Value Wagering and ROI

[traynor]

I have been analyzing the results of my own wagering patterns, and ran into some anomalies related to the concept of value wagering. In some cases, it may be counterproductive--rather than increasing profit, value wagering can diminish profit.

In a situation in which one has a positive ROI (at least on paper or printout), it is easy to say, "If I only bet the biggies, I am sure to profit even more." That may not be the case at all.

There is a (relatively) undisputed claim of a direct correlation between win frequency and final odds. What I have found (and continue to find) is that a big chunk of the ROI in any given process comes from the high win percentage/low mutuel end of the spectrum, rather than the (much) lower win percentage/high(er) mutuel end.

I use cleaned mutuel data, so outliers are not an issue. I suggest you may find it useful to break down the overall ROI by odds categories. You may find--as I did in a number of cases--that "value betting" may be misleading because it may toss out the major portion of the winners (and the profit from those winners that goes into creating the ROI in the first place). When ROI is recalculated on the basis of number of races actually won in the odds categories defined as value wagering, it may disappear altogether.

[DeltaLover]

Excellent thread topic as in touches the most challenging part of the creation of a valid fitness function (FF) which is the most challenging piece of the puzzle when trying to create a computerized betting model.

It should be clear that in theory, winning frequency is supposed to be completely irrelevant to the value of a betting model. This would have been the case if we were trying to solve the problem with a purely analytical approach, based in the concept of discovering overlays comparing a custom odds line against the public. For reasons that are not related to the topic, in real world such a puristic approach might not be feasible and this is why a more empirical approach, based on some type of a finite state machine might be preferable and easier to implement. The most fundamental question we need to answer towards the construction of such a FSM is how to decide between two of them which one is better than the other.

As I have stated in another thread an obvious rule to create a good FF is going to be something looking similar to the following:

Fitness = Final Balance / max drawdown

Although some tweaking of this based rule might be needed.

One of these tweaks, dramatically improving the behavior of this rule is if we somehow manage to consider the win – lose ratio (WLR) as another factor.

I have repeatably experience the situation where a very high performance model will behave poorer than a more modest one if the later appears to have a higher WLR when both are applied to fresh data, that were never before presented to them.

The definition of a good FF is so important, that I have concluded to even create models where the unknown was exactly this. One approach that I have concluded might be one of the best, is to consider each component of the FF (for example Final Balance, max drawdown and WLR) as separate vectors in a multidimensional space and use their Euclidean distance as the fitness. Such an approach tends to outperform almost any other although it adds significant implementation complexity, as the model needs to be retrained very often as new data are added to its universe.

[DeltaLover]

To make my approach easier to understand:

http://www.themindofagambler.com/fitness-example.png

[thaskalos]

To make my approach easier to understand:

http://www.themindofagambler.com/fitness-example.png

Oh...yeah.

Now I get it! :)

[proximity]

I have been analyzing the results of my own wagering patterns, and ran into some anomalies related to the concept of value wagering. In some cases, it may be counterproductive--rather than increasing profit, value wagering can diminish profit..........

.........I suggest you may find it useful to break down the overall ROI by odds categories.

after delta's excellent post, my post will probably read like ogre's (revenge of the nerds 2) contribution to the tri-lam desert island philosophy conversation..... what if c-a-t really spelled dog???... but here goes:

i recently did this too but broke the overlays down into four ranges and basically found after a rather minimal overlay level, that bigger wasn't necessarily better. conclusion: i'd be better off betting more on some smaller overlays and less on some larger overlays. (i looked separately at win, place, and show)

[DeltaLover]

To continue what I was saying before, trying to measure the performance of a model we need a fitness function receiving several parameters resulting to an arbitrary number than should be either maximized (or minimized). A major challenge we are facing going towards this direction is the degree of correlation among the various components. At this time we have a couple of threads going on, about the effectiveness of the trainer of jockey statistics in the final profitability. It is my constant impression that most of the posters seem to ignore the correlation between connections and the crowd's behavior. Similar is the reason, why 'formulator' stats seem to be so popular today among handicappers and gamblers, who try to think about the game. Elimination rules, also represent the same fallacy and again there is another thread going on right now about exactly this.

The reality is that the game needs way deeper and complex thinking in order to be resolved analytically. I am using the word analytically, just because I believe that there is still some room for an intuitive handicapper to beat the game, although as time goes by the margins for such a bettor are becoming narrower and thinner.

[dnlgfnk]

A layman here...

Is this similar to observing that an overlay at 6-1, whose "true odds" of 4-1 (approx. 16% win rate) produces a 12% ROI, but a 4-1 horse whose true odds are deemed to be 2-1 ( approx. 26% win rate) produces a 30% ROI?

[vegasone]

Seems like this gets beaten to death. If you have a system and you decide to exclude some horses from your system for betting purposes then it is no longer the same system.

[DeltaLover]

A layman here...

Is this similar to observing that an overlay at 6-1, whose "true odds" of 4-1 (approx. 16% win rate) produces a 12% ROI, but a 4-1 horse whose true odds are deemed to be 2-1 ( approx. 26% win rate) produces a 30% ROI?

1) What exactly are you referring to as been similar?

2) A 4-1 true odds does not translate to 16% win rate.

When expressing odds in the American way (like 4-1) you can think the first component as how many times you expect the horse to lose while the second how many times to win. Based in this :

We expect a 4-1 to win 1 time and lose 4, or win 1 in every 5 races which translates to:

4-1 ~ 20% win percentage

Using the same approach a 1 – 5 will win 5 out of 6 so

1-5 ~ 83% win percentage

Based in your example if you get 6-1 while your true odds are 4-1 the ROI produced will be 40% or 1.4 per wagerer dollar:

You expect to win 2 out of 10. Each time you win you collect $7 so every 10 races you collect $14 while your total bet is $10 so:

ROI = $14 / $10 = 1.4 or 40%

Your second example has 0.33 or 33% win rate (not 26%), you can do the rest of the calculation using the previous case.

You can also do the reverse: A 16% horse will lose 84 times for every 16 wins so its odds should be:

84 – 16 which is the same as

5 - 1

[traynor]

To continue what I was saying before, trying to measure the performance of a model we need a fitness function receiving several parameters resulting to an arbitrary number than should be either maximized (or minimized). A major challenge we are facing going towards this direction is the degree of correlation among the various components. At this time we have a couple of threads going on, about the effectiveness of the trainer of jockey statistics in the final profitability. It is my constant impression that most of the posters seem to ignore the correlation between connections and the crowd's behavior. Similar is the reason, why 'formulator' stats seem to be so popular today among handicappers and gamblers, who try to think about the game. Elimination rules, also represent the same fallacy and again there is another thread going on right now about exactly this.

The reality is that the game needs way deeper and complex thinking in order to be resolved analytically. I am using the word analytically, just because I believe that there is still some room for an intuitive handicapper to beat the game, although as time goes by the margins for such a bettor are becoming narrower and thinner.

I agree with the way deeper and complex thinking as a prerequisite. It is not so much that a more complex version of what is already out there is the solution (or a step toward a solution), but rather that a different way of thinking is needed--the proverbial "paradigm shift." There are so many things in horse racing that bettors seem to accept uncritically as "true" and to then behave as if they were true.

I recently started layering models by odds categories--not just "stuff out there" but my own models and my own performance. A hypothetical model: Strike Rate 50% plus, cleaned average mutuel over $5, and lots of plays. In theory, anything over $4 (at 50% win rate) is a good thing. To make a good thing even better, I (sometimes) pay attention to the many, many advocates of value betting, and pass on any race at even money or less. The theory is good, but it doesn't work that way in the real world.

In many of the models I have built in the last few months, both the majority of wins and the majority of profits are on the low side, not the high side. Those models closely approximate reality and are relatively predictive. Whether running simulations or betting in the real world, there is a distinct tendency for profit to drop dramatically when betting is restricted to the high end. According to the general principle of value betting, that should not be the case.

[traynor]

Seems like this gets beaten to death. If you have a system and you decide to exclude some horses from your system for betting purposes then it is no longer the same system.

Exactly.

[traynor]

Excellent thread topic as in touches the most challenging part of the creation of a valid fitness function (FF) which is the most challenging piece of the puzzle when trying to create a computerized betting model.

It should be clear that in theory, winning frequency is supposed to be completely irrelevant to the value of a betting model. This would have been the case if we were trying to solve the problem with a purely analytical approach, based in the concept of discovering overlays comparing a custom odds line against the public. For reasons that are not related to the topic, in real world such a puristic approach might not be feasible and this is why a more empirical approach, based on some type of a finite state machine might be preferable and easier to implement. The most fundamental question we need to answer towards the construction of such a FSM is how to decide between two of them which one is better than the other.

As I have stated in another thread an obvious rule to create a good FF is going to be something looking similar to the following:

Fitness = Final Balance / max drawdown

Although some tweaking of this based rule might be needed.

One of these tweaks, dramatically improving the behavior of this rule is if we somehow manage to consider the win – lose ratio (WLR) as another factor.

I have repeatably experience the situation where a very high performance model will behave poorer than a more modest one if the later appears to have a higher WLR when both are applied to fresh data, that were never before presented to them.

The definition of a good FF is so important, that I have concluded to even create models where the unknown was exactly this. One approach that I have concluded might be one of the best, is to consider each component of the FF (for example Final Balance, max drawdown and WLR) as separate vectors in a multidimensional space and use their Euclidean distance as the fitness. Such an approach tends to outperform almost any other although it adds significant implementation complexity, as the model needs to be retrained very often as new data are added to its universe.

