Value that counts in theory

[Overlay]

Originally Posted by Capper Al

Okay let's try this from another way. Let's say your odds line on the type of race is 4/1. Your odds line on your top selection is 2/1. The horse is going off at 3/1. Your R&D of 10,000 races was two sets of 5,000 races with the second set verifying the conclusions from the first set. What would you do and why?

I size my wager based on the degree of overlay that the actual odds are offering on the horse in comparison to my personal full-field odds line for that specific individual race. I ignore the fact that the horse's actual odds are lower than the historical overall average odds for winners of that type of race.

[Overlay]

Originally Posted by jpren37

Anyone,

Would it be possible to find overlays in the win pools in exacta combinations? Back in 1984, I purchased a gauge created by Andy Anderson (I believe he was an accountant/statistician) which purported to be able to identify overlay exactas based on the tote board odds. Apparently he had done extensive research of California tracks covering many years/races evaluating exacta payoffs in relation to final odds. He further stated that his math and gauge were validated by Bill Ziemba.

He did very little handicapping; just watched the tote board. If an exacta combination was paying above 10% of what the gauge said it should pay, he would bet that exacta....e.g. if the potental winner's odds were 9-1 and the potential placer odds were 4-1, that combination should pay $130....if it paid less, it's an underlay and a bad bet.

Possibly back then, the tote board was more stable with just a few minutes to post so this might work? I experimented briefly with the gauge but found that it wasn't identifying enough overlays but perhaps I did not stick with it long enough.

Does anyone have any thoughts or experience about the gauge or concept?

I can't speak to the specific gauge you mentioned, although you might try doing a search for it on the board, using Andy Anderson's name, or some other appropriate term, such as exacta gauge (or calculator) or exacta payoff gauge (or calculator).

You seemed to start out asking about finding overlays in the win pool using exacta combinations, but then your discussion shifted to finding exacta combinations that are projected to pay more than they should, based on the win-pool odds of the horses involved. If you're interested in exacta betting rather than win-pool betting, there's been a fair amount of past discussion on the board concerning the use of formulas such as Harville and modified Harville for calculating the probabilities of exacta combinations. There have also been threads dealing with isolating place-horse "types" to fill out exactas, apart from considerations of the fair or actual odds of the horses involved.

Aside from the issue with the stability of the tote-board odds that you noted, I would think that there would also be the question of which exacta combinations to figure payoffs for (unless one were intending to calculate all possible payoffs). Do the instructions for the Anderson gauge prioritize or place a limit on that somehow?

[Capper Al]

Originally Posted by Overlay

I size my wager based on the degree of overlay that the actual odds are offering on the horse in comparison to my personal full-field odds line for that specific individual race. I ignore the fact that the horse's actual odds are lower than the historical overall average odds for winners of that type of race.

Why should one ignore the fact? Are we saying one set of facts is more valid than the other? Facts are facts either way.

[jorcus99]

Originally Posted by jpren37

Anyone,

Would it be possible to find overlays in the win pools in exacta combinations? Back in 1984, I purchased a gauge created by Andy Anderson (I believe he was an accountant/statistician) which purported to be able to identify overlay exactas based on the tote board odds. Apparently he had done extensive research of California tracks covering many years/races evaluating exacta payoffs in relation to final odds. He further stated that his math and gauge were validated by Bill Ziemba.

He did very little handicapping; just watched the tote board. If an exacta combination was paying above 10% of what the gauge said it should pay, he would bet that exacta....e.g. if the potental winner's odds were 9-1 and the potential placer odds were 4-1, that combination should pay $130....if it paid less, it's an underlay and a bad bet.

Possibly back then, the tote board was more stable with just a few minutes to post so this might work? I experimented briefly with the gauge but found that it wasn't identifying enough overlays but perhaps I did not stick with it long enough.

Does anyone have any thoughts or experience about the gauge or concept?

I think the win pool vs the exacta pool has become much more efficient in the last decade. The win payouts may swing some at the last minute at the tracks I play but the exacta payouts often swing much more. A 10% swing is not much room for error and exacta payouts can swing 50% on the last flash.

The other issue you have to deal with is that if you are going to exploit a statistical advantage you have to look for it in every race on every day at the tracks you built your data base on. The day you miss might be the day you were waiting for to get that positive ROI.

[Capper Al]

Originally Posted by Dave Schwartz

Al,

The first thing one must determine is whether or not the horses one makes 2/1 are profitable at 3/1, 4/1 or whatever.

