universe of bettors...

For example the Tampa Bay crowed consisting from a high percentage of retirees and casual gamblers will bet totally different from an elite group of Belmont professional players even if they are handicapping exactly the same race.....

Your definition of competitiveness:

A race with a heavy favorite should be classified as less competitive.
A race with more balanced odds should be classified as more competitive.

does not really present any handicapping or betting value.

As a bettor I do not care to introduce a new index called competitiveness. Only the detection of an overlay and nothing else can lead me to a winning strategy. The odds of the horse as an individual factor are totally meaningless as is 'strike rate'. A horse that should have been 24-1 and is bet in the pools as a 35-1 is a great bet by any means even if only has a 5% strike rate....

If the heavy favorite you are referring is a 1-5 chalk that should have been 4-5 then we have a playable race that although non competitive by your definition, indeed creates some betting opportunities to higher odds horses. The reason this happens is exactly the fact that the information is not estimated properly by the public which could be due to insider bets (or absence of them), a single big bettor or simply to the fact that the correct interpretation of the data is deceiving for the betting public.

So to conclude:

1)You approach is conceptually wrong since the odds as formed by the public do not always project all information with the same degree of effectiveness .

2)Even if the previous was true there is no value to the introduction of a new race attribute called competitiveness or anything else since what we really care in this game as bettors and handicappers is to detect the crowd mistakes and this can only be done by the discovery of the overlays and nothing else

[Robert Goren]

It seems to me that you reaching for something that doesn't tell you much.

[TrifectaMike]

Dave, maybe you can help me here. Do you sum all the odds for the horses and then compute the result or do you solve the statement for each horse and then sum the results ?
Thank you,
TD
Final Odds from Race Result Chart (6 Horse Race(s))
N H1 H2 H3 H4 H5 H6
Example 1 6 8.60 0.6 3.1 20.4 6.3 14.7
Example 2 6 4.1 1.6 2.2 10.2 3.5 91.9

Compute 1/(Odds +1) times 1/(Odds +1) for each horse
H1 H2 H3 H4 H5 H6
Example 1 0.0109 0.3906 0.0595 0.0022 0.0188 0.0041
Example 2 0.0384 0.1479 0.0977 0.0080 0.0494 0.0001

Sum the terms and divide by number of horses
Sum Terms Divide The Sum by Number of Horses in the Race
Example 1 0.4860 0.0810 Race Index
Example 2 0.3415 0.0569 Race Index


Mike (Dr Beav)

[thaskalos]

It has been widely reported - and varified using databases containing many thousands of race results - that the lower the horse's closing odds are...the more likely that horse is to win the race.

That is, the race favorite wins more races than the second-favorite...who, in turn, wins more races than the third-favorite...who also wins more races than the fourth-favorite, etc...

The closing odds - lowest to highest - have predicted the actual "long run" race results EXACTLY.

However...there has been NO STUDY - as far as I know - to indicate how accurate these odds are in predicting how these horses finish WITHIN the race.

Is the second-favorite in a race more likely to finish SECOND...and is the third-favorite more likely to finish THIRD, etc?

I seriously doubt it...

And if the closing odds fail to predict how the horses perform WITHIN the race...how then can they be expected to predict the race's "competitiveness level"?

[TrifectaMike]

I simply have to shake my head in disbelief at some of the responses.

Mike (Dr Beav)

[thaskalos]

I simply have to shake my head in disbelief at some of the responses.

Mike (Dr Beav)
Is this your definition of a constructive comment?

Or are you auditioning to be the "Don Rickles" of this board?

[TrifectaMike]

Is this your definition of a constructive comment?

Or are you auditioning to be the "Don Rickles" of this board?

Defintion
It is a metric measuring competitiveness as determined by the bettors. A concentration or dispersion of the odds in a horse race.

Mike (Dr Beav)

[thaskalos]

Defintion
It is a metric measuring competitiveness as determined by the bettors. A concentration or dispersion of the odds in a horse race.

Mike (Dr Beav)
Mike...

I think that you have a mania for the unimportant...

[Robert Fischer]

Defintion
It is a metric measuring competitiveness as determined by the bettors. A concentration or dispersion of the odds in a horse race.

Mike (Dr Beav)

This is a great idea, and you present it in a very clean formula.

I would also like to see an example for dummies.

Some have claimed that because the formula doesn't take into account data that perfectly measures the horses true ability and therefore measure a true competitiveness, that it is somehow worthless.
In fact it is valuable. ALL racing figures have flaws. With this figure there will be times (such as races with a false favorite, that was truely false and didn't just happen to run poorly,) that the numbers will be wrong - in the case of a false favorite, it may measure the race as less competitive than it actually was.
The majority of the time it will be a fairly accurate measure and will provide at-a-glance insight into what the final odds suggest the competitiveness was.
I like it.

[Robert Fischer]

Actually, Mike has ideas turning in my head. For example...

I'd like to use the top 3 horses for public choice only rather than the whole field. This becomes a metric of the "energy" (if you will) at the top of the food chain.

In other words, a race with 2/1, 2/1, 3/1 is far different that 6/5, 8/5, 5/2 or 3/5, 4/1, 9/2.



I find Mike's ideas very stimulating. So stimulating that I hope to get the chance to discuss a problem I am working on with him.


Dave

yes :ThmbUp:

there are a number of ways to tailor a metric like Mike's to "concentrate" on a specific area of the race (top 3 choice...).