That is a really interesting way to look at it. I would think that it is not just that the model needs to be retrained very often, but that the information provided to that model for the retraining would need to be considerably better than most that is available.

[traynor]

A layman here...

Is this similar to observing that an overlay at 6-1, whose "true odds" of 4-1 (approx. 16% win rate) produces a 12% ROI, but a 4-1 horse whose true odds are deemed to be 2-1 ( approx. 26% win rate) produces a 30% ROI?

Are you referring to the win rate and ROI of your own "true odds"?

[DeltaLover]

That is a really interesting way to look at it. I would think that it is not just that the model needs to be retrained very often, but that the information provided to that model for the retraining would need to be considerably better than most that is available.

Can you please give the definition of better data and how you can improve the publicly used Equibase ?

[dnlgfnk]

1) What exactly are you referring to as been similar?

2) A 4-1 true odds does not translate to 16% win rate.

When expressing odds in the American way (like 4-1) you can think the first component as how many times you expect the horse to lose while the second how many times to win. Based in this :

We expect a 4-1 to win 1 time and lose 4, or win 1 in every 5 races which translates to:

4-1 ~ 20% win percentage

Using the same approach a 1 – 5 will win 5 out of 6 so

1-5 ~ 83% win percentage

Based in your example if you get 6-1 while your true odds are 4-1 the ROI produced will be 40% or 1.4 per wagerer dollar:

You expect to win 2 out of 10. Each time you win you collect $7 so every 10 races you collect $14 while your total bet is $10 so:

ROI = $14 / $10 = 1.4 or 40%

Your second example has 0.33 or 33% win rate (not 26%), you can do the rest of the calculation using the previous case.

You can also do the reverse: A 16% horse will lose 84 times for every 16 wins so its odds should be:

84 – 16 which is the same as

5 - 1

Thanks, Delta.

I lean heavily upon a table Jeff P. posted some time ago regarding, I think, the So. Cal. tracks.

My thinking: Horse A is 6-1 on the board. If he truly ought to be a 4-1 risk (= 16% per Jeff P.), I'm betting $2 one hundred times and getting a $14 win price 16 times, = $224...or 12% return.

Horse B is 4-1 on the toteboard, wins like he should be 2-1 (26%, thanks Jeff), so I'm betting 100 races x $2, winning 26 of them at $10 = $260...=30%?

[dnlgfnk]

Are you referring to the win rate and ROI of your own "true odds"?

Hi, Traynor. Hope I explained a bit better above? I may be missing something.

[DeltaLover]

Thanks, Delta.

I lean heavily upon a table Jeff P. posted some time ago regarding, I think, the So. Cal. tracks.

My thinking: Horse A is 6-1 on the board. If he truly ought to be a 4-1 risk (= 16% per Jeff P.), I'm betting $2 one hundred times and getting a $14 win price 16 times, = $224...or 12% return.

Horse B is 4-1 on the toteboard, wins like he should be 2-1 (26%, thanks Jeff), so I'm betting 100 races x $2, winning 26 of them at $10 = $260...=30%?

you bet $2 100 times: TOTAL BET $200
Based in 4-1 he will win 20 and lose 80 so you collect $14 20 times (NOT 16) which is: $280

ROI = 280 / 200 = 1.4 or 40% return

[dnlgfnk]

Delta, are you allowing for takeout, without which I would agree? Jeff P.'s table is a year's compilation of the hit rate at each odds increment.

[DeltaLover]

Delta, are you allowing for takeout, without which I would agree? Jeff P.'s table is a year's compilation of the hit rate at each odds increment.

No, I do not consider takeout as this was not part of our hypothesis which referred to true odds. The way the problem is presented, it talks about true odds (which do not account for take out of course) and offered odds (which already have take out incorporated).

[DeltaLover]

Delta, are you allowing for takeout, without which I would agree? Jeff P.'s table is a year's compilation of the hit rate at each odds increment.

If by true odds you are referring after subtracting take out then your ROI is EVEN LARGER. Let's see why.

First of all, let's define true odds as the odds needed to have a fair game where there is a zero expectation.

Assume a pre- takeout pool of $100.
After a 20% take out you have left in the pool $80.

Horse A has 4-1 TRUE odds after the take out.

This means that the horse has a $16 bet on it AFTER APPLYING THE TAKE OUT and still wins often enough to be a fair bet.

So, before takeout the horse had on it:

$19.2

For simplicity lets assume that this horse in bet only be one person.

So, every race he bets the same amount: $19.2

When he wins he collects the whole pool after take out, which is $80 for a winning of 60.8

Since the odds are TRUE we expect after 100 races to break even.

Lets call the number of wins T. In this case number of loses are 100 - T

Obviously we need:

$60.8 * T = $19.2 * (100 – T) <=> T = 24

So the winning percentage of the horse should be 24% to present a fair bet when his TRUE ODDS are 4-1 after takeout!

Remember that in the previous case the winning percent was only 20%. If we are lucky enough to find this horse at 6-1 as the original example had it then:

After 100 bets costing $100 we will collect $7 X 24 = $168 for a 1.68 ROI or 0.68 (more than the 1.40 of the first example).

[JoeLong]

The way I see it is, in every race there is a 100% probability that there will be a winner. If you have a valid method of judging each individual contenders probability you can see how much probability is contained in your top-three choices. When that number is above 66% there is little reason to go beyond this top-three group when searching for value. The race then becomes pass/play based on any overlays in this group alone.

[Red Knave]

you bet $2 100 times: TOTAL BET $200
Based in 4-1 he will win 20 and lose 80 so you collect $14 20 times (NOT 16) which is: $280

ROI = 280 / 200 = 1.4 or 40% return

I think he is allowing for toteboard odds of 4-1 to be between 4.00 to 1 and 4.99 to 1 and using the higher odds (i.e. ~16%) and the lower payoff ($10) to be conservative.

[traynor]

Can you please give the definition of better data and how you can improve the publicly used Equibase ?

I came from a different background than most horse race bettors. My initial exposure to computer-based race analysis was a project contract to synthesize qualitative and quantitative data gathered by a group of bettors from Argentina. Basic triangulation, but interesting conceptually. One result of that project is that I have never considered that "everything I need to know" is contained in the Equibase data.

[DeltaLover]

I think he is allowing for toteboard odds of 4-1 to be between 4.00 to 1 and 4.99 to 1 and using the higher odds (i.e. ~16%) and the lower payoff ($10) to be conservative.

OK, this make sense then.

[PICSIX]

You guys are making this way too complicated....win betting with a 40% strike rate @ 3-1 odds or higher and you are printing money!

You can expect your maximum losing streak to be as many as 18 in a row if your proven strike is 40% (Degree of Certainty 99.99%).

The flip-side, your longest win streak could be 10 in a row. :jump:

We are betting on living, breathing horses. Not calculating Pi to the 1,000th decimal place!

3.141592653589793238462643383279502884197169399375 10
58209749445923078164062862089986280348253421170679
82148086513282306647093844609550582231725359408128
48111745028410270193852110555964462294895493038196
44288109756659334461284756482337867831652712019091
45648566923460348610454326648213393607260249141273
72458700660631558817488152092096282925409171536436
78925903600113305305488204665213841469519415116094
33057270365759591953092186117381932611793105118548
07446237996274956735188575272489122793818301194912
98336733624406566430860213949463952247371907021798
60943702770539217176293176752384674818467669405132
00056812714526356082778577134275778960917363717872
14684409012249534301465495853710507922796892589235
42019956112129021960864034418159813629774771309960
51870721134999999837297804995105973173281609631859
50244594553469083026425223082533446850352619311881
71010003137838752886587533208381420617177669147303
59825349042875546873115956286388235378759375195778
18577805321712268066130019278766111959092164201989

Good luck in the pursuit :ThmbUp:

[traynor]

I think the application of the result of large datasets to a smaller sample may be as conceptually impoverished as the reverse. I am far more interested in how my own selections fare at various odds levels than in how the overall selections of the overall general public fare. Basically, I am betting on my selections, not theirs, and the key points (for me, from a profit standpoint) are the win rate and return of my selections at specific odds ranges.

I think tossing the performance of the general public into the mix adds little of value, and serves more as a source of distraction than of illumination. I am not searching for The Answer or Truth-With-a-Big-T. I just want to cash more tickets at better prices.

[DeltaLover]

I came from a different background than most horse race bettors. My initial exposure to computer-based race analysis was a project contract to synthesize qualitative and quantitative data gathered by a group of bettors from Argentina. Basic triangulation, but interesting conceptually. One result of that project is that I have never considered that "everything I need to know" is contained in the Equibase data.

Does that mean that 'what you know' contains custom primary data, collected by you or someone else by watching the races and collecting them directly from the field?

Given the high number of race tracks and racing days, it is my impression that something like this should be a major commitment requiring a whole company of the size of Equibase.

[traynor]

You guys are making this way too complicated....win betting with a 40% strike rate @ 3-1 odds or higher and you are printing money!

You can expect your maximum losing streak to be as many as 18 in a row if your proven strike is 40% (Degree of Certainty 99.99%).

The flip-side, your longest win streak could be 10 in a row. :jump:

Good luck in the pursuit :ThmbUp:

That is the underlying assumption that is being challenged. It seems simple, it seems logical, it seems reasonable--and yet people lose buckets of money every day on that 99.99% Degree of Certainty.

The strike rate is NOT based on cherry picking. That changes the whole scenario. Unless the strike rate is broken down by odds categories, and is sufficient in the range of 3-1 odds and higher to be profitable, it can be a losing proposition.