Second, when you say "stats are built on 10,000 races," how many races are in the independent test of the method?

IOW, you build your system on "stats" from 10,000 races. Now, as you test forward, how many races have you tested against to verify that your system really can tell the difference between an overlay and an underlay?

Let's assume all your test have been made by your standers. That testing your 2/1 at 3/1 should be profitable overall, but now you are aware this race type needs 4/1. What do you do and why?

[jpren37]

Originally Posted by Overlay

I can't speak to the specific gauge you mentioned, although you might try doing a search for it on the board, using Andy Anderson's name, or some other appropriate term, such as exacta gauge (or calculator) or exacta payoff gauge (or calculator).

You seemed to start out asking about finding overlays in the win pool using exacta combinations, but then your discussion shifted to finding exacta combinations that are projected to pay more than they should, based on the win-pool odds of the horses involved. If you're interested in exacta betting rather than win-pool betting, there's been a fair amount of past discussion on the board concerning the use of formulas such as Harville and modified Harville for calculating the probabilities of exacta combinations. There have also been threads dealing with isolating place-horse "types" to fill out exactas, apart from considerations of the fair or actual odds of the horses involved.

Aside from the issue with the stability of the tote-board odds that you noted, I would think that there would also be the question of which exacta combinations to figure payoffs for (unless one were intending to calculate all possible payoffs). Do the instructions for the Anderson gauge prioritize or place a limit on that somehow?

Overlay,

If I'm not mistaken, Anderson used the Harville formulas to create his gauge. Before a race he would write the initial exacta payouts quickly across the top and down the side on a blank paper (similarly to the way it appears on the monitors now). At that time (his presentation at the meadowlands,1984) he felt the exacta payouts were fairly stable a few minutes before post time.

He felt it reasonable to bet #'s of combinations based on field size e.g. 6 horse race you might reasonably bet 5 combinations, 6 horse fields-7 combinations etc. By sliding/manipulating his gauge, you could match-up the odds and payouts fairly quickly. The point was that you simply only played those combinations that were "overlays"- even if you decided to bet only one combination.

What intrigued me was the fact that there was no guess work. On a straight win pool bet, as I interpret comments in this thread, you never really know if you're betting an overlay. (I guess you can analyze a large sample of your win bet results to see if you have a positive roi but how large a sample would that require.) Using his gauge, you knew where the exacta overlays were in any race; add in handicapping factors and you might reduce playable combinations even further.

There was a tape that came with the gauge that's locked away on my bookshelf somewhere. If there's any interest, I can look for it and provide more information

[jorcus99]

Originally Posted by Capper Al

Why should one ignore the fact? Are we saying one set of facts is more valid than the other? Facts are facts either way.

This fact is probably more general than the other facts and therefore less effective for a particular event.

[Overlay]

Originally Posted by Capper Al

Why should one ignore the fact? Are we saying one set of facts is more valid than the other? Facts are facts either way.

I'm not disputing that it is a fact, or that it is valid. But I don't see it as having any relevance to the way I handicap a race. (In other words, I don't believe that my handicapping performance will be adversely affected if I disregard it.) If you find a way to incorporate it into your handicapping that makes it is useful to you, go for it.

[Overlay]

Originally Posted by jpren37

What intrigued me was the fact that there was no guess work. On a straight win pool bet, as I interpret comments in this thread, you never really know if you're betting an overlay. (I guess you can analyze a large sample of your win bet results to see if you have a positive roi but how large a sample would that require.) Using his gauge, you knew where the exacta overlays were in any race; add in handicapping factors and you might reduce playable combinations even further.

I agree that you can determine whether horses comprising an exacta combination are receiving disproportionately more play in the exacta pool than as a win bet based solely on the amounts wagered. But that says nothing about whether the odds or projected payoffs on the horses/combinations represent betting value to begin with. My handicapping approach is based on the assumption that it is in fact possible to both identify horses or combinations that are overlaid in the win or exotic pools, and also determine the degree to which they are overlaid.

[jpren37]

Originally Posted by Overlay

I agree that you can determine whether horses comprising an exacta combination are receiving disproportionately more play in the exacta pool than as a win bet based solely on the amounts wagered. But that says nothing about whether the odds or projected payoffs on the horses/combinations represent betting value to begin with. My handicapping approach is based on the assumption that it is in fact possible to both identify horses or combinations that are overlaid in the win or exotic pools, and also determine the degree to which they are overlaid.