I like his original base metric at least to start with and get comfortable with. Learn the ins/outs.

Then you can try it with different groups like the top3 choices. Your suggestion goes beyond averaging to best ask what you seem to be asking, so the metric will need some tweaking. There is also the idea of just using the top 3 choices or just using all less than a certain cut-off (like 5-1) and using the original metric.

I experimented some with Final Odds when looking at the pace of a race, and what I basically did was average the pace of the top3 choices. The idea was to differentiate between races where two longshots dueled up front(with the contenders measuring a sensible pace), and races where the favorites dueled up front - when both races would look the same numerically but the latter could be a more grueling pace, and the former could be just a moderate pace.
I didn't know how to automate the system and eventually I learned to do the equivalent by looking at the charts.

[HUSKER55]

ok boys and girls. I got lost....again.

you can look at the tote and see if the public has any idea who will win the race or if you have a real race where any one can win.

How do you apply this index to the tote so you have an edge?

thanks guys

[ALL CIRCUITS GO]

This is a great idea, and you present it in a very clean formula.

I would also like to see an example for dummies.



d'uh yeah!

[TexasDolly]

Mike,
Thanks for your example ,it was clear and easy to follow.
TD

Dave,
Thanks to you for your comment as well.
TD

[thaskalos]

This is a great idea, and you present it in a very clean formula.

I would also like to see an example for dummies.

Some have claimed that because the formula doesn't take into account data that perfectly measures the horses true ability and therefore measure a true competitiveness, that it is somehow worthless.
In fact it is valuable. ALL racing figures have flaws. With this figure there will be times (such as races with a false favorite, that was truely false and didn't just happen to run poorly,) that the numbers will be wrong - in the case of a false favorite, it may measure the race as less competitive than it actually was.
The majority of the time it will be a fairly accurate measure and will provide at-a-glance insight into what the final odds suggest the competitiveness was.
I like it.
Robert, explain something to me in layman's terms please...

How can "measuring competitiveness as determined by the bettors" really be used for future reference...as is implied by TrifectaMike in his original post here?

Do the closing odds REALLY reflect how "competitive" the race turns out to be?

Has it been your experience that the most hotly contested races on the tote board also materialize as such during the running of the race?

[Dave Schwartz]

How can "measuring competitiveness as determined by the bettors" really be used for future reference...as is implied by TrifectaMike in his original post here?

Do the closing odds REALLY reflect how "competitive" the race turns out to be?

Has it been your experience that the most hotly contested races on the tote board also materialize as such during the running of the race?


(Robert, PMFJI.)

Thaskalos,

Let's go off-topic for just a moment.

Randy Giles came up with a great idea when he said that what mattered were not so much the number of Quirin ES points in a race, but rather the sum of the 5+ early speed points. It gives a pretty clear picture of the race...


Race A: 8,7,6,5 = 26
Race A: 8,8,8 =24

Race B: 8,7 =15
Race B: 5,5,5 =15

Race C: ... =0


Each of these races gives a handicapper a picture of a race.

What wins an "A" race is probably different than what it takes to win a "B" or "C" race.

We can learn what wins an "A" race by looking at our database, isolating on A-type races and studying the winners (and the losers).


Okay, back to the subject at hand. To keep this simple -

Race A
Prg C Odds
7 C 0.40 0.5102
9 C 5.40 0.0244
8 C 5.80 0.2163
4 f 12.90 0.0052
* 2 17.50 0.0029
1 C 17.70 0.0028
* 3 48.20 0.0004
6 79.40 0.0002
5 111.90 0.0000
-------
0.7572

/ 9 =0.0841 per horse

Take SqRt of that = 0.29

approximately 5/2 (2/7 = 0.2857)


Race B
Prg C Odds
3 C 2.10 0.1041
4 C 2.10 0.1041
7 C 2.90 0.0657
5 C 5.20 0.0260
2 12.80 0.0525
6 14.70 0.0041
1 17.60 0.0029
------
0.3594

/ 7 = .0513 per horse

Take SqRt of that = 0.2266

approximately 7/2 (2/9 = .2222)



So, the question is, "If we look in the database at races which are like Race A, do we find a different way to bet than if we study races that are more like Race B?"


Personally, I find this approach absolutely brilliant. (Frankly, I am surprised I never thought of it myself. :rolleyes: )


Regards,
Dave Schwartz

[classhandicapper]

There is one problem I see with this kind of thing.

Assume a deep competitive field of stakes horses and Secretariat is running the race. The field is no less strong because Secretariat went off at 3-5. His presence made the field even stronger.

The odds of the favorite don't necessarily tell you if the rest of the field was weak or if the favorite was simply a killer.

I occasionally use past odds to evaluate a horse that ran a dud race in its last start.

If a horse ran poorly out of town and I am not familiar with the quality of the horses in the race, if the horse went off at long odd I'll be more forgiving of the race on the assumption that the field was very strong than I would if the horse went off at 2-1. This is the same kind of thing.

[Dave Schwartz]

Assume a deep competitive field of stakes horses and Secretariat is running the race. The field is no less strong because Secretariat went off at 3-5. His presence made the field even stronger.

The odds of the favorite don't necessarily tell you if the rest of the field was weak or if the favorite was simply a killer.


Class,

He isn't trying to measure the strength of field. H is trying to measure the public's perception of the competitiveness of the field.