[DeltaLover]

win betting with a 40% strike rate @ 3-1 odds or higher and you are printing money!

Sounds good on paper but... it ain't gonna happen in real life...

Sorry... Expecting to consistently get a 6-5 real odds horse at 3-1 is clearly a fantasy.

The crowd is pretty good in our days to allow something like this to happen..

[traynor]

Does that mean that 'what you know' contains custom primary data, collected by you or someone else by watching the races and collecting them directly from the field?

Given the high number of race tracks and racing days, it is my impression that something like this should be a major commitment requiring a whole company of the size of Equibase.

Not really. That would assume one was (interested in) betting every track every day. I have explained my approach to wagering a number of times, but basically, it is a lot of smaller wagers based on computer models, with larger wagers restricted to the tracks and circuits on which additional information (my own of from others that have proven to be reliable researchers) is available.

[traynor]

Sounds good on paper but... it ain't gonna happen in real life...

Sorry... Expecting to consistently get a 6-5 real odds horse at 3-1 is clearly a fantasy.

The crowd is pretty good in our days to allow something like this to happen..

Totalmente de acuerdo.

[DeltaLover]

Not really. That would assume one was (interested in) betting every track every day. I have explained my approach to wagering a number of times, but basically, it is a lot of smaller wagers based on computer models, with larger wagers restricted to the tracks and circuits on which additional information (my own of from others that have proven to be reliable researchers) is available.

OK, this means that you collect primary data directly from the field, creating a custom db of some racing attributes that are overlooked by the main data providers of the industry.

Something like this makes perfect sense and might put you well ahead of the curve.

[traynor]

OK, this means that you collect primary data directly from the field, creating a custom db of some racing attributes that are overlooked by the main data providers of the industry.

Something like this makes perfect sense and might put you well ahead of the curve.

I am not the only one. With all due respect to the software developers and data providers, there are a number of very serious bettors who regard wagering on any horse one cannot view close up as extremely foolish. For recreational or hobby handicapping, computers are nice. For serious betting, more is needed. Of course, that is highly subjective--there may be "whales" off in the hinterlands somewhere who rely exclusively on computer massaging of Equibase data as sufficient. I doubt it, but it may be possible.

I have a good friend who develops (very expensive) data mining and data analysis apps for a group that favors the Sha Tin/Happy Valley circuit. Those apps are considered augmentation of direct observation--not as replacement.

You may recall a discussion on a previous thread on the topic of "coding"--creating repeatable, teachable standards to facilitate quantitiative analysis of qualitative data.

[DeltaLover]

I am not the only one. With all due respect to the software developers and data providers, there are a number of very serious bettors who regard wagering on any horse one cannot view close up as extremely foolish. For recreational or hobby handicapping, computers are nice. For serious betting, more is needed. Of course, that is highly subjective--there may be "whales" off in the hinterlands somewhere who rely exclusively on computer massaging of Equibase data as sufficient. I doubt it, but it may be possible.

I have a good friend who develops (very expensive) data mining and data analysis apps for a group that favors the Sha Tin/Happy Valley circuit. Those apps are considered augmentation of direct observation--not as replacement.

You may recall a discussion on a previous thread on the topic of "coding"--creating repeatable, teachable standards to facilitate quantitiative analysis of qualitative data.

There is no doubt that having additional data not known to the public is the
best way to go about the game.

The question is what kind of data exist that are not part of the public domain
and how they can be gathered.

As an example, when I was betting full time, I experimented with an equipment
data base where I was storing things like tong ties, shadow rolls, nosebands
etc hoping to extract some edge out of it. The data gathering process was too
laborious though and the experiment lasted only a week or so.

In the past, I also have been thinking about the idea of collecting data like
wind measurement, moisture, the depth of the sand etc that could be used for the
creation of a better track variant. I have never materialized such plans
though, mainly because of the labor required but also because of my doubts about
the value of this approach compared to widely used figures like beyer or bris.

More that these I cannot think of any other kind of data that might present a
considerable handicapping value, so if you can give an example this might be
very helpful.

Another similar but still different type of data, would be a community data base
storing trip notes by many observers, using a specific language making it easy
for data mining and use. Something like this might be a great idea for an open
source project where anyone can contribute and use the data. I have presented
this idea in the past, here in PA, but it seems that nobody cares about it.

[traynor]

There is no doubt that having additional data not known to the public is the
best way to go about the game.

The question is what kind of data exist that are not part of the public domain
and how they can be gathered.

As an example, when I was betting full time, I experimented with an equipment
data base where I was storing things like tong ties, shadow rolls, nosebands
etc hoping to extract some edge out of it. The data gathering process was too
laborious though and the experiment lasted only a week or so.

In the past, I also have been thinking about the idea of collecting data like
wind measurement, moisture, the depth of the sand etc that could be used for the
creation of a better track variant. I have never materialized such plans
though, mainly because of the labor required but also because of my doubts about
the value of this approach compared to widely used figures like beyer or bris.

More that these I cannot think of any other kind of data that might present a
considerable handicapping value, so if you can give an example this might be
very helpful.

Another similar but still different type of data, would be a community data base
storing trip notes by many observers, using a specific language making it easy
for data mining and use. Something like this might be a great idea for an open
source project where anyone can contribute and use the data. I have presented
this idea in the past, here in PA, but it seems that nobody cares about it.

It is not so much that nobody cares about it as it is that very few are willing to do the work involved. Jim Selvidge discovered that many years ago when he was trying to set up an "observer network to share data." In essence, bettors want to "play," not work. Fiddling around with a computer is recreational, interesting, and (in the case of most users) not cognitively demanding. Punch a few buttons, stare at a few screens, and expect miracles to happen every day and twice a day on the weekends.

The information is out there for those willing to exert the effort to gain an edge. The information can be quantified, measured, and applied to modify and compliment computer analysis. The problem in an open-source project such as you describe is that the normal human tendency is to exert minimal effort to gain maximum reward--everyone expects someone else to do the heavy-lifting.

Whatever I define as a source of information would accomplish little more than putting a Korzybskian label on something to enable the reader to stop processing it. If I said "trips" the response would be, "Oh, yeah, I get all that in the charts." If I say "body language" the response would be, "Oh, yeah, I subscribed to Joe Takach's stuff years ago and lost a bundle." No matter what I said, the immediate, knee-jerk response would be to interpret whatever was said through a subjective perceptual filter to make it "understandable." It is not so much the information per se as it is the perceptual paradigm with which that information is processed.

For those now thinking, "Oh, blah blah, just give us some solid examples" I can only say that response clearly defines the problem. It's out there, but it don't come easy.

[traynor]

It is a bit like a software developer believing he or she can hire a "quant" to perform miracles, and that she or he will then be able to milk cash cows until the end of time. In the real world, if the quant is smart enough to develop the processes (or to discover which processes need to be developed, which seems to be the obstacle for most developers and the reason they think they need quants in the first place--they have no real clue what they are looking for or what they need), he or she will take the front money, go through a dazzling display of smoke and mirrors, then either keep the goodies to herself or himself, or sell it to the highest bidder. Or both, with the highest bidder getting only a limited version with a poison pill to assure future income for the quant when it self-destructs a little down the road.

I love this world. And everything in it. It seems so normal to me.

[MJC922]

It is not so much that nobody cares about it as it is that very few are willing to do the work involved. Jim Selvidge discovered that many years ago when he was trying to set up an "observer network to share data." In essence, bettors want to "play," not work. Fiddling around with a computer is recreational, interesting, and (in the case of most users) not cognitively demanding. Punch a few buttons, stare at a few screens, and expect miracles to happen every day and twice a day on the weekends.

The information is out there for those willing to exert the effort to gain an edge. The information can be quantified, measured, and applied to modify and compliment computer analysis. The problem in an open-source project such as you describe is that the normal human tendency is to exert minimal effort to gain maximum reward--everyone expects someone else to do the heavy-lifting.

Whatever I define as a source of information would accomplish little more than putting a Korzybskian label on something to enable the reader to stop processing it. If I said "trips" the response would be, "Oh, yeah, I get all that in the charts." If I say "body language" the response would be, "Oh, yeah, I subscribed to Joe Takach's stuff years ago and lost a bundle." No matter what I said, the immediate, knee-jerk response would be to interpret whatever was said through a subjective perceptual filter to make it "understandable." It is not so much the information per se as it is the perceptual paradigm with which that information is processed.

For those now thinking, "Oh, blah blah, just give us some solid examples" I can only say that response clearly defines the problem. It's out there, but it don't come easy.

:ThmbUp: You're touching on a lot of truths in here. If it's too much like work then most people want nothing to do with it. The same thing goes for the research side, most people don't do any research, they accept the literature at face value because that's much easier. They assume the research is valid because it's in a book. For trips and other visual observation, people who are good at it can't afford to share it, it's one of the only things left tipping the scales. Let's also mention pre and post race observation because the balance of what's out there with respect to the literature has done more harm than good, these things can be very valuable, just not every race. People are all about give me the winner of the next race, well most of the time you're going to bump into it by choosing contenders using speed, class, connections and all of the other stuff that's out in the open, unfortunately those races aren't where the long term profits come from, it's the one or two a day where you know something that's mostly off the radar and that doesn't necessarily mean 'price' horses, sometimes it's chalk but that's not all bad. As your research is suggesting the low odds overlay can be the foundation of profits.

[Overlay]

Sounds good on paper but... it ain't gonna happen in real life...