Overlay,

I don't think Anderson is suggesting identifying exacta combinations that are receiving disproportionate play. He's simply saying that by noting the odds and projected payoffs, his gauge will precisely tell you whether or not a particular exacta bet has value. If it is paying above breakeven (according to the gauge) it is a value bet! He is just confirming your assumption that it is possible to find overlaid combinations in the exotic pools and to what degree...I think the point is that his gauge identifies such overlays based on extensive research backed by math formulas and not on handicapping factors.

[Capper Al]

Originally Posted by Overlay

I'm not disputing that it is a fact, or that it is valid. But I don't see it as having any relevance to the way I handicap a race. (In other words, I don't believe that my handicapping performance will be adversely affected if I disregard it.) If you find a way to incorporate it into your handicapping that makes it is useful to you, go for it.

The fact about race type predicts how effective handicappers are in general with this type of race. The only way to come to grips with this situation is to have statistics based only on this type of race. This would be the intersection of both sets of stats.

[Jeff P]

Originally Posted by Capper Al

Let's assume all your test have been made by your standers. That testing your 2/1 at 3/1 should be profitable overall, but now you are aware this race type needs 4/1. What do you do and why?

Imho, an important objective would be the ability to test the performance of a model under a variety of conditions including (player defined) race type.

Hint: Think Bayes...

The first test "your 2/1 at 3/1"... that's A.

The second test "now you are aware this race type needs 4/1"... that's B.

I would develop a stand alone test for the intersection of A and B.



-jp

.

[Greyfox]

I've enjoyed following this thread.

Certainly I admire the amount of effort that some of you have put into horse racing and can appreciate the number of hours that it must take to come up with the data bases that you are developing.

I'm sure a number of neophytes can get ideas from this thread and perhaps push the boundaries of knowledge in this field even further.

Having said that, I do surgery with a very blunt scalpel and depending upon the race, I can fish with a very wide net. Other times, I can be much more precise and single a pick or not bet a race at all.
I trust the "computer" of my "gut," "mind," and previous experience as much as anything.
The only statistic, (aside from Past Performance lines and algorithms that I've developed) that I rely on is the growth or depreciation of my bank roll.

In my own floundering way, I've done okay for years at this sport, but it is not within my disposition to try and get rich at it. Large wagers on a continuous daily basis would take me out of my comfort level and I'm not so sure I'd enjoy the sport as much, even if I kept winning.

At any rate, gang - good discussion and good points that you are making.

[thaskalos]

Originally Posted by Capper Al

BRIS' PPs give average odds for the race type. For example, BRIS might say that with a 6 furlong race on the turf the average pay out might be $10.00. That would be 4/1 odds. Now what is the point if your odds line or the trackman has your horse going off at 3/1? Does one wait for 3/1 or 4/1. Let's keep it simple and say all other necessary markups have been included for this exampe.

That's like saying that the average PaceAdvantage member has an annual income of $50,000. While it appears to be useful information at first...it reveals absolutely nothing about the members individually.

[thaskalos]

Originally Posted by jpren37

Anyone,

Would it be possible to find overlays in the win pools in exacta combinations? Back in 1984, I purchased a gauge created by Andy Anderson (I believe he was an accountant/statistician) which purported to be able to identify overlay exactas based on the tote board odds. Apparently he had done extensive research of California tracks covering many years/races evaluating exacta payoffs in relation to final odds. He further stated that his math and gauge were validated by Bill Ziemba.

He did very little handicapping; just watched the tote board. If an exacta combination was paying above 10% of what the gauge said it should pay, he would bet that exacta....e.g. if the potental winner's odds were 9-1 and the potential placer odds were 4-1, that combination should pay $130....if it paid less, it's an underlay and a bad bet.

Possibly back then, the tote board was more stable with just a few minutes to post so this might work? I experimented briefly with the gauge but found that it wasn't identifying enough overlays but perhaps I did not stick with it long enough.

Does anyone have any thoughts or experience about the gauge or concept?

I am not familiar with the Anderson gauge...but, IMO, the concept is flawed.

The tote board often reveals information that is not apparent in the past performances...and the exacta payoffs do as well. You cannot automatically assume that your 5/2 shot is an overlay just because it is offered at odds of 8/1...nor can you automatically assume that your $85 exacta payoff is an overlay, just because it figures to pay $40.

Assesing value in this game is a lot trickier than that.

That's why winning remains as elusive as ever...regardless of how many "competent" oddslines and exacta gauges we find available in the horse handicapping marketplace.