I beat Saratoga almost every year. The primary reason is because the "typical" odds on the favorite are much higher than at other tracks. Not in every race, but there are so many such race as to be able to concentrate on the races where there is a good chance of making money.


Consider the Secretariat race - If Mike worked that race, I'd bet you'd get a very low number for the race, indicating that the public feels the race is not very competitive. (They turned out to be right, of course.)

Which kind of race would you TYPICALLY like to bet into: One where the favorite is 2/5 or a race where the favorite was 7/2? I think most people would agree that they would prefer the 7/2 race.

(Of course, if one is really good at picking very low odds favorites to bet against, one might actually enjoy that 2/5 race, but usually not.)

What Mike has offered is a way to measure more than just the favorite to determine the public's perception of the competitiveness.


Again, I say, "brilliant."

[thaskalos]

Okay, back to the subject at hand. To keep this simple -

Race A
Prg C Odds
7 C 0.40 0.5102
9 C 5.40 0.0244
8 C 5.80 0.2163
4 f 12.90 0.0052
* 2 17.50 0.0029
1 C 17.70 0.0028
* 3 48.20 0.0004
6 79.40 0.0002
5 111.90 0.0000
-------
0.7572

/ 9 =0.0841 per horse

Take SqRt of that = 0.29

approximately 5/2 (2/7 = 0.2857)


Race B
Prg C Odds
3 C 2.10 0.1041
4 C 2.10 0.1041
7 C 2.90 0.0657
5 C 5.20 0.0260
2 12.80 0.0525
6 14.70 0.0041
1 17.60 0.0029
------
0.3594

/ 7 = .0513 per horse

Take SqRt of that = 0.2266

approximately 7/2 (2/9 = .2222)



So, the question is, "If we look in the database at races which are like Race A, do we find a different way to bet than if we study races that are more like Race B?"


Personally, I find this approach absolutely brilliant. (Frankly, I am surprised I never thought of it myself. :rolleyes: )


Regards,
Dave Schwartz
Yes...but aren't we then just lumping together all those "Race A" (or Race B) types - as defined by the closing odds - as if they were similar in style?

What happened to the argument that each and every race is unique...because of the different "dynamics" that define it?

[Elliott Sidewater]

Until Trifecta Mike shows us how to use his metric, I have nothing more to say about it. It's only a proposed metric at this point in time. Maybe it's good, maybe not. Conceptually, I think that the odds are not as good a measuring stick for competitiveness as other measures. Those that I think are more important are:

1. the margins between the horses at the finish, along with the (adjusted) final time
2. (pre-race) class ratings for the horses that finished in the top half of the field

The odds measure what the bettors think will happen, the result charts show what did happen. One problem I see with TM's concept is that if you have a 3-5 shot that breaks down in the race, and then a 20-1 shot wins in slow time in a blanket finish, his metric will be identical for that race as if the 3-5 had won for fun.

[thaskalos]

Until Trifecta Mike shows us how to use his metric, I have nothing more to say about it. It's only a proposed metric at this point in time. Maybe it's good, maybe not. Conceptually, I think that the odds are not as good a measuring stick for competitiveness as other measures. Those that I think are more important are:

1. the margins between the horses at the finish, along with the (adjusted) final time
2. (pre-race) class ratings for the horses that finished in the top half of the field

The odds measure what the bettors think will happen, the result charts show what did happen. One problem I see with TM's concept is that if you have a 3-5 shot that breaks down in the race, and then a 20-1 shot wins in slow time in a blanket finish, his metric will be identical for that race as if the 3-5 had won for fun.
Right!

If we are concerned with creating a way to measure competitiveness for future reference...why not rely on something more "concrete" than the closing odds?

There is pertinent information available to us AFTER the race is run, which reflects the race's competitiveness level?

Why do we have to rely on information available BEFORE the race?

[Dave Schwartz]

Yes...but aren't we then just lumping together all those "Race A" (or Race B) types - as defined by the closing odds - as if they were similar in style?

What happened to the argument that each and every race is unique...because of the different "dynamics" that define it?

Each and every race is unique.

However, to develop any type of strategy, we most SOMEHOW lump races together into "like categories." You do this all the time, don't you?

Don't you treat turf races different than dirt races?
Sprints different than routes?
2yr olds different than 5 year olds?


Well, you guys can suit yourself. I am going to LOOK CLOSER at this.

[thaskalos]

Each and every race is unique.

However, to develop any type of strategy, we most SOMEHOW lump races together into "like categories." You do this all the time, don't you?

Don't you treat turf races different than dirt races?
Sprints different than routes?
2yr olds different than 5 year olds?


Well, you guys can suit yourself. I am going to LOOK CLOSER at this.
We will just have to agree to disagree on this one Dave...

Yes...I do "group" races into "like categories"...but they have to share common characteristics -- like the "dirt/turf" and "sprints/routes" that you mentioned.

It hadn't yet occurred to me to lump races together just because of their pre-race odds layout.

That's what's so great about this game though...we learn something new every day...

[GameTheory]

So, the question is, "If we look in the database at races which are like Race A, do we find a different way to bet than if we study races that are more like Race B?"


Personally, I find this approach absolutely brilliant. (Frankly, I am surprised I never thought of it myself. :rolleyes: )
Epstein in "The Theory of Gambling and Statistical Logic" proposes a betting system based solely on the entropy of the post-time odds, suggesting that by looking at the entropy just before the race goes off and comparing it to a database of races with the same (or close) entropy score, you could simply bet on those horses with a positive expectation on that basis. So for example, you might find that in races with entropy X (or maybe >= X), any 6-1 horse might be worth a bet. (The idea goes basically untested though, he just mentions that a small study shows potential.)