Sorry... Expecting to consistently get a 6-5 real odds horse at 3-1 is clearly a fantasy.

The crowd is pretty good in our days to allow something like this to happen..
Despite its remarkable aggregate accuracy and increasing sophistication, the public still routinely overbets individual horses (as well as exotic combinations involving those horses) on a race-by-race basis based on big figures, hot jockeys, "angles du jour", etc. -- not to mention the instances where (for whatever reason) the public is just flat-out wrong. And for every underlay, there has to be an overlay.

In my opinion, we will never reach a condition where the public sends every horse (and every exotic-wager combination) of at odds/payoffs corresponding precisely to its true chance of success. Overlays will always continue to exist, as long as the player doesn't limit his search for them to the horse that's also most likely to win, or to specific odds ranges, or to a particular type of wager.

[traynor]

Despite its remarkable aggregate accuracy and increasing sophistication, the public still routinely overbets individual horses (as well as exotic combinations involving those horses) on a race-by-race basis based on big figures, hot jockeys, "angles du jour", etc. -- not to mention the instances where (for whatever reason) the public is just flat-out wrong. And for every underlay, there has to be an overlay.

In my opinion, we will never reach a condition where the public sends every horse (and every exotic-wager combination) of at odds/payoffs corresponding precisely to its true chance of success. Overlays will always continue to exist, as long as the player doesn't limit his search for them to the horse that's also most likely to win, or to specific odds ranges, or to a particular type of wager.

It's basically an ego thing. Most bettors want to be right more than they want to win a profit. Especially in New York and SoCal. That is the big reason why many bettors depend on outside sources of income, and keep their wagers "manageable"--sometimes the desire to be right (or to be regarded by others as an "expert") far outweighs the desire to make money. Pretty strange behavior. Typical, but really strange.

[traynor]

:ThmbUp: You're touching on a lot of truths in here. If it's too much like work then most people want nothing to do with it. The same thing goes for the research side, most people don't do any research, they accept the literature at face value because that's much easier. They assume the research is valid because it's in a book. For trips and other the next race, well most of the time you're going to bump into it by choosing contenders using speed, class, connections and all of the other stuff that's out in the open, unfortunately those races aren't where the long term profits come from, it's the one or two a day where you know visual observation, people who are good at it can't afford to share it, it's one of the only things left tipping the scales. Let's also mention pre and post race observation because the balance of what's out there with respect to the literature has done more harm than good, these things can be very valuable, just not every race. People are all about give me the winner of something that's mostly off the radar and that doesn't necessarily mean 'price' horses, sometimes it's chalk but that's not all bad. As your research is suggesting the low odds overlay can be the foundation of profits.

It's actually a good thing. It fuels that endless rainbow chasing of developing just one more number crunching approach to picking winners as the be all and end all. It gives endless opportunities for creative "explanations" of why that approach should ultimately triumph because "they" are off doing it big time with their super-sophisticated computer apps and endless millions of bankroll.

The simple truth is that if they had the super-sophisticated computer apps and endless millions of bankroll, 98% of them would still lose. It is pretty much impossible to go mindlessly drifting through lala land while leveraging every advantage to win. That effort introduces a cold, hard reality that most lala landers avoid like the plague.

They can have lala land. I'll take the money.

[thaskalos]

It's basically an ego thing. Most bettors want to be right more than they want to win a profit. Especially in New York and SoCal. That is the big reason why many bettors depend on outside sources of income, and keep their wagers "manageable"--sometimes the desire to be right (or to be regarded by others as an "expert") far outweighs the desire to make money. Pretty strange behavior. Typical, but really strange.
I think that many players have a strong desire to make a profit in this game...but they don't know how to go about it. They create nice, logical methods of play...which get them "close" enough times to make them think that they are onto something. And then human nature takes over...and the self deception commences. I know...because I was one of them for more years than I care to admit.

It took me many years to finally realize that "logic" does not necessarily win out in this game.

[Phantombridgejumpe]

Money isn't the main goal of some that play.

[thaskalos]

Money isn't the main goal of some that play.
True, but the ones who play for the fun of it don't need much advice...do they?

[Phantombridgejumpe]

Each race is a puzzle. I love reading other opinions.

I studied some if the BC races for hours, listened to the radio about the races and wagered a total of under $80 on the races and cleared $200

I had tons of fun. I love trying to solve the mystery, and enjoy the sport.

Of course I would love to hit something big someday, but if it never happens I won't be depressed.

I don't do this as a job, it is a hobby. A serious and fun hobby.

I don't think I'm any better or worse handicapper because money isn't my ultimate goal.

[Dave Schwartz]

Money isn't the main goal of some that play.

We need more of these.

:ThmbUp:

[raybo]

Let's face it. There are lots of players out there that want to make money at this game. "You can lead them to water, but you can't make them drink".

There are also lots of excellent handicappers out there. However, they continuously lose money. Again, "You can lead them to water, but you can't make them drink".

Most players "self-destruct". They may have a good method, but they don't use it consistently, in a very disciplined way. No matter if they take Traynor's advice, or Delta's, or Thaskalos's, or Dave's, or mine, or anyone else's, they still must implement it, in real time, with the money on the line, without emotion or angst. How many out there can do that? Some say 2% or less. Regardless of the exact percentage, it is absolutely a very, very small percentage.

All this talk about which handicapping method, or which statistical model they use, or which theoretical/scientific approach they use, or which wagering method is "the truth and the light", it still comes down to the individual making the proper selections and wagering the correct way, consistently. It will always be this way, sure the public's hit rate on favorites may continue to increase, but the odds on those horses will also continue to decline, so the "public" is no better off by becoming better at selecting horses, they will still lose, because they refuse to get off the favorite and onto horses that pay more. They will never learn because they want to cash tickets, even if they are losing ones.

One member stated the truth already about samples, a "model" or "method" based on past results includes all the plays in that sample. Removing a subset of that, based on odds, destroys the model/method, unless you have multiple subsets based on multiple odds ranges. Then you have a multifaceted "model" or "method", and one can then effectively discard those odds ranges that do not produce net long term profit. Without having multiple subsets within the model, if any portion of that model is removed, the model no longer exists. So, you have gained nothing, but rather, have lost much.

[CincyHorseplayer]

I have been analyzing the results of my own wagering patterns, and ran into some anomalies related to the concept of value wagering. In some cases, it may be counterproductive--rather than increasing profit, value wagering can diminish profit.

In a situation in which one has a positive ROI (at least on paper or printout), it is easy to say, "If I only bet the biggies, I am sure to profit even more." That may not be the case at all.

There is a (relatively) undisputed claim of a direct correlation between win frequency and final odds. What I have found (and continue to find) is that a big chunk of the ROI in any given process comes from the high win percentage/low mutuel end of the spectrum, rather than the (much) lower win percentage/high(er) mutuel end.

I use cleaned mutuel data, so outliers are not an issue. I suggest you may find it useful to break down the overall ROI by odds categories. You may find--as I did in a number of cases--that "value betting" may be misleading because it may toss out the major portion of the winners (and the profit from those winners that goes into creating the ROI in the first place). When ROI is recalculated on the basis of number of races actually won in the odds categories defined as value wagering, it may disappear altogether.

Traynor,I'm not trying to be a jerk here but this is kindergarten.Anybody that has kept betting records knows that odds ranges have expected win percentages.When you talk about odds lower than 1-1 people do indulge that as an ego thing.Realistically we could all change our win percentages to be gaudy if we indulged those brackets.Over time you realize one immortal and undisputable truth=F-that.But I agree with you absolutely about the lower odds spectrum of say $6-8 winners.Whatever qualifiers anybody uses these horses should win about 40% of the time.I've kept records since 2001 and that's the case.This,as you say balances out the longshot chase.I've been critical about pure overlay play just because I realize that the % distribution so inexorably plummets as you descend your hierarchy of contenders.Is an overlay still an overlay if your 4th choice wins 8% over time?I don't think so.

But while I don't indulge the pure pie in the sky approach I do avoid the garbage bin of sub 2-1 horses.There are so many countless reasons for why horses end up at those odds we don't even need to go there.But they are just normal bet againsts for all of us.The only reason I ever pass races is if I think there is a legitimate favorite and their odds are sub 2-1.Once every 100 bets I bet a horse lower,and no lower than 1-1.This was just a humdrum philosophy til I realized the obvious.That I eliminate those particular races, favorites in races I do play lose 85% of the time.I could almost throw darts and win.There is such a such a fine balance in all of this but it is a balance and both odds spectrums are necessary.I don't involve the extreme lower bracket merely because of personal taste.But I think somewhere Delta has a mathematical formula for my intuition!

[Capper Al]

Good topic. The best way to play a niche has always been a hot topic for me. If one has a niche with a 40% or more strike rate, would it be better just to go for the win instead of the value? Obviously beating a 5/2 average payout would be profitable at 40%. But the old idea of just playing those horses over 5/2 might not work. Let's say the hit rate was 70% under 5/2 and, therefore, 30% over and including 5/2. Now 5/2 loses money at over 5/2 and is profitable under 5/2. Ainsle warned about averages. His joke was about the man that drowned crossing a river with an average depth of 4 feet.

[CincyHorseplayer]

Good topic. The best way to play a niche has always been a hot topic for me. If one has a niche with a 40% or more strike rate, would it be better just to go for the win instead of the value? Obviously beating a 5/2 average payout would be profitable at 40%. But the old idea of just playing those horses over 5/2 might not work. Let's say the hit rate was 70% under 5/2 and, therefore, 30% over and including 5/2. Now 5/2 loses money at over 5/2 and is profitable under 5/2. Ainsle warned about averages. His joke was about the man that drowned crossing a river with an average depth of 4 feet.

