A theory of maximum confusion based on win pool and exacta pool betting

[formula_2002]

A theory of maximum confusion based on win pool and exacta pool betting

It seems that in 26% of the race sample, the public is displaying a bit of confusion between the separate win and exacta pools, to such an extent that leads me to think that this confusion may indicate when a bet on the favorite is a very bad idea and when it is better idea.

SAMPLE SIZE IS NOW 296 RACES
93 favorites* won in 296 races for a 31% win rate.
The average odds on these favorites was $1.376 to 1

Based on their final odds, 128 of them should have won.

However, base on "Rigorous Data Analysis", a small cluster of them outperformed the entire set.

The cluster contained 36 winning favorites in 78 races.
Based on their odds, 37 of them should have won.

ALL FAVORITE ROI= 93/128= .73 ROI
CLUSTER FAVORITES = 36/37 = .98 ROI
All favorites outside of the cluster =57/91=.63 roi.


THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE

[DeltaLover]

How you define this cluster?

[formula_2002]

How you define this cluster?
By establishing a numerical string for the relationship of win pool betting and exacta pool betting,then finding, if possible, a significant group which exceeded the normal roi.
The numerical expressions that fall within the middle of the string seem to produice the higher roi that either end of the string.

Now don't ask how I generated the numerical string :) (just yet)

[DeltaLover]

By establishing a numerical string for the relationship of win pool betting and exacta pool betting,then finding, if possible, a significant group which exceeded the normal roi.
The numerical expressions that fall within the middle of the string seem to produice the higher roi that either end of the string.

Now don't ask how I generated the numerical string :) (just yet)

when you say numerical string what are you refering to?

Do you mean an expression which evaluates to a number?

[formula_2002]

when you say numerical string what are you refering to?

Do you mean an expression which evaluates to a number?

the short answer is yes

[DeltaLover]

thx

[formula_2002]

SAMPLE SIZE IS NOW 343 RACES
104 favorites* won in 343 races for a 30% win rate.

Based on their final odds, 146 of them should have won.

However, base on "Rigorous Data Analysis" a small cluster of them outperformed the entire set.

The cluster contained 39 winning favorites in 85 races.
Based on their odds, 40 of them should have won.

ALL FAVORITE ROI= 98/128= .71 ROI
CLUSTER FAVORITES = 36/37 = .98 ROI.

When eliminating the cluster plays form the entire set, the remaining plays produced a .61 roi, who's average final odds were 1.43-1.
I have a smaller sample of those races where I viewed the 2nd choice's returns
21 winners in 36 races returned a 2.11 roi. only 10 of those horses were expected to win.

THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE
__________________

[iceknight]

Has this theory been validated with experiments?

[formula_2002]

Has this theory been validated with experiments?
yep. I've been posting Live plays on selections forum.
However I'm unable to post all the plays.

The program runs automatically using Twinspire's web site, but the posting to PA is manual and I can not get all the plays posted.

[DeltaLover]

SAMPLE SIZE IS NOW 343 RACES
104 favorites* won in 343 races for a 30% win rate.

Based on their final odds, 146 of them should have won.

However, base on "Rigorous Data Analysis" a small cluster of them outperformed the entire set.

The cluster contained 39 winning favorites in 85 races.
Based on their odds, 40 of them should have won.

ALL FAVORITE ROI= 98/128= .71 ROI
CLUSTER FAVORITES = 36/37 = .98 ROI.

When eliminating the cluster plays form the entire set, the remaining plays produced a .61 roi, who's average final odds were 1.43-1.
I have a smaller sample of those races where I viewed the 2nd choice's returns
21 winners in 36 races returned a 2.11 roi. only 10 of those horses were expected to win.

THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE
__________________


I am sure that you know and realize that your sample is extremely small to justify any valid conclusion. The cluster that appears to be rich in winners most likely will regress to the mean as you sample will grow.... Unfortunately this type of behavior most of the times it just represents a random rather than a systematic behavior..

But again I am sure you know that, so I might be missing something from your posting.... Do you really mean something different?

[formula_2002]

I am sure that you know and realize that your sample is extremely small to justify any valid conclusion. The cluster that appears to be rich in winners most likely will regress to the mean as you sample will grow.... Unfortunately this type of behavior most of the times it just represents a random rather than a systematic behavior..

But again I am sure you know that, so I might be missing something from your posting.... Do you really mean something different?