The relevant page can actually be seen on google books:

http://books.google.com/books?id=1cho4FiE_NMC&lpg=PP1&dq=gambling%2Bstatistical%2Blogic%2Bepstein%2Bentr opy&pg=PA303#v=onepage&q=entropy&f=false

If the link works correctly, you should see page 303 on top, which is the page to look at. The formula for entropy is not shown on that page, but is similar to Mike's formula. Instead of summing the squares of the odds, you'd first convert the odds to probabilities and normalize (so they sum to one) and sum the result of p*log(p) for each horse. (The log should be base2 as the entropy formula is about measuring information in bits -- i.e. binary.) Races of different field sizes are not really comparable under that formula, though, as you do not divide by the number of entrants (and it wouldn't work if you did). Mike's formula might have potential as a replacement using the same idea. (Or likely neither would work since you need to figure it and bet before the race and it is likely to change drastically by the time all the money comes in.)

[BobD]

I think this all has promise but I think it would be helpful to point out a couple of things.

First of all, your race A computation has a math error: program number 8 should be .0216 not .2163, but more importantly you have moved away from the original posted formula. The example of the posted race index formula computed the average of the sum of the squares of the calculation derived from the odds but the formula posted in your example computes and uses the square root of that number.

Is that intended as a suggested modification/improvement?

[pondman]

I believe that we can all agree that the final odds "contain" ( is a refection of) all known information about a particular race.

Final odds include known facts, assumptions, erroneous information, and unknowns. I believe most of the known is not of quality.

Threfore, would it not be an unreasonable assumption that the level of race competiveness can be inferred by the final tote odds?


Do you mean closer to being a die?

I'm not sure why you would want a race to be competitive?

[Dave Schwartz]

Is that intended as a suggested modification/improvement?

Errors being what they are - sorry. You are correct.

However, I did not post the Square Root. I posted the square.

As an example:

5.40:1 1/6.4=.15625^2 = .0244

Square root would have been .3953

[gm10]

Mike, it's a nice theoretical concept, but I must admit I'm struggling to see myself using it in practice. Do you use it yourself?

On the other hand, I can see a lot of practical value in applying the idea to my own odds-line rather than the public odds, since I already know those personal odds when I'm handicapping the race. The entropy number could - conceivably - show a correlation with the reliability and profitability of those personal odds.

[classhandicapper]

Class,

He isn't trying to measure the strength of field. H is trying to measure the public's perception of the competitiveness of the field.

What Mike has offered is a way to measure more than just the favorite to determine the public's perception of the competitiveness.



I'm not sure why you would want to know whether the public thought a field was competitive or not unless you are hoping to translate that into a measurement of quality.

[Dave Schwartz]

The point is that fields are different.

It is the building of subsets of data that allows one to find a way to win at the races in the first place.

[TrifectaMike]

Mike, it's a nice theoretical concept, but I must admit I'm struggling to see myself using it in practice. Do you use it yourself?

On the other hand, I can see a lot of practical value in applying the idea to my own odds-line rather than the public odds, since I already know those personal odds when I'm handicapping the race. The entropy number could - conceivably - show a correlation with the reliability and profitability of those personal odds.

Yes, I do use it. Primararily in my Bayesian Belief Networks. I have also used as a predictor for race times.

I like your idea of using the metric on your own odds-line. An interesting area to explore.

Mike (Dr Beav)

[Some_One]

Class,

He isn't trying to measure the strength of field. H is trying to measure the public's perception of the competitiveness of the field.

I beat Saratoga almost every year. The primary reason is because the "typical" odds on the favorite are much higher than at other tracks. Not in every race, but there are so many such race as to be able to concentrate on the races where there is a good chance of making money.


Consider the Secretariat race - If Mike worked that race, I'd bet you'd get a very low number for the race, indicating that the public feels the race is not very competitive. (They turned out to be right, of course.)

Which kind of race would you TYPICALLY like to bet into: One where the favorite is 2/5 or a race where the favorite was 7/2? I think most people would agree that they would prefer the 7/2 race.

(Of course, if one is really good at picking very low odds favorites to bet against, one might actually enjoy that 2/5 race, but usually not.)

What Mike has offered is a way to measure more than just the favorite to determine the public's perception of the competitiveness.


Again, I say, "brilliant."

I don't like playing races with high favs, usually those races the odds are quite flat (think the Preakness was like this) where the public hasn't made a choice and thus really can't be that wrong. The way we profit is when the public is really wrong, so I want the evens or 3/2 fav which I can play against.

[ALL CIRCUITS GO]

how truly unfortunate that he did not care to enlighten me.

:confused:

[Show Me the Wire]

I think the idea is to measure for example whether a 5/2 horse in a low level claiming race is the same risk as a 5/2 horse in an allowance race.

If I am wrong please correct my thinking.

[Dave Schwartz]

I think the idea is to measure for example whether a 5/2 horse in a low level claiming race is the same risk as a 5/2 horse in an allowance race.

I can tell you that this is not what I am thinking.


It is more like:

"...whether a 5/2 horse in a race where he is the favorite is the same risk as if he is 3rd choice in a race."

And possibly more important to many of us would be how the 6/1 horse does.