No it doesn't.If the bottom level of the over 5/2 group is $7 the average mutual is going to be higher.My 10 year average mutual is 9.47.This year it's 13.13.I'd have to eliminate the 2-1's from that group though.

[Phantombridgejumpe]

With odd fluctuations at many tracks at the last second - or ten seconds after the race starts - how do you know you are betting at 5-2 and not 2-5?

Do you only play large-pool tracks?

[CincyHorseplayer]

With odd fluctuations at many tracks at the last second - or ten seconds after the race starts - how do you know you are betting at 5-2 and not 2-5?

Do you only play large-pool tracks?

I only experience the late odds drops roughly twice per 100 bets.It's usually of the 4-1 to 5/2 variety but occasionally the 2-1 to 1-1.It's usually at small tracks too.Since I love turf racing I usually bet New York tracks or Kentucky and Florida,Fair Grounds.Pure dirt I usually only bet Ohio and Oaklawn.

[Phantombridgejumpe]

Given the conversations on the board, and my own observations the 2/100 seems low. I'm glad you have avoided that issue for the most part.

[CincyHorseplayer]

Given the conversations on the board, and my own observations the 2/100 seems low. I'm glad you have avoided that issue for the most part.

I think it's simply because I detest and avoid favorites!Being on the horses that end up favorites is a miserable experience from every angle.It's a lose-lose.My experience with odds is I had to talk my way out of low odds winners and talk myself back into taking 2-1 when the horse is dominant.I'm almost thinking about starting a thread called "The 2nd Life of a Handicapper".We all go through that 1st learning curve,however many years it takes,and you come out on the other end doing the same things but with a vastly different perspective.

[shouldacoulda]

You guys are making this way too complicated....
Agreed. I didn't play for a good part of the year. I believe I stopped in March and started again in early October. I did think about it much of that time and trying to figure out why I could pick winners and yet lose money. I believe I had my epiphany this past week.

Instead of concentrating on the winner, I try to look for the value horse that will blow up the exacta and the triple. Those are my two wagers that I am profitable in. I don't like horizontal wagers even though I have come close to hitting the place 9 too many times. I don't play it much but have had 8 out of 9 at least 4 times. Last effort was 7 winners 2 place and 1 show. I eliminated the winner from the last race selection to keep the wager down. I play with a small bankroll and can't afford long losing streaks or sacrificing too much of my bankroll on a single wager. Actually I can, but I refuse to burn money in an attempt to prove a point. I work too hard for it. That mistake cost me $1893.00 but I digress.

Since January 1 my ROI is -.10 over all wager types. Total wagers 1172.
Since October 1 to yesterday my roi for all wager types is -.01. Total wagers 923.

Since January 1 my roi for exacta wagers is +.31 total wagers 326 hit rate 10.2%.

Since October 1 my roi for exacta wagers is +.45 total wagers 278 hit rate 10.43%.

Since January 1 my roi for triples is +.27. Total wagers 78, hit rate 10.26%

Since October 1 my roi for triples is +.35. Total wagers 69, hit rate 10.14%.

Here in lies the key. Betting to win only produced a negative roi of -.27 with a hit rate of 14.78% over 978 wagers since January 1. Since October 1st it's slightly better but still negative. However, since January 1 Place wagers are -.02 for 109 wagers and a hit rate of 24.77. Betting win/place for the same time period (something I started doing recently after reviewing my play and a post by CJ) The hit rate goes to 26.72% with a roi of -.06 over 116 wagers.

Before October 1st I just played 2 horses boxed usually for a dollar. After October 1st I started keying horse with multiple horses. Usually 2 others sometimes (rarely) 3. Also for a dollar. I know some will scoff at a 2 dollar bettor but If I can't do it with a dollar and two dollar wagers I most likely won't do it with more decimal places.

What does this mean? The same thing that Thaskalos has been saying since I joined the board. Exotic wagers! Andy Beyer also said trying to grind out a profit on win place and show betting isn't going to happen.

The object of the game for me has turned into who will be on or close to the lead at the wire with high odds. I could care less about what a made up odds line is. Tote odds matter most. It dictates the payoff. Favorites lose 63% of the time. That leaves room for a good exotic payoff 64% of the time. Most people will key the chalk. I key anti chalk and don't bother if the chalk looks like a walk off winner at even money. I play all kinds of tracks from the big circuits to Los Alimitos and Turf Paradise and Mountaineer. I work all week and play a lot at night. They are usually the smaller tracks with cheaper less consistent runners. There are still too many races to bother trying to beat a solid favorite. Believe me, I am a timid bettor. A bunch of times my exacta key selections turn out to make up the triple. I don't like going for more than 6-8 dollars a race right now. I think that will change soon though. I think if I passed more races my percentages would go up too. Lets face it, 923 wagers from 10/6 to 11/10 is a bit much.

Great thread BTW

[CincyHorseplayer]

Agreed. I didn't play for a good part of the year. I believe I stopped in March and started again in early October. I did think about it much of that time and trying to figure out why I could pick winners and yet lose money. I believe I had my epiphany this past week.

Instead of concentrating on the winner, I try to look for the value horse that will blow up the exacta and the triple. Those are my two wagers that I am profitable in. I don't like horizontal wagers even though I have come close to hitting the place 9 too many times. I don't play it much but have had 8 out of 9 at least 4 times. Last effort was 7 winners 2 place and 1 show. I eliminated the winner from the last race selection to keep the wager down. I play with a small bankroll and can't afford long losing streaks or sacrificing too much of my bankroll on a single wager. Actually I can, but I refuse to burn money in an attempt to prove a point. I work too hard for it. That mistake cost me $1893.00 but I digress.

Since January 1 my ROI is -.10 over all wager types. Total wagers 1172.
Since October 1 to yesterday my roi for all wager types is -.01. Total wagers 923.

Since January 1 my roi for exacta wagers is +.31 total wagers 326 hit rate 10.2%.

Since October 1 my roi for exacta wagers is +.45 total wagers 278 hit rate 10.43%.

Since January 1 my roi for triples is +.27. Total wagers 78, hit rate 10.26%

Since October 1 my roi for triples is +.35. Total wagers 69, hit rate 10.14%.

Here in lies the key. Betting to win only produced a negative roi of -.27 with a hit rate of 14.78% over 978 wagers since January 1. Since October 1st it's slightly better but still negative. However, since January 1 Place wagers are -.02 for 109 wagers and a hit rate of 24.77. Betting win/place for the same time period (something I started doing recently after reviewing my play and a post by CJ) The hit rate goes to 26.72% with a roi of -.06 over 116 wagers.

Before October 1st I just played 2 horses boxed usually for a dollar. After October 1st I started keying horse with multiple horses. Usually 2 others sometimes (rarely) 3. Also for a dollar. I know some will scoff at a 2 dollar bettor but If I can't do it with a dollar and two dollar wagers I most likely won't do it with more decimal places.

What does this mean? The same thing that Thaskalos has been saying since I joined the board. Exotic wagers! Andy Beyer also said trying to grind out a profit on win place and show betting isn't going to happen.

The object of the game for me has turned into who will be on or close to the lead at the wire with high odds. I could care less about what a made up odds line is. Tote odds matter most. It dictates the payoff. Favorites lose 63% of the time. That leaves room for a good exotic payoff 64% of the time. Most people will key the chalk. I key anti chalk and don't bother if the chalk looks like a walk off winner at even money. I play all kinds of tracks from the big circuits to Los Alimitos and Turf Paradise and Mountaineer. I work all week and play a lot at night. They are usually the smaller tracks with cheaper less consistent runners. There are still too many races to bother trying to beat a solid favorite. Believe me, I am a timid bettor. A bunch of times my exacta key selections turn out to make up the triple. I don't like going for more than 6-8 dollars a race right now. I think that will change soon though. I think if I passed more races my percentages would go up too. Lets face it, 923 wagers from 10/6 to 11/10 is a bit much.

Great thread BTW

I grew up with Andy Beyer books too and bet accordingly.In the process and being blown out of a lot of wagers where my horse won plenty at 4-1,9-1,16-1,I knew that it was bullshit.Andy,at least from what you read from his books doesn't play seriously very often and wants roller coaster rides when he does.That doesn't apply to a guy playing a steady game trying to evaluate where he's at,even if it's once a week.Having multiple,well thought out attacks on this game is where it's at.If anyone is serious why don't you want to be the best?I want to hit for average.I want to hit for power.I want to get on base(cash rate).I want to have a .900+ OPS.I want to win at all my separate wagers.Achieving multiple levels of steady IMO is the only way to make that happen.Including win wagers.

[Capper Al]

No it doesn't.If the bottom level of the over 5/2 group is $7 the average mutual is going to be higher.My 10 year average mutual is 9.47.This year it's 13.13.I'd have to eliminate the 2-1's from that group though.

I stand corrected. All I could say was the 5/2 and over would not be profitable. The under 5/2 group would be another study.

[CincyHorseplayer]

I stand corrected. All I could say was the 5/2 and over would not be profitable. The under 5/2 group would be another study.

It is distributed differently.The $6-8 group I finds wins 40% of the time but that 7/2 to 6-1 group drops to 25%.It gets more precipitous after that obviously.I think the biggest coming of age process for all of us is not being pissed off and bitter when we see good horses win races that represent bad bets.Once over that hump it's all gravy!