Copy of my first note in the Selections forum

"What can a small sample tell you?

19 favorites won in 57 races for a 33% win rate.
Based on their final odds, 27 of them should have won.
However, base on "Rigorous Data Analysis" a small cluster of them outperformed the entire set.
The cluster contained 11 winning favorites in 20 races.
Based on their odds, 11 of them should have won and they did."

In almost 300 additional races, the cluster ROI stays about the same..
That continues to interest me

[DeltaLover]

Happy Holidays to all..

What can a small sample tell you?
19 favorites won in 57 races for a 33% win rate.
Based on their final odds, 27 of them should have won.
However, base on "Rigorous Data Analysis" :) a small cluster of them outperformed the entire set.
The cluster contained 11 winning favorites in 20 races.
Based on their odds, 11 of them should have won and they did.

What can a small sample tell you?
19 favorites won in 57 races for a 33% win rate.
Based on their final odds, 27 of them should have won.
However, base on "Rigorous Data Analysis" a small cluster of them outperformed the entire set.
The cluster contained 11 winning favorites in 20 races.
Based on their odds, 11 of them should have won and they did.

Formula,

let me see if I follow you:

You say that based in final odds 27 out of 57 should have won.

What exactly do you mean 'should have won'?

The most probable explanation is the following:

Based in final odds the corresponding probability of all the favorites was such to assume 27/57 winners. (In the same way that out of 10 coin flips we expect 5 heads).

In other words the mean probability (always based on odds) for each favorite was:

27/57 ~ 0.47

What you observed was:

19/57 ~ 0.33

A cluster of 20 contained 11 winners so we have the following fragmentation:

11 / 20 ~ 0.55

8 / 37 ~ 0.21


I have the following questions here:

1) what was the expected winning frequency based on odds for each cluster?

2) how many factors you are using to form your clusters? From what I can see the chance of over fitting seems pretty high, considering the limited sample.

3) how diverse are the handicapping factors represented by your sample? Obviously the larger the universe of factors the greatest is the chance to be looking at a pure random situation.

As a rule of thumb you need at least five winners for each factor you are using (this is the minimum accepted for chi square use). This applies to primitive and composite factors.

Overall I think you are following a correct approach but we are missing the necessary implementation details to form a more complete opinion about your work...

[formula_2002]

Deltalover, there are three figures I use. one is develope in the win pool the other in the exacta pool and the third is...magic :) . Their relationship determine the play.

It's all math based on the public odds

the odds cluster range hovers aroun 1.40-1.
Too few plays to make an incremental odds analysis.

The expected winners = 1/((odd)+1).


It's been my experience in this game to believe nothing will make a long term profit, but I'm always willing to look further into it. :)

I hope this answers some of your comments.

[formula_2002]

SAMPLE SIZE IS NOW 371 RACES
115 favorites* won in 371 races for a 31% win rate.

Based on their final odds, 160 of them should have won.

However, base on "Rigorous Data Analysis" a small cluster of them outperformed the entire set.

The cluster contained 43 winning favorites in 93 races.
Based on their odds, 44 of them should have won.

ALL FAVORITE ROI= 115/160= .72 ROI
CLUSTER FAVORITES = 43/44 = .98 ROI.
WHEN ELIMINATING THE CLUSTER PLAYS FROM THE ENTIRE SET, THE REMAINING FAVORITES RETURNED ONLY A .62 ROI.

WHAT A GREAT LAY BET!!!

WHICH BEGS THE QUESTION WHY BET SUCH FAVORITES??
OH, WHEN WILL BET FAIR BE LEGAL IN THE USA??



THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE
__________________

[dkithore]

Formula,

Most of what you write is above my head but your intent to share your findings are truly commendable. Thanks. I hope you inspire others on this board.

[turninforhome10]

That is nice work. Agree with you on Betfair. Have you done anything with Australian pools?

[davew]

ALL FAVORITE ROI= 115/160= .72 ROI
CLUSTER FAVORITES = 43/44 = .98 ROI.
WHEN ELIMINATING THE CLUSTER PLAYS FROM THE ENTIRE SET, THE REMAINING FAVORITES RETURNED ONLY A .62 ROI.

WHAT A GREAT LAY BET!!!

WHICH BEGS THE QUESTION WHY BET SUCH FAVORITES??
OH, WHEN WILL BET FAIR BE LEGAL IN THE USA??