As someone alluded to earlier, there are races where the public isn't THAT wrong.

Imagine that there are races where the 6/1 horse might be (logically) around 4th choice, while there are races where he COULD be 2nd choice. The question is, "Does he perform differently?"



how truly unfortunate that he did not care to enlighten me.

ALL CIRCUITS GO,

Actually, I was noticing a change in the original poster's attitude. If memory serves, he used to get pretty angry when someone disagreed with him or even didn't understand what he was saying.

It seems to me that he has done a much better job of staying on topic. If this continues, there is no telling what I might be able to learn from him.

Just my opinion.


Regards,
Dave Schwartz

[ALL CIRCUITS GO]

Actually, I was noticing a change in the original poster's attitude. If memory serves, he used to get pretty angry when someone disagreed with him or even didn't understand what he was saying.

It seems to me that he has done a much better job of staying on topic. If this continues, there is no telling what I might be able to learn from him.

Just my opinion.


Regards,
Dave Schwartz

I agree with you. He does seem to have changed his attitude. I remain the patient student.

ACG

[Dark Target]

Do you want to know how competitive the race could be, or how competitive it actually was?

As mentioned earlier, surley the final margins are a much greater indictaion of how competitive the race actually was.... If this is a post race stastistic, surely you would be better served using post race data :confused:

[Dave Schwartz]

Do you want to know how competitive the race could be, or how competitive it actually was?

Neither. I want to know how competitive the public thought it would be.

[Greyfox]

The original metric looks like a measure of bettor confusion (or indecisiveness) vs decisiveness (perceived). Obviously it does not measure race competitiveness.

[TrifectaMike]

Class

He isn't trying to measure the strength of field. He is trying to measure the public's perception of the competitiveness of the field.

What Mike has offered is a way to measure more than just the favorite to determine the public's perception of the competitiveness.

Dave,

You have nailed it. I am truly amazed at how quickly you have internalized the metric.

I'm certain that somewhere in the back of your mind, you had thought of this in some fashion.

Mike (Dr Beav)

[HUSKER55]

So you are saying that a 6:1 horse in a race with an index of .150 has a better chance than a 6:1 horse in a race with an index of 2.50 ? In other words the horse in the second race is closer to what his odds should be?

Am I on the right track?

[LottaKash]

Neither. I want to know how competitive the public thought it would be.

Dave and all, I have been monitoring this thread with great interest and enthusiasm, but, I have some reservations about how this would help me my in own particular way of going at the races....Perhaps it is a bit too obscure or abstract to my own way of thinking, so I was a bit perplexed about that....

What I mean to say is, odds are odds, so how would I be able to use it (knowing how competitive the public thought a race to be) to my advantage....I have been comparing my own thoughts to the general crowd's for a good long while now, so what would be any different in that regard ?

Maybe in certain races ?....Maybe a differing track to track mentality ?.....Smart Money races ?....

I am not sure if I am asking, enough of, or the right or pertinent questions regarding this subject....

Any elaboration would be greatly welcome....

best,

[Dave Schwartz]

Kash,

Maybe the best answer is to address Husker's question:

So you are saying that a 6:1 horse in a race with an index of .150 has a better chance than a 6:1 horse in a race with an index of 2.50 ? In other words the horse in the second race is closer to what his odds should be?

Am I on the right track?

The answer is, "I don't know, but with a study I might be able to find out."

From my current point of view I will simply say that I expect their chances are different.

A good way to test that there MAY be something there to look at is to ask the question:

"Do 3/1 horses perform differently based upon their public choice number?"

This would address whether or not there were horses of lower odds than the 3/1 shot we are studying.


last 10000.
201-PubCh+1-rPubCh
WIN BETS
Odds PubCh Starts Pays Pct $Net IV PIV
3/1 1 192 45 23.4 $1.94 2.31 1.19
3/1 2 2,007 416 20.7 $1.73 1.63 1.06
3/1 3 1,111 215 19.4 $1.62 1.34 1.00
3/1 4 61 13 21.3 $1.82 1.34 1.11

Total 3,371 689 20.4 $1.71 1.55 1.02

Based upon the last 10,000 races from my database, I would say that there IS a difference, although the sample size does not allow for a very high correlation - still 80+%.


Regards,
Dave Schwartz

[LottaKash]

Kash,

Maybe the best answer is to address Husker's question:

The answer is, "I don't know, but with a study I might be able to find out."

Based upon the last 10,000 races from my database, I would say that there IS a difference, although the sample size does not allow for a very high correlation - still 80+%.


Regards,
Dave Schwartz

Thanks for that Dave, :ThmbUp: :ThmbUp: :ThmbUp:

I guess for me tho, not having any significant database exepting what is in this ole head of mine, I can't see where this line of questioning would have much impact on my bottom line at all...I just don't have the numbers to crunch...

In these past few years, my action has slowed to a snail's crawl, as I don't bet nearly as often as the way I used to....I have managed to get "my way of going", down to just a handful of carefully screened angles and such, so my plays are often few and far in between, or they will come in bunches and then quiet down for several days at a time....Given that, it would'nt be of much use in me knowing about something that, at first consideration, to be of only limited value to begin with....Actually, it may even tighten my play up even more, by inhibiting my play with those "new" odds considerations and/or limitations....

I just need to stay focused, and, wait and wait and wait....They are money tho, when they are there...I suppose I am like Mark Cramer, in that, I have always been searching for the "Automatic"....Gettin close these days...

