Why Value Wagering May Be Misguided

[traynor]

The results below are typical, not selected to make a point. The technique is used in some very sophisticated applications used by some very serious bettors. The results are current for Buffalo, but could as easily apply to (almost) any other track. The results can (and do) also apply (using the same processes) to thoroughbred races.

Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is filter criteria

Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is number of races in which selections went to post at odds of less than 3/1 (56%), the number of those selections that won the race (55%), and the return for flat bets on those selections (1.30 ROI), with no other factors considered.

Of particular note to those rainbow chasers who advocate "value wagering" based on inappropriate (or possibly misinterpreted) data analysis of past events:
Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is number of races in which selections went to post at odds of 3/1 or greater (44%), the number of those selections that won the race (11%), and the return for flat bets on those selections (0.62 ROI).

As mentioned in other threads, this particular technique is used in conjunction with "automated wagering apps." The data--used inappropriately or uncritically, most notably by failing to correct for anomalies and outliers in order to inflate "theoretical" ROI--can lead bettors astray by suggesting profit that only exists in a small sample with a few aberrant mutuels tossed in.

Used appropriately, as suggested on another thread by bettors at Fraser Downs among other tracks, a very healthy ROI can be earned, with lots of action, lots of wins, lots of fun, and very little stress, anxiety, and complaints of onerous track takeout or chicanery by trainers, drivers, or track management.

Again, the example is representative rather than carefully selected. The same criteria can be used to isolate sections of profitable wagers available at other tracks. For those who do not wager sufficient amounts to earn the privilege of cancelling wagers at the last moment unless everything is to their liking, it may not be as convenient at "conditional wagering." It may even involve the inconvenience of actually going to a racetrack or simulcast outlet to wager, rather than clicking buttons on a computer app in the comfort of one's home or office.

However, the rewards are there like low-hanging fruit waiting to be plucked by anyone willing to make the effort.

[traynor]

It is not so much that I have something against value wagering as it is that I think bettors are passing a lot of opportunities for potential profit to chase rainbows they can never quite reach. There is a place for value wagering, and there are circumstances in which value wagering can be very profitable.

The problem arises when value wagering is inappropriately applied to schemes that seem (on paper or computer output) to be profitable, but in reality are not. Some very serious bettors are wagering some very serious sums of money on circumstances in which many bettors (who consider primarily the dismal percentage of winners they select correctly) choose not to wager. I suggest that one might be advised to consider what those very serious bettors are doing, and why.

One might also consider what potential is offered in the exotic pools in races in which the winner is predictable 55-70% of the time. "Swinging for the bleachers" is all well and good as a metaphor, but wagering on horse races is NOT baseball. Chasing "overlays" and "big scores" may blind one to the reality that consistent profit is (relatively) readily available to anyone willing to be a bit less dogmatic and pedantic in his or her race analysis strategies.

[thaskalos]

Would these "select" races retain their profitability in a subsequent test run...or could it be that this profit only materializes in retrospect? I occurs to me that a betting strategy needs to offer REPEATABLE results if it's to be considered effective.

Also...given the size of the pools in these harness races...how does the late-late money, that gets wagered at the last second (or, perhaps even after) affect the usability of this method in "real life"? It seems to me that there is a difference between testing a method after the fact, with the odds already known...and trying to isolate these horses BEFORE the start of the race -- when the odds are still in flux.

I don't want you to misunderstand my intentions here...or to misinterpret the wording of my posting. I am not the eternal skeptic...and I happen to enjoy reading your thoughts. It's just that I've been in this game a long time...and, as a result, certain warning signs pop up in my head sometimes...

[traynor]

Originally Posted by thaskalos

Would these "select" races retain their profitability in a subsequent test run...

Yes. I am as aversed as you are to things that only work in retrospect. I only use models that perform at least as well in the real world as they perform in theory. Most perform better because of the automatic elimination of outliers in thne modeling process that would otherwise distort the results.

or could it be that this profit only materializes in retrospect? I occurs to me that a betting strategy needs to offer REPEATABLE results if it's to be considered effective.

Also...given the size of the pools in these harness races...how does the late-late money, that gets wagered at the last second (or, perhaps even after) affect the usability of this method in "real life"? It seems to me that there is a difference between testing a method after the fact, with the odds already known...and trying to isolate these horses BEFORE the start of the race -- when the odds are still in flux.

That is (and has been) a problem, as well as an area of substantial research. Simply stated, the last-minute odds changes are not a serious problem.

I don't want you to misunderstand my intentions here...or to misinterpret the wording of my posting. I am not the eternal skeptic...and I happen to enjoy reading your thoughts. It's just that I've been in this game a long time...and, as a result, certain warning signs pop up in my head sometimes...