It would be a great lay bet if you were at the same odds, but BETFAIR is a different market, sometimes with different favorites. If a favorite is overbet in the mutual pool, it does not necessarily mean it is overbacked in the betting market.

With simultcasting, many overbet favorites were not when they started the race. They get pounded late and the money shows up after the race starts (and sometimes after the race ends).

[Capper Al]

A theory of maximum confusion based on win pool and exacta pool betting

It seems that in 26% of the race sample, the public is displaying a bit of confusion between the separate win and exacta pools, to such an extent that leads me to think that this confusion may indicate when a bet on the favorite is a very bad idea and when it is better idea.

SAMPLE SIZE IS NOW 296 RACES
93 favorites* won in 296 races for a 31% win rate.
The average odds on these favorites was $1.376 to 1

Based on their final odds, 128 of them should have won.

However, base on "Rigorous Data Analysis", a small cluster of them outperformed the entire set.

The cluster contained 36 winning favorites in 78 races.
Based on their odds, 37 of them should have won.

ALL FAVORITE ROI= 93/128= .73 ROI
CLUSTER FAVORITES = 36/37 = .98 ROI
All favorites outside of the cluster =57/91=.63 roi.


THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE

Are you saying that if the horse is the favorite on the tote-board and the exact will pay it is a preferred wager?

[formula_2002]

[QUOTE=turninforhome10]That is nice work. Agree with you on Betfair. Have you done anything with Australian pools?[/QUOT

I think I may have.

If I did the pools were too small to consider.

[formula_2002]

Are you saying that if the horse is the favorite on the tote-board and the exact will pay it is a preferred wager?

I'm not sure what you mean.
I just look at the win pool favorite and go from there.

[formula_2002]

ALL FAVORITE ROI= 115/160= .72 ROI
CLUSTER FAVORITES = 43/44 = .98 ROI.
WHEN ELIMINATING THE CLUSTER PLAYS FROM THE ENTIRE SET, THE REMAINING FAVORITES RETURNED ONLY A .62 ROI.

WHAT A GREAT LAY BET!!!

WHICH BEGS THE QUESTION WHY BET SUCH FAVORITES??
OH, WHEN WILL BET FAIR BE LEGAL IN THE USA??



It would be a great lay bet if you were at the same odds, but BETFAIR is a different market, sometimes with different favorites. If a favorite is overbet in the mutual pool, it does not necessarily mean it is overbacked in the betting market.

With simultcasting, many overbet favorites were not when they started the race. They get pounded late and the money shows up after the race starts (and sometimes after the race ends).

My play would be a propostion bet.
I'd pick the horse to lose.
Should the horse win I'd pay track odds plus , say 25%, to any takers .
I don't know if Betfare can accomodate that proposition.

[Capper Al]

My play would be a propostion bet.
I'd pick the horse to lose.
Should the horse win I'd pay track odds plus , say 25%, to any takers .
I don't know if Betfare can accomodate that proposition.


What's a cluster favorite?

[formula_2002]

I am sure that you know and realize that your sample is extremely small to justify any valid conclusion. The cluster that appears to be rich in winners most likely will regress to the mean as you sample will grow.... Unfortunately this type of behavior most of the times it just represents a random rather than a systematic behavior..

But again I am sure you know that, so I might be missing something from your posting.... Do you really mean something different?


It rears it's ugly head..and it is recongnized in the win and lay bets posted in the selctions forum.. muck the system wth all previous 1000 :rolleyes:

the back bets have an actual to expected win ratio of .77
while the lay bets ratio is 1.18... just the opposite of what was expected

[formula_2002]

SAMPLE SIZE IS NOW 436 RACES
146 favorites* won in 436 races for a 33% win rate.
The average odds on these favorites was $1.52 to 1

Based on their final odds, 189 of them should have won.

However, base on "Rigorous Data Analysis", a small cluster of them outperformed the entire set.

The cluster contained 54 winning favorites in 119 races.
Based on their odds, 57 of them should have won.

ALL FAVORITE ROI= 146/189= .77 ROI
CLUSTER FAVORITES = 54/57 = .95 ROI
All favorites outside of the cluster =92/132=.70 roi.

WORTH STAYING WITH IT FOR A FEW HUNDRED RACES MORE :)
THE FAVORITE IS DETERMINED WITH 1 MINUTE TO POST TIME, WHICH MAY NOT BE THE POST TIME FAVORITE.MOST OFTEN, IT IS THE FAVORITE