Thx again Dave, now it's back to "Monty Hall"...I am still struggling with that a bit...Trying to integrate that way of thinking without losing what I already have thus far, in my search for "a new, and possibly better way"...

best,....

[andicap]

I get it!!! (I think)
But I'm still a bit fuzzy on one aspect of this -- the application. (see the end of this post for that question, Dave/Mike)

Maybe I'm wrong but here's my interpretation.

First to those who espousing the faults -- racing is not an exact science. Nothing wrong with healthy skepticism but you might be missing the forest for the trees. You could pick apart ANY handicapping theory. It doesn't have to be perfect. It just has to be good enough. Some of those tearing apart Mike's formula might have dismissed the Ragozin figs if he had explained them back in 1980. "What, you don't account for pace or trainer? You're crazy!!"

So I'm not sure if some of these responses fall into the category of healthy skepticism or corrosive withering cynicism.

Second, I believe Dave and Mike are trying to categorize races in order to put them in orderly sub-groups that can analyzed on their own. Yes, no two races are alike, but you've got to group them somehow or instead of having a database of say, 15 to 20 different types of races, you'd have 1 million databases of one race each.

Particular groupings of race types yield disparate results which Mike is trying to analyze. In Dave's example he referred to Randy Giles' idea of pace pictures: That certain types of pace pictures yield their own peculiar set of winners. Not all the time. But enough to exploit them and make a profit with smart contender selection and wagering.

This idea is Giles' pace pictures wrapped in different clothing. Instead of looking at Quirin Speed Points we are analyzing the public's perception of a race. Are there patterns in certain types of betting scenarios that repeat themselves enough to be predictable?

In effect, should races that yield results of say, 5-2 on Mike's formula (or perhaps Dave's tweaks) be evaluated and bet differently than races that average 7-2?

Here is my question: Would you apply this by looking at the odds two minutes before a race and assigning this race to a particular grouping of odds? If so, I'm assuming that the resulting odds changes until post time would be modest enough to allow for valid interpretations two minutes out.

Or is this just an exercise to appraise races that have been run? If so, I'm not sure I see the practical application. There's no use in putting races into sub-groups if you can't assign that data to help evaluate today's race.







.

[HUSKER55]

ALRIGHT DAVE AND MIKE, let me try this again, as I seem to be having a problem getting my head around this.

suppose we take the horses last two outs and find that the rate is 1.2 and then the last out the rate is 4.7.

would we be correct to assume that the horses in the last race got beat by a horse that won by 10 lengths ( or superior horse) and the rest came in 2nd, 3rd and 4th at 12 lengths off lead for all three.

Maybe the second race back is fairer odds of what the horse should be at because there was one superior horse in that last race.

[Dave Schwartz]

Husker, you are lost but Andy is getting pretty close to being found.

The idea is to answer this question about today's race:

"How do I bet?"


Remember that racing is a 2-phased problem: Handicapping and wagering.

We are addressing what to do after the handicapping phase here.

Imagine if I told you that in this kind of race horses over (say) 12/1 rarely get into exactas. Would that (perhaps) change your betting style if one of the horses you liked was 15/1? Of course it would.


Taking a slightly different explanatory path:

If I said that a race had a certain "pace shape" you would know what I meant. You also have an idea of how to modify your wagering strategy based upon that shape. (Whether you are good at that or not is another story, but you probably have an idea of how to do it.)


Well, imagine that we might eventually come to understand what a wagering shape meant. In a race that is defined as a "2/1" race we may wager differently than a "4/1" race.

I have no idea at this point what the differences might be - as I said: study is the key.


Dave

PS: Seems like I keep taking over, Mike. Sorry, you still get full credit for the great idea.

[TrifectaMike]

Husker, you are lost but Andy is getting pretty close to being found.

The idea is to answer this question about today's race:

"How do I bet?"


Remember that racing is a 2-phased problem: Handicapping and wagering.

We are addressing what to do after the handicapping phase here.

Imagine if I told you that in this kind of race horses over (say) 12/1 rarely get into exactas. Would that (perhaps) change your betting style if one of the horses you liked was 15/1? Of course it would.


Taking a slightly different explanatory path:

If I said that a race had a certain "pace shape" you would know what I meant. You also have an idea of how to modify your wagering strategy based upon that shape. (Whether you are good at that or not is another story, but you probably have an idea of how to do it.)


Well, imagine that we might eventually come to understand what a wagering shape meant. In a race that is defined as a "2/1" race we may wager differently than a "4/1" race.

I have no idea at this point what the differences might be - as I said: study is the key.


Dave

PS: Seems like I keep taking over, Mike. Sorry, you still get full credit for the great idea.

Don't be sorry. I'm enjoying your journey. I have to admit I find your ability to rationally assemble what appear to be disjointed facts uncanny.

Andicap, you do get it.

Keep in mind that the metric is an index. Post race indexing is neccesary to set-up baselines. Otherwise, you are there.

Mike (Dr Beav)

[HUSKER55]

IF a player doesn't have a data base is this topic going to help?

[Dave Schwartz]

It will if we share our findings. <G>

[HUSKER55]

DAVE, you're enjoying this,...aren't you?:D

[Dave Schwartz]

I am fortunate to say that I enjoy most of life.

I was serious, though. And you may have noticed that I share well with others.

[traveler]

Will we get a different answer based on whether the whales are betting into a pool or not? In other words, if they don't bet into a given race we can have a highly different metric than if they do.