That is not unusual, nor is it unexpected. I have nothing to sell, and I am not trying to convince anyone to adopt my views. I am simply suggesting that certain approaches may be so widely used, and certain other approaches may be so widely avoided, that substantial opportunities for profit exist for those who choose to go against the conventional wisdom.

In specific regard to odds changes at post time, in the particular scenarios presented above, it is more often that the odds go up at the last minute than that they go down (substantially). That may be the result of "wagers not made on otherwise solid choices with odds unacceptably low at 0 minutes to post time." In that area, the use of conditional wagering by online bettors may actually be an advantage.

[traynor]

For those who may not be aware of my use of the word SELECT, it refers to the group of races from which Maiden, NW1, NW2, and NW of X$ Last howevermany races have been culled.

[Ray2000]

A typical example from a different track would be useful.

I say this because Buffalo is a weird track and gets weirder every year.

So far in 2014 the rail is winning at a 25% clip, (.83 roi), favorites at 43% and 4 driver's take almost all the money (what little there is )
Cummings, Beback, Dobson, Wrenn


Finding positive expectation bets at low odds is not impossible but I'd look elsewhere.

[Capper Al]

You're right about other threads. The handicapper gets second choice after the public. In other words, leftovers.

[traynor]

Originally Posted by Ray2000

A typical example from a different track would be useful.

I say this because Buffalo is a weird track and gets weirder every year.

So far in 2014 the rail is winning at a 25% clip, (.83 roi), favorites at 43% and 4 driver's take almost all the money (what little there is )
Cummings, Beback, Dobson, Wrenn


Finding positive expectation bets at low odds is not impossible but I'd look elsewhere.

Current, Maywood:
Jun23 In 260 SELECT Pace races 36 13 22 in 131 HM 51% HW 55% HR 1.26 VHM 49% VHW 13% VHR 0.77 55.40 9

Again, these are "clean" results, with all mutuels truncated to 1.5 times the interquartile mean (of mutuel prices in that specific group of races). The 55.40 is the high mutuel in the set. That mutuel and others like it, uncritically accepted, would make the higher mutuel ranges more appealing. Appealing on paper does not readily translate to cash in the pocket. I prefer a steady run of winners at an acceptable rate of profit to dumping buckets of cash I may never see again chasing rainbows and "big scores."

[traynor]

The Maywood sample is actually a good demonstration of why it is more useful to truncate outliers than to (uncritically) accept them as if they will endlessly repeat. That is, including outliers would mislead the researcher into believing that all odds ranges would be profitable, when the majority of available profit is elsewhere. Specifically, unless one has enough good sense to realize it is broken, there is little motivation to fix it.

Those using HTR and similar apps can readily run similar queries. Find a profitable set of filters--using all odds ranges--then re-run those filters with odds ranges set to lower ranges. Sometimes you may find the "profit" was in one or two outliers in a clump of races. However, the reverse happens frequently as well--eliminating the high odds ranges may toss a large mutuel or two, but it often more than compensates by eliminating losers.

The Pareto Principle (80% of this is in 20% of that) definitely applies to horse races. The trick is to locate the 20% of the races that contain 80% of profit.

[traynor]

I think the bottom line is that a bit more thought and a bit less pedantry might be useful to bettors. "Value wagering" is a commonly accepted principle that in many cases works in reverse by ignoring many (lower priced payoffs) in favor of a few (higher priced payoffs). When the return on the lower priced payoffs easily exceeds the return on the (much less frequent and possibly missing entirely) higher priced payoffs, that may not be a very good wagering strategy.

One should be aware that the pursuit of higher mutuel prices is fueled as much by the ego as it is by the desire to win. The less accurate the selection method, the higher the mutuel prices required to support it in theory. The less accurate the selection method, the more likely it is to be based on a few (unlikely to repeat) wagering anomalies, and the less likely it is to produce profit over time.

For bettors, believing one's self to be a better race analyst than one actually is can be detrimental to the bank account.

[mrroyboy]

I still don't understand exactly what you mean Tray. What betting patterns etc are you talking about?

As far as Overlay/Value theory it is correct in theory but way harder in practice. And you are right many people have inflated egos about their ability.

[traynor]

Originally Posted by mrroyboy

I still don't understand exactly what you mean Tray. What betting patterns etc are you talking about?

As far as Overlay/Value theory it is correct in theory but way harder in practice. And you are right many people have inflated egos about their ability.

If something happens 55-70 times out of a hundred, while another may (or may not) happen once or twice in a hundred, which is more likely to provide a dependable return? That is not just a rhetorical question. Many seem to believe the two scenarios would be equivalent if the ROI for a small sample of races is identical.