[traynor]

Traynor,I'm not trying to be a jerk here but this is kindergarten.Anybody that has kept betting records knows that odds ranges have expected win percentages.When you talk about odds lower than 1-1 people do indulge that as an ego thing.Realistically we could all change our win percentages to be gaudy if we indulged those brackets.Over time you realize one immortal and undisputable truth=F-that.But I agree with you absolutely about the lower odds spectrum of say $6-8 winners.Whatever qualifiers anybody uses these horses should win about 40% of the time.I've kept records since 2001 and that's the case.This,as you say balances out the longshot chase.I've been critical about pure overlay play just because I realize that the % distribution so inexorably plummets as you descend your hierarchy of contenders.Is an overlay still an overlay if your 4th choice wins 8% over time?I don't think so.

But while I don't indulge the pure pie in the sky approach I do avoid the garbage bin of sub 2-1 horses.There are so many countless reasons for why horses end up at those odds we don't even need to go there.But they are just normal bet againsts for all of us.The only reason I ever pass races is if I think there is a legitimate favorite and their odds are sub 2-1.Once every 100 bets I bet a horse lower,and no lower than 1-1.This was just a humdrum philosophy til I realized the obvious.That I eliminate those particular races, favorites in races I do play lose 85% of the time.I could almost throw darts and win.There is such a such a fine balance in all of this but it is a balance and both odds spectrums are necessary.I don't involve the extreme lower bracket merely because of personal taste.But I think somewhere Delta has a mathematical formula for my intuition!

I think way too many conclusions are drawn from generic data that are not (as) applicable to subsets of data. I am referring to my own selections, and my own analysis process, rather than a simplistic tote board view. I am layering by the actual odds of my selections--not of selections in general. In that layering, I find that the majority of profit is NOT in the higher ranges, but rather in the lower ranges.

Again, this ONLY applies to my own analysis of my own selection and wagering processes, but MAY apply to the selection and wagering processes of others. What I find is that by restricting wagers to "value bets" over some specific odds value, my (actual, real world) profits drop. Not go away--drop. YMMV.

What I am suggesting is that when database results and paper wagering results are discounted as interesting but not necessarily relevant to events in the real world, profits may increase by focusing on areas other than "value betting" or "only bet at some odds range or above." Some of the best returns I have had were in races that odds were below a given cutoff point at post time, but a last minute infusion of cash (on other entries) elevated the odds to an "acceptable range" after betting was closed. Conditional wages would have failed, and yet the results would be "included" in regressions.

While the concept of only wagering at 2-1 or higher is interesting, I find that there are major earning opportunities in other areas. Most notably, consistent long-term gains from horses BELOW 2-1. YMMV.

[traynor]

We need more of these.

:ThmbUp:

There is nothing wrong with playing at race analysis. As long as those who are playing make it clear that their advice is intended only for other players, and NOT for those intent on making a profit.

[traynor]

It may be that so many are focused on value betting that they have tipped the scales relative to those willing to get a winner at any price, and actually increased the earning opportunities in lower odds ranges. That phenomena may be a result of last-minute conditional wagers dumped into the mutuel pools.

I am not suggesting that anyone "accept" my ideas or observations. Only that you may find it useful to look a bit more closely at the potential in lower odds ranges.

How things were 10 years ago, or what my own personal records from last year or the year before indicate, is not the issue. I am betting today, in the real world, not in (what has now become) the purely conceptual world that may have existed at some time in the past. In that real world of today's wagering, I find that there is a LOT of profit to be made in lower odds ranges. Again, YMMV.

[Dave Schwartz]

Money isn't the main goal of some that play.

Oops. That was the post I was responding to when I said "We need more of these."



Sorry.

[Phantombridgejumpe]

I don't charge for my picks or make any claims.

That being said, I will not add any disclaimers.

[raybo]

There is nothing wrong with playing at race analysis. As long as those who are playing make it clear that their advice is intended only for other players, and NOT for those intent on making a profit.

That would be unneeded, the people who are in it only for profit will know when such a player gives bad advice. So, in effect, they would only be advising those who don't know better.

[MJC922]

It may be that so many are focused on value betting that they have tipped the scales relative to those willing to get a winner at any price, and actually increased the earning opportunities in lower odds ranges. That phenomena may be a result of last-minute conditional wagers dumped into the mutuel pools.

I am not suggesting that anyone "accept" my ideas or observations. Only that you may find it useful to look a bit more closely at the potential in lower odds ranges.

How things were 10 years ago, or what my own personal records from last year or the year before indicate, is not the issue. I am betting today, in the real world, not in (what has now become) the purely conceptual world that may have existed at some time in the past. In that real world of today's wagering, I find that there is a LOT of profit to be made in lower odds ranges. Again, YMMV.

Well stated. I would echo the same YMMV with the caveat that in my case the approach used to detect the low odds overlays was comprehensive and I suspect those using a less comprehensive approach need not apply. i.e. If someone is an angle player for instance who looks to pull in bombs from horses showing four negatives to every one positive then minimum acceptable odds becomes mandatory to profitable play, otherwise low odds are just a drain on profits because negatives are being glossed over. e.g. When I tested using moderately less comprehensive approaches I found the 2-1 cutoff became important.

[traynor]

Well stated. I would echo the same YMMV with the caveat that in my case the approach used to detect the low odds overlays was comprehensive and I suspect those using a less comprehensive approach need not apply. i.e. If someone is an angle player for instance who looks to pull in bombs from horses showing four negatives to every one positive then minimum acceptable odds becomes mandatory to profitable play, otherwise low odds are just a drain on profits because negatives are being glossed over. e.g. When I tested using moderately less comprehensive approaches I found the 2-1 cutoff became important.

The weird thing (again with the caveat that I am only looking at my own selections and wagers) is that this was not happening two years ago, and it was not happening for most of last year. The algorithms that make the selections are essentially the same. However, when layered in odds ranges, the profit from the lower end odds has increased substantially. I have no idea why, but it may be worthwhile considering.

The ROI of higher odds ranges is essentially consistent. It is only the ROI in the lower odds ranges that seems to have increased. The only reason I can think of offhand is that more bettors may be wagering online, and using conditional wagers set to the higher odds ranges. Either that or some whales have a bug in their apps that is putting them on the wrong horses.

[MJC922]

The only reason I can think of offhand is that more bettors may be wagering online, and using conditional wagers set to the higher odds ranges. Either that or some whales have a bug in their apps that is putting them on the wrong horses.

Yeah, conditional is definitely part of it, I think it's common and probably an increasing portion of the money is being bet by hobbists recently getting their heads around automated processes. I'll blast in 70 per day or more before I go to work just toying with it and find maybe 8 or 9 went to tote. It's just easy to do so people are probably demanding higher prices than they should, and don't forget all of that bad software out there too with linemaking coming out of it, every race spits out five horses between 3-1 and 5-1.

Just as an observer you can see these corrections, anything with decent numbers that's legit 3-1 but going off at 5-1 with 2 MTP is really just DOB but inevitably marches its way down to 5-2 or even favored by the time the gate opens. I notice one somewhat infrequent way to combat this from time to time is to look for horses that are very solid on paper legit 2/5 or 1/2 get drilled early and open 2/5 or 3/5 and then up a notch from there to 4-5, they seem to at least hold steady at 4-5 or even money from all of the late conditional 'dumb' money coming in. I'll take even money on 2/5 all day long, unfortunately horses that I make 2/5 don't come around very much. The mad bomber at MTN is all about playing into this stuff, there's a method to that madness but pulling it out late is really blatant abuse of the 'conditional' crowd which shouldn't be allowed. Dahlman I think mentioned it too, bury them early so people jump off. By the time it's official people are kicking themselves the horse paid $4.00 and should've paid $2.80. It's brilliant but you gotta be good and be right about it, not pulling money out with cancellations which I doubt he would ever do.

[traynor]

Yeah, conditional is definitely part of it, I think it's common and probably an increasing portion of the money is being bet by hobbists recently getting their heads around automated processes. I'll blast in 70 per day or more before I go to work just toying with it and find maybe 8 or 9 went to tote. It's just easy to do so people are probably demanding higher prices than they should, and don't forget all of that bad software out there too with linemaking coming out of it, every race spits out five horses between 3-1 and 5-1.

Just as an observer you can see these corrections, anything with decent numbers that's legit 3-1 but going off at 5-1 with 2 MTP is really just DOB but inevitably marches its way down to 5-2 or even favored by the time the gate opens. I notice one somewhat infrequent way to combat this from time to time is to look for horses that are very solid on paper legit 2/5 or 1/2 get drilled early and open 2/5 or 3/5 and then up a notch from there to 4-5, they seem to at least hold steady at 4-5 or even money from all of the late conditional 'dumb' money coming in. I'll take even money on 2/5 all day long, unfortunately horses that I make 2/5 don't come around very much. The mad bomber at MTN is all about playing into this stuff, there's a method to that madness but pulling it out late is really blatant abuse of the 'conditional' crowd which shouldn't be allowed. Dahlman I think mentioned it too, bury them early so people jump off. By the time it's official people are kicking themselves the horse paid $4.00 and should've paid $2.80. It's brilliant but you gotta be good and be right about it, not pulling money out with cancellations which I doubt he would ever do.


The only time I really consider odds is when I am at the track. Because I bet on a lot of races using models, it is not really feasible to spend the time watching each race "to see how it all turns out." If I picked the right horse(s), I win, otherwise I don't. Watching the races doesn't change much for me, because I depend on my own models of what the "average" return should be. Some go off at higher, some go off at lower, but when it all shakes out, the projections are accurate enough that I don't pay much attention to the odds in individual races as a bet-no bet decision criteria.