[Dave Schwartz]

I would say that is probably true. It is safe to say that it would be more pronounced at tracks where whale money is more dominant, like the night tracks on the one extreme and CD, SA, KEE, GG, NYRA on the other.

[Tom]

In effect, should races that yield results of say, 5-2 on Mike's formula (or perhaps Dave's tweaks) be evaluated and bet differently than races that average 7-2?

Andy, sounds like the V.I.?

[098poi]

My original impression of this concept was this. Let's say you have 2 races of relatively equal class. 1 race has an extreme fav and the 2nd fav is the only other contender (based on the odds) and the others are way up there. In the other race you have a lukewarm fav 3-1 or so and the others are somewhat more even as you move up the odds ladder. If the 2nd fav wins in both races and then they meet in the future you could give extra credit to the horse from the more evenly matched race because he was in a race that was ( I am not sure about this next part) perceived(?) to be more competitive. Kind of simplistic but I am not a rocket scientist like some here. Not a put down just reality.

[zerosky]

someone mentioned Shannon's equation which I sometimes tamper with.
Its a measure of uncertainty i.e the more uncertain we are about the outcome
of an event the higher the number or Entropy (H)

H = sum of Pi(-log2*Pi)

Where Pi equals the probability of horse (in our case) i as reflected in the odds.

For example in Daves first example H= 2.05
in the second example H = 2.51 (Theoretically a more confusing race despite having fewer runners)

An information theory primer...
http://tinyurl.com/67yhoos

[Dave Schwartz]

Is that not field-size specific?

[TrifectaMike]

Dave,

It might be helpful (gain some insight) if you generate some joint probability distributions (let's keep them discrete, it'll be easier)

Let me suggest the following:

For race winning Favorites

Odds Index Intervals
2/5 [ao-a1] [a2-a3]..........[an-1 - an]
..... .................................................
..... .................................................

For race winning second Favorites

Odds Index Intervals
9/5 [ao-a1] [a2-a3]..........[an-1 - an]
..... .................................................
..... .................................................

For race winning third Favorites

Odds Index Intervals
7/2 [ao-a1] [a2-a3]..........[an-1 - an]
..... .................................................
..... .................................................

Mike (Dr Beav)

[Dave Schwartz]

Mike,

Sorry, but I do not today have the capability of doing this.

I am in deep development and it will likely be a month or more before I will be ready to play with this.


Dave

[sjk]

For what its worth. I don't think much going on that would point to profits.

from to group 0 1 2 3 4
0 0.99 fav -13.01%
1 1.99 2nd fav -14.83%
1 1.99 fav -14.99%
2 2.99 2nd fav -15.96%
2 2.99 3rd fav -20.15%
2 2.99 fav -17.67%
3 3.99 2nd fav -17.99%
3 3.99 3rd fav -18.83%
3 3.99 4th fav -21.01%
3 3.99 fav -19.85%
4 4.99 2nd fav -18.57%
4 4.99 3rd fav -19.42%
4 4.99 4th fav -19.45%

Pcts are returns

Here are the win pcts:

from to group 0 1 2 3 4
0 0.99 fav 52.55%
1 1.99 2nd fav 31.59%
1 1.99 fav 35.29%
2 2.99 2nd fav 24.22%
2 2.99 3rd fav 21.78%
2 2.99 fav 24.98%
3 3.99 2nd fav 18.79%
3 3.99 3rd fav 18.08%
3 3.99 4th fav 17.01%
3 3.99 fav 18.95%
4 4.99 2nd fav 15.24%
4 4.99 3rd fav 14.88%
4 4.99 4th fav 14.63%

[traveler]

Kash,

Maybe the best answer is to address Husker's question:



The answer is, "I don't know, but with a study I might be able to find out."

From my current point of view I will simply say that I expect their chances are different.

A good way to test that there MAY be something there to look at is to ask the question:

"Do 3/1 horses perform differently based upon their public choice number?"

This would address whether or not there were horses of lower odds than the 3/1 shot we are studying.


last 10000.
201-PubCh+1-rPubCh
WIN BETS
Odds PubCh Starts Pays Pct $Net IV PIV
3/1 1 192 45 23.4 $1.94 2.31 1.19
3/1 2 2,007 416 20.7 $1.73 1.63 1.06
3/1 3 1,111 215 19.4 $1.62 1.34 1.00
3/1 4 61 13 21.3 $1.82 1.34 1.11

Total 3,371 689 20.4 $1.71 1.55 1.02

Based upon the last 10,000 races from my database, I would say that there IS a difference, although the sample size does not allow for a very high correlation - still 80+%.


Regards,
Dave Schwartz

Here is another 10,000 race sample looking at just 3-1 shots x-refd with Public Choice - don't know what conclusion one would draw from this.