Extrapolating to a larger sample, 10 times the size of the original sample, the first scenario is likely to repeat in distribution (and it will be obvious if it does not) while the second scenario may have 20 wins, 10 wins, or no wins, because the infrequency relative to the model size is usually indicative of anomalies and distortions caused by outliers.

A "model" based on events in a sample is dependable as a predictor in direct proportion to the ratio of events to sample. The 2, 3, 5, 10 or whatever number of events may be a cluster of unevenly distributed anomalies in a sample that "seems" to form a pattern. In the example above, the frequency of wins is a MUCH stronger indicator of a predictive model than the ROI.

In the previous examples from Buffalo and Maywood, the major portion of the profit available is in the mutuel range under 3/1. Many may not realize that, look at a particular strategy as "profitable," and decide to only make conditional wagers at 3/1 or better. I think they would find it much easier to make a profit (as well as winning much more frequently) by concentrating on the mutuel range below 3/1. If it works for the whales, it should work as well (or better, because they have a wider selection of tracks) for the non-whale bettors.

[DeltaLover]

Originally Posted by traynor

Many may not realize that, look at a particular strategy as "profitable," and decide to only make conditional wagers at 3/1 or better.

I think that a conditional wager at any odds is by no means a guarantee of a final payout. In today's game quite the opposite is truth, a 3-1 just a few seconds prior to the bell, in most of the cases will translated into something like 9-5 or less if the horse wins.

[traynor]

Originally Posted by DeltaLover

I think that a conditional wager at any odds is by no means a guarantee of a final payout. In today's game quite the opposite is truth, a 3-1 just a few seconds prior to the bell, in most of the cases will translated into something like 9-5 or less if the horse wins.

I could not agree more, with one exception. There are specific tracks, and specific types of wagering patterns, in which it is either less noticeable (final drop) or even reverses. One example is Yonkers--9/5 would be a generous payoff for an entry 3/1 at post time. In contrast, at Fraser Downs (and a number of other tracks) it is often that an entry "held" at close to even money for most of the wagering period goes up in odds after the wagering is closed. A payoff of 8/5, 9/5, or even 2/1 are typical for entries that have rarely been on the board at more than 6/5.

My best (and possibly completely erroneous) guess is that the depressed (or suppressed) odds discourage those who would otherwise bet on that entry--and the conditional wagers do not enter the pool to dilute it, or may even go on an alternate choice.

[imofe]

My results at some tracks from last year and this year so far indicate that my profit is from 5/2 to 5/1 range on horses I believe should be lower. Horses that I believe should be around 7/1 or higher and go off as handicapped overlays have been losing propositions overall for me. Very interesting stuff Traynor.

Originally Posted by traynor

The results below are typical, not selected to make a point. The technique is used in some very sophisticated applications used by some very serious bettors. The results are current for Buffalo, but could as easily apply to (almost) any other track. The results can (and do) also apply (using the same processes) to thoroughbred races.

Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is filter criteria

Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is number of races in which selections went to post at odds of less than 3/1 (56%), the number of those selections that won the race (55%), and the return for flat bets on those selections (1.30 ROI), with no other factors considered.

Of particular note to those rainbow chasers who advocate "value wagering" based on inappropriate (or possibly misinterpreted) data analysis of past events:
Jun21 In 388 SELECT Pace races 14 12 35 36 in 138 HM 56% HW 55% HR 1.30 VHM 44% VHW 11% VHR 0.62 22.00 8
Bolded section is number of races in which selections went to post at odds of 3/1 or greater (44%), the number of those selections that won the race (11%), and the return for flat bets on those selections (0.62 ROI).

As mentioned in other threads, this particular technique is used in conjunction with "automated wagering apps." The data--used inappropriately or uncritically, most notably by failing to correct for anomalies and outliers in order to inflate "theoretical" ROI--can lead bettors astray by suggesting profit that only exists in a small sample with a few aberrant mutuels tossed in.

Used appropriately, as suggested on another thread by bettors at Fraser Downs among other tracks, a very healthy ROI can be earned, with lots of action, lots of wins, lots of fun, and very little stress, anxiety, and complaints of onerous track takeout or chicanery by trainers, drivers, or track management.

Again, the example is representative rather than carefully selected. The same criteria can be used to isolate sections of profitable wagers available at other tracks. For those who do not wager sufficient amounts to earn the privilege of cancelling wagers at the last moment unless everything is to their liking, it may not be as convenient at "conditional wagering." It may even involve the inconvenience of actually going to a racetrack or simulcast outlet to wager, rather than clicking buttons on a computer app in the comfort of one's home or office.

However, the rewards are there like low-hanging fruit waiting to be plucked by anyone willing to make the effort.