I have been considering using conditional wagers, and that started the whole investigative process that turned up (what I consider to be) counter-intuitive results. The basic process I use to clean outliers always makes higher mutuels a pleasant surprise. I know they will be there, but if the process is not productive without those higher mutuels, I don't use it. The surprising thing to me is that the lower odds ranges seem to have become more productive. Unfortunately--other than the previously mentioned conjecture--I don't have a clue why.

I know that many would say, "Just take the money." Perhaps it is a character defect, but I want to know why things happen--not just that they happened.

[Magister Ludi]

You are observing the favorite-longshot bias (flb). However, that doesn’t answer your question as to why that bias exists. There have been many conjectures as to why the flb exists. I surmise that it’s because of the bets made by recreational (wag) bettors whose bets are based upon random variables such as lucky numbers and horse colors. The distribution of the wag bets is linear and skews the distribution of the remaining (swag) bets. These wag bets cause the distribution of the swag bets to rotate at the point 1/n, where n is the field size. That causes the entries with odds <1/n to be overbet and those with odds >1/n to be underbet.

[traynor]

You are observing the favorite-longshot bias (flb). However, that doesn’t answer your question as to why that bias exists. There have been many conjectures as to why the flb exists. I surmise that it’s because of the bets made by recreational (wag) bettors whose bets are based upon random variables such as lucky numbers and horse colors. The distribution of the wag bets is linear and skews the distribution of the remaining (swag) bets. These wag bets cause the distribution of the swag bets to rotate at the point 1/n, where n is the field size. That causes the entries with odds <1/n to be overbet and those with odds >1/n to be underbet.

I am not quite sure that is what I am seeing. To clarify a bit, assume a cleaned average mutuel of $5 for a given model, with a 50% win rate overall. Because the mutuels are cleaned (to 1.5 times the mean of the interquartile range), "normal" behavior is that the larger (actual) ROI would be higher for those races in which the mutuel is over $5 (because it would include outliers truncated in the modeling process).

However, what I am seeing is not a disproportionate tipping at the average mutuel value--the high end still returns more than the expected ROI. It is that the low end ROI has increased (without a proportional increase or decrease in the number of winners in each category that might explain it). This is a relatively new--it was not happening last year, or the year before.

[raybo]

I am not quite sure that is what I am seeing. To clarify a bit, assume a cleaned average mutuel of $5 for a given model, with a 50% win rate overall. Because the mutuels are cleaned (to 1.5 times the mean of the interquartile range), "normal" behavior is that the larger (actual) ROI would be higher for those races in which the mutuel is over $5 (because it would include outliers truncated in the modeling process).

However, what I am seeing is not a disproportionate tipping at the average mutuel value--the high end still returns more than the expected ROI. It is that the low end ROI has increased (without a proportional increase or decrease in the number of winners in each category that might explain it). This is a relatively new--it was not happening last year, or the year before.

I think that your "truncating" is the problem. You are in effect destroying the individuality of the samples. No 2 samples will be exactly the same, as you know, so those individualities express themselves differently in the overall results.

[SchagFactorToWin]

There is a (relatively) undisputed claim of a direct correlation between win frequency and final odds. What I have found (and continue to find) is that a big chunk of the ROI in any given process comes from the high win percentage/low mutuel end of the spectrum, rather than the (much) lower win percentage/high(er) mutuel end.

I did this with exactas. My 1|23 exactas were profitable over a 3 year period. So I broke them down by odds levels. Over the past year, I have been betting my 1|2 exactas only when they pay LESS than $24 and the 1|3 when they pay MORE than $25. Result: win rate almost the same with amount bet reduced by almost 50%.

[Capper Al]

I am not quite sure that is what I am seeing. To clarify a bit, assume a cleaned average mutuel of $5 for a given model, with a 50% win rate overall. Because the mutuels are cleaned (to 1.5 times the mean of the interquartile range), "normal" behavior is that the larger (actual) ROI would be higher for those races in which the mutuel is over $5 (because it would include outliers truncated in the modeling process).

However, what I am seeing is not a disproportionate tipping at the average mutuel value--the high end still returns more than the expected ROI. It is that the low end ROI has increased (without a proportional increase or decrease in the number of winners in each category that might explain it). This is a relatively new--it was not happening last year, or the year before.

Could the whales be causing this?

[traynor]

Could the whales be causing this?

Strong possibility. It is not (apparently) an anomaly, but something that seems to be happening across the board. As if a lot of money that previously had been dumped into pools is no longer being dumped in, probably because of conditional wagers. It may be useful to take a closer look at the lower odds end of the spectrum for any wagering process. Things may have changed that can be leveraged profitably.

[traynor]

I did this with exactas. My 1|23 exactas were profitable over a 3 year period. So I broke them down by odds levels. Over the past year, I have been betting my 1|2 exactas only when they pay LESS than $24 and the 1|3 when they pay MORE than $25. Result: win rate almost the same with amount bet reduced by almost 50%.

I think that is a process that anyone who bets exotics should continually monitor. However, as you seem to clearly understand, consistency is mandatory--one cannot change their selection process continually and then claim, "I woulda had that one!" Once one's selection process is stable, monitoring exotic patterns becomes extremely useful.

Many bettors routinely reverse exactas for no better reason than that way back in the olden days they lost a reverse exacta that they didn't bet. In most cases (if they keep accurate records) they would find they are throwing away buckets of money on losing propositions.

As you indicate above, there is a different scenario available--if you closely monitor your own results, you could (almost) double your return by wagering the same amount on select exactas as you had wagered before on less select choices.

[thaskalos]

I think that is a process that anyone who bets exotics should continually monitor. However, as you seem to clearly understand, consistency is mandatory--one cannot change their selection process continually and then claim, "I woulda had that one!" Once one's selection process is stable, monitoring exotic patterns becomes extremely useful.

Many bettors routinely reverse exactas for no better reason than that way back in the olden days they lost a reverse exacta that they didn't bet. In most cases (if they keep accurate records) they would find they are throwing away buckets of money on losing propositions.

As you indicate above, there is a different scenario available--if you closely monitor your own results, you could (almost) double your return by wagering the same amount on select exactas as you had wagered before on less select choices.

I have a question, traynor...

When you DO monitor your own results...do you find that these results remain stable over time?

What I mean is...let's assume that, upon surveying several hundred exacta-wagered races, we find that we are throwing away money by reversing our exactas. Does that mean that the next several hundred exacta-wagered races will yield the same results?

How many exacta-wagered races should a person use as a base before the declaration can be made that reversing the exactas is a bad idea?

And even if a sample of several thousand races proves that reversing the exactas is a bad idea...how certain are we that the NEXT several thousand races will also prove the same point?

[traynor]

I have a question, traynor...

When you DO monitor your own results...do you find that these results remain stable over time?

What I mean is...let's assume that, upon surveying several hundred exacta-wagered races, we find that we are throwing away money by reversing our exactas. Does that mean that the next several hundred exacta-wagered races will yield the same results?

How many exacta-wagered races should a person use as a base before the declaration can be made that reversing the exactas is a bad idea?

And even if a sample of several thousand races proves that reversing the exactas is a bad idea...how certain are we that the NEXT several thousand races will also prove the same point?

The bottom line is that there are no cetainties (other than that it is probably more profitable to wager on four-legged horses rather than three-legged horses, and even that could change).

I understand your question, and you are quite right--individual experience is statistically trivial, as you pointed out in the discussion regarding Schlesinger's comments about Anderson's three years of full-time blackjack play.

However, it depends on how one goes about making selections, and what those selections are based on. If one approaches races as dynamic, ever-changing, unknowable processes--then virtually every race analysis or "handicapping" computer application out there is irrelevant trivia crunching nonsense numbers. As is every "system" or "method" one has developed over the years.

Alternatively, when selection processes are based on objective, repeatable patterns rather than individually subjective evaluations of each and every new race, extrapolating the results of one's own betting processes is fairly simple. Ignoring the tote board and "value wagering" helps a lot.

[raybo]

I have a question, traynor...

When you DO monitor your own results...do you find that these results remain stable over time?

What I mean is...let's assume that, upon surveying several hundred exacta-wagered races, we find that we are throwing away money by reversing our exactas. Does that mean that the next several hundred exacta-wagered races will yield the same results?

How many exacta-wagered races should a person use as a base before the declaration can be made that reversing the exactas is a bad idea?

And even if a sample of several thousand races proves that reversing the exactas is a bad idea...how certain are we that the NEXT several thousand races will also prove the same point?

Of course, things continually change in horse racing, we all know that. If your method has no capability of changing along with it then you are stuck with things not working, sometimes for quite a while. If however, your method evolves with the changing racing environment your bad times will usually be shorter lived and cost you less in the long term. If one's exacta play, for example, all of a sudden needs to be reversed, and that need persists more than a couple of weeks or so, then something involving your selection process is not compatible with the changes that have occurred, or are occurring. So, maybe you need to make adjustments in your selection process, or simply stop playing that track, or those tracks, until things return to "normal", if they ever do. Almost inevitably, one's method becomes less efficient and accurate as time goes by and we then have to make adjustments, or lose our edge, if we have one to begin with, of course. If we don't adjust and keep doing what we have been doing, then we have to suffer the consequences, whatever they may be.