3-1.
1-rPubCh
----------------------------------------------------------------
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV HV
---------------------------------------------------------------
1 193 37 19.2 $1.61 1.86 0.97 1.16
2 1,976 403 20.4 $1.72 1.68 1.04 1.19
3 1,122 217 19.3 $1.62 1.45 1.00 1.11
4 91 20 22.0 $1.89 1.62 1.15 1.27
5 8 0.0 $0.00 0.00 0.00 0.00

Total 3,390 677 20.0 $1.68 1.60 1.0

[TrifectaMike]

For what its worth. I don't think much going on that would point to profits.

from to group 0 1 2 3 4
0 0.99 fav -13.01%
1 1.99 2nd fav -14.83%
1 1.99 fav -14.99%
2 2.99 2nd fav -15.96%
2 2.99 3rd fav -20.15%
2 2.99 fav -17.67%
3 3.99 2nd fav -17.99%
3 3.99 3rd fav -18.83%
3 3.99 4th fav -21.01%
3 3.99 fav -19.85%
4 4.99 2nd fav -18.57%
4 4.99 3rd fav -19.42%
4 4.99 4th fav -19.45%

Pcts are returns

Here are the win pcts:

from to group 0 1 2 3 4
0 0.99 fav 52.55%
1 1.99 2nd fav 31.59%
1 1.99 fav 35.29%
2 2.99 2nd fav 24.22%
2 2.99 3rd fav 21.78%
2 2.99 fav 24.98%
3 3.99 2nd fav 18.79%
3 3.99 3rd fav 18.08%
3 3.99 4th fav 17.01%
3 3.99 fav 18.95%
4 4.99 2nd fav 15.24%
4 4.99 3rd fav 14.88%
4 4.99 4th fav 14.63%

Thanks for the data. I'm not sure I understand what the data represents. Can you give a brief explanation of the data. For example, what is the meaning of 0 1 2 3 4 in row 1.

Thanks,
Mike (Dr Beav)

[sjk]

The column headings are the bottom of the odds range (which is the same as the left-most row heading). The next row heading is the top of the odds range. Sorry if the formatting is clumsy.

I only provided values where I had at least 1000 data points. The only group in the 0-.99-1 odds group which had at least 1000 was the favorites and not the 2nd, 3rd or 4th favorites. When it gets to 4 to 4.99-1 there were fewer than 1000 favorites so the only values relate to the 2nd, 3rd and 4th.

To pick a sample data point, for those horses which were 3rd favorite and which went off at between 3-1 and 3.99-1 the ROI (top table) was -18.83% and the win percentage (bottom table) was 18.08%.

My conclusion was that in each odds range the metrics related to favorites, 2nd favorites, 3rd favorites and 4th favorites were quite similar and the returns were generally near the take in all cases.

[TrifectaMike]

The column headings are the bottom of the odds range (which is the same as the left-most row heading). The next row heading is the top of the odds range. Sorry if the formatting is clumsy.

I only provided values where I had at least 1000 data points. The only group in the 0-.99-1 odds group which had at least 1000 was the favorites and not the 2nd, 3rd or 4th favorites. When it gets to 4 to 4.99-1 there were fewer than 1000 favorites so the only values relate to the 2nd, 3rd and 4th.

To pick a sample data point, for those horses which were 3rd favorite and which went off at between 3-1 and 3.99-1 the ROI (top table) was -18.83% and the win percentage (bottom table) was 18.08%.

My conclusion was that in each odds range the metrics related to favorites, 2nd favorites, 3rd favorites and 4th favorites were quite similar and the returns were generally near the take in all cases.

SJK,

OK. It's clear now. I'm not surprised of the results or your conclusions based on your data.

What I was hoping to see was a joint distribution of odds range and the Competitiveness Index (metric binned).

Mike (Dr Beav)

[sjk]

Something like this?

Expr2 0 1 2 3 4 5 6 7
1 28.76% 23.72% 18.20% 15.08% 12.96% 11.57% 10.17%
2 45.80% 32.34% 24.26% 18.49% 15.38% 12.55% 11.11% 9.84%
3 46.96% 34.67% 24.50% 18.81% 14.95% 12.51% 11.06% 9.52%
4 48.65% 35.08% 24.45% 18.70% 14.68% 12.70% 11.03% 9.50%
5 50.24% 35.66% 24.31% 18.28% 14.58% 12.13% 10.58% 9.53%
6 51.38% 36.90% 24.14% 17.49% 14.52% 11.83% 10.75% 8.88%
7 53.32% 35.09% 23.51% 17.81% 14.48% 12.18% 10.80% 9.63%
8 54.69% 36.83% 22.88% 17.82% 14.86% 12.55% 9.83% 8.56%
9 58.11% 35.02% 24.87% 17.49% 13.46% 9.99% 9.95% 7.35%

The column headings are odds groups 3 for 3 to 3.99 for example.
As I figure it your competitiveness index generally runs between .01 and .1 so the rows are groups based on int(100*[ci]).

Returns:

Expr2 0 1 2 3 4 5 6 7
1 -20.72% -17.27% -19.28% -17.98% -16.74% -13.88% -14.32%
2 -13.70% -16.73% -16.72% -18.28% -16.59% -19.33% -17.41% -16.89%
3 -14.24% -14.65% -16.17% -16.89% -19.03% -19.55% -17.81% -19.69%
4 -13.96% -14.79% -16.45% -17.42% -20.43% -18.41% -18.08% -19.88%
5 -13.39% -14.24% -17.12% -19.24% -21.09% -22.03% -21.37% -19.57%
6 -13.19% -13.51% -17.53% -22.74% -21.25% -23.80% -20.05% -25.16%
7 -12.65% -15.32% -19.57% -21.32% -21.34% -21.53% -19.72% -18.55%
8 -13.39% -14.18% -21.76% -20.62% -19.38% -19.31% -26.85% -27.85%
9 -10.32% -15.88% -14.46% -22.82% -26.97% -35.84% -26.12% -37.51%