I don't mean that we need to make adjustments, just for the sake of making adjustments, but sooner or later, if those changes remain or get worse, then it's probably wise to re-evaluate what we are doing and try to find out what we need to do to get back on track, sometimes that is in our method and sometimes it's within ourselves (as you well know).

[thaskalos]

Of course, things continually change in horse racing, we all know that. If your method has no capability of changing along with it then you are stuck with things not working, sometimes for quite a while. If however, your method evolves with the changing racing environment your bad times will usually be shorter lived and cost you less in the long term. If one's exacta play, for example, all of a sudden needs to be reversed, and that need persists more than a couple of weeks or so, then something involving your selection process is not compatible with the changes that have occurred, or are occurring. So, maybe you need to make adjustments in your selection process, or simply stop playing that track, or those tracks, until things return to "normal", if they ever do. Almost inevitably, one's method becomes less efficient and accurate as time goes by and we then have to make adjustments, or lose our edge, if we have one to begin with, of course. If we don't adjust and keep doing what we have been doing, then we have to suffer the consequences, whatever they may be.

I don't mean that we need to make adjustments, just for the sake of making adjustments, but sooner or later, if those changes remain or get worse, then it's probably wise to re-evaluate what we are doing and try to find out what we need to do to get back on track, sometimes that is in our method and sometimes it's within ourselves (as you well know).

I understand what you are saying...but I wasn't talking about making changes in our handicapping approach in order to deal with the changes presented by a dynamic game. I was talking about the effectiveness of what I call "formula betting"...which is what I call blanket betting advice like "don't box your exactas".

IMO...the right way to approach the exotics is by being flexible in our approach -- and letting the circumstances of the race dictate how they should be bet. Advice like "don't box your exactas"...or "don't bet two horses to win in the same race", don't serve the player very well...IMO.

The RACE should tell us how it should be bet...and there is room on the wagering menu for a great deal of variety.

[raybo]

I understand what you are saying...but I wasn't talking about making changes in our handicapping approach in order to deal with the changes presented by a dynamic game. I was talking about the effectiveness of what I call "formula betting"...which is what I call blanket betting advice like "don't box your exactas".

IMO...the right way to approach the exotics is by being flexible in our approach -- and letting the circumstances of the race dictate how they should be bet. Advice like "don't box your exactas"...or "don't bet two horses to win in the same race", don't serve the player very well...IMO.

The RACE should tell us how it should be bet...and there is room on the wagering menu for a great deal of variety.

I agree, and that's the reason I do my superfecta stuff more "old school", because each race presents different obstacles, which become exaggerated in the exotics ("one size does not fit all"). Exactas have the same problems, albeit of lesser magnitude. Win play has a degree of that also, but less problematic, allowing "formula betting"/play more efficiently and profitably than do the exotics.

[MJC922]

I agree, and that's the reason I do my superfecta stuff more "old school", because each race presents different obstacles, which become exaggerated in the exotics ("one size does not fit all"). Exactas have the same problems, albeit of lesser magnitude. Win play has a degree of that also, but less problematic, allowing "formula betting"/play more efficiently and profitably than do the exotics.

One of the larger studies that I did this year showed profits on the top choice when min odds criteria were applied but a severe drop off from there. The top choice made solid money while the bottom choice lost far more than the take. This makes sense, the money from profitable horses comes from somewhere, it comes from those with the increasingly opposite types of characteristics.

Given this research my current thoughts on the matter are that I find it unlikely horses beyond second rank of any 'profitable' approach are apt to add any real value to the bottom line in the long term. I suspect that we reach a point very quickly with exotics where 'spreading' is just buying insurance and if taken too far it may destroy profitability altogether. I'm not saying some insurance is all bad, but it's generally what it is in real life, you're largely paying for protection against loss.

[Boca Poppy]

Gents,

I've enjoyed reading all of the comments in this thread, in which smart points are stated with proper mathematics to validate your wagering preferances, however recently I have found myself focusing on why the best horse doesn't always win.

The best horse is sometimes the crowd favorite and sometimes its not.

Through your models and research and ROI calculations, whether your selection is 1/2, 3/1 or 10/1. There's also this X factor - Stumbling, bumping, traffic, checking ect that easily causes your horse the race and I feel it's the main deterrent to offset any/all ROI's.

How do you address this in your mathematical approach? Do you assign that a number?

[Overlay]

Gents,

I've enjoyed reading all of the comments in this thread, in which smart points are stated with proper mathematics to validate your wagering preferances, however recently I have found myself focusing on why the best horse doesn't always win.

The best horse is sometimes the crowd favorite and sometimes its not.

Through your models and research and ROI calculations, whether your selection is 1/2, 3/1 or 10/1. There's also this X factor - Stumbling, bumping, traffic, checking ect that easily causes your horse the race and I feel it's the main deterrent to offset any/all ROI's.

How do you address this in your mathematical approach? Do you assign that a number?
The only way that I could think of to address that would be if a horse had a history of a particular running pattern that would have the potential (if repeated) of affecting its own winning chances, or those of one or more other horses in the race. However, such problems might very well be encompassed by the evaluation that I do to assess a horse's overall chance of winning, and would make it an unlikely wager. Other than that, I would view such contingencies as random occurrences that can't be predicted, controlled, or otherwise taken into account, and that would tend to "even out" over time from a perspective of whether their occurrence would have a positive or negative effect on the results of wagers that I had made.

However, I'm also reminded of the late Dick Mitchell leaving 20% of his 100% betting line unallocated among the horses in a field, which he referred to as his "SH" ("stuff happens", as he euphemistically put it) pile, and which required him to get higher odds on the horses in the race before he would bet them, in order to account for such random occurrences.

[thaskalos]

Gents,

I've enjoyed reading all of the comments in this thread, in which smart points are stated with proper mathematics to validate your wagering preferances, however recently I have found myself focusing on why the best horse doesn't always win.

The best horse is sometimes the crowd favorite and sometimes its not.

Through your models and research and ROI calculations, whether your selection is 1/2, 3/1 or 10/1. There's also this X factor - Stumbling, bumping, traffic, checking ect that easily causes your horse the race and I feel it's the main deterrent to offset any/all ROI's.

How do you address this in your mathematical approach? Do you assign that a number?
There are many reasons why the "best horse doesn't always win"...chief of which might be the fact that the "best horse" on paper is not always the best horse on the track. Couple this with the fact that there are often wide areas of disagreement on who this "best horse" really is...and factor in the fact that we are talking about horses whose abilities are separated by scant fractions of a second, who are asked to compete in a chaotic environment where precious fifths of a second can be lost in any number of ways...and one begins to wonder how the "best horse" wins as often as it does.

You address this "chaos" by demanding a certain price for the risk that you undertake.

[traynor]

I understand what you are saying...but I wasn't talking about making changes in our handicapping approach in order to deal with the changes presented by a dynamic game. I was talking about the effectiveness of what I call "formula betting"...which is what I call blanket betting advice like "don't box your exactas".

IMO...the right way to approach the exotics is by being flexible in our approach -- and letting the circumstances of the race dictate how they should be bet. Advice like "don't box your exactas"...or "don't bet two horses to win in the same race", don't serve the player very well...IMO.

The RACE should tell us how it should be bet...and there is room on the wagering menu for a great deal of variety.

Of course there is. I was not referring to that at all, but rather to the carved-in-stone, sacrosanct truisms that some feel they have discovered, and that may be a lot of why the supposed 98% of bettors lose.

[overthehill]

based on my own experience over many years and different tracks I find that difficult to believe unless you are specifically targeting races going against horses that are guaranteed to take late money. as a value player i would say that about 80% of the time when i spot a situation where i think the horse has a 40% chance or better of winning the odds are 2-1 or less and going down from there. I usually look for odds of 6-1 or better on contenders and frequently i am shocked by how much late play they get between the time the horses come on to the track and the time the get loaded in. twice at gulfstream yesterday horses that i saw as contenders went from 6-1 to 7-2 with no time to the race supposedly. and neither of these horses were favored or second choices in the morning line.

[pondman]

Through your models and research and ROI calculations, whether your selection is 1/2, 3/1 or 10/1. There's also this X factor - Stumbling, bumping, traffic, checking ect that easily causes your horse the race and I feel it's the main deterrent to offset any/all ROI's.

How do you address this in your mathematical approach? Do you assign that a number?

There's a Benter video somewhere on this Site. Might be your thing. You might be able to create your own logic to create your own epsilon.

I reason the epsilon for the majority of my play to be just under 60% for most of my play. A horse need to strictly qualify before the game gets my money. It needs to be at what I believe to be at +40%. At that point it could be played at 9/5. But this year I'm going at +8-1. I avoid any singles with less, which is many horses and races.

[pondman]

and that may be a lot of why the supposed 98% of bettors lose.

The vast majority of the 98% really have no clear reason for making a bet. Truthfully, they should not be at the track.

Only a few player go through the steps-- losing, preventing loss, breaking even, making money, to a total Eureka and blowing the game out of the water. And to do this, it's going to take a creative approach and an understanding that the other 98% of the bettors do not have successful methods. So try not to listen to the roar of the crowd.

[raybo]

The vast majority of the 98% really have no clear reason for making a bet. Truthfully, they should not be at the track.

Only a few player go through the steps-- losing, preventing loss, breaking even, making money, to a total Eureka and blowing the game out of the water. And to do this, it's going to take a creative approach and an understanding that the other 98% of the bettors do not have successful methods. So try not to listen to the roar of the crowd.

Well said! Approach is the key, IMO. While patience, discipline, and consistency are very close seconds.