Showing posts with label value betting. Show all posts
Showing posts with label value betting. Show all posts

  • Beating an Efficient Market

    POSTED Aug 21, 2014 By Derek Simon
    Despite what various racetrack touts and system peddlers say, one of the things that makes consistently beating the races so tough is that, for the most part, pari-mutuel betting markets are efficient. This means that all the relevant and available information affecting the outcome of a horse race is generally known and accounted for. Hence, the final odds are an accurate reflection of a horse’s chances of winning — minus the track take and breakage, of course.

    Sure, there are pockets of inefficiency and irrationalism. Scholars have long documented the existence of a “favorite-longshot bias,” whereby shorter-priced horses are slightly underbet and longer-priced horses slightly overbet. However, such inefficiencies are but ripples on the pari-mutuel ocean.

    Still, the fact that inefficiencies can and do exist provides hope that the races can be beaten — just like grainy, out-of-focus video footage provides hope to some that bigfoot lives among us (often disguised as a broken tree branch).

    In this article, I will attempt to show readers how they can spot and capitalize on inefficient markets — as well as efficient markets — to make more moolah at the racetrack.

    Follow the Money 


    Not to go all “Deep Throat” on everybody, but the simplest way to spot an efficient or inefficient market is to follow the money. Consider the following scenario:

    John Doe is given $2 to bet to win on any horse running at Saratoga on Saturday.

    * How does he choose what race to bet?
    * How does he choose which horse to bet?

    Well, assuming Mr. Doe is logical, one would expect him to play the race and horse that (he believes) give him the best chance of winning. However, even if Doe possessed the superior handicapping acumen of a dart-throwing monkey or one of those omnipotent racetrack touts mentioned earlier, it is clear that any market comprised solely of his wager would have to be inefficient. For, even if we ignored the fact that Doe’s horse would be 1-9, we are stuck with the unfortunate detail that all the other horses in the field — those that didn’t receive any of Doe’s dough — would be lumped together at 99-1.

    Obviously, this is not an accurate assessment of each horse’s chances.

    Thus, even though this was an extreme example, it should be self-evident that less money and fewer wagers equal a less efficient market. Might the opposite also be true? Does more money and more wagers lead to a more efficient market?

    I decided to find out.

    To provide a baseline, I first looked at all sole betting favorites (no favored entries) from a variety of races run across the fruited plain from September to December of 2013:

    Number – 7,996
    Winners – 2,904
    Win Rate – 36.3%
    $2 Net – $1.67
    IV – 2.80
    OBIV – 0.84

    Next, I analyzed favorites in races with the lowest straight (win, place and show) handle on the card (provided the total pool was less than $10,000). As expected, the numbers took a nosedive, giving credence to my hypothesis that less betting/money results in a less efficient market:

    Number – 185
    Winners – 61
    Win Rate – 33.0%
    $2 Net – $1.50
    IV – 2.11
    OBIV – 0.75

    Lastly, I looked at races with the greatest straight handle on the card (provided the total pool exceeded $10,000). Not surprisingly (at least to me), the figures were fantastic:

    Number – 830
    Winners – 320
    Win Rate – 38.6%
    $2 Net – $1.73
    IV – 3.35
    OBIV – 0.89

    In races featuring above-average betting action, favorites won 38.6 percent of the time, lost just 13 cents per dollar wagered (compared to 16 cents for favorites overall) and had an impact value (IV) of 3.35 (versus 2.80 for favorites on the whole).

    The Efficient Data Hypothesis

    Now, I know what some of you are thinking: big deal, Derek, your “fantastic figures” still produced a loss of 13 percent. What good does it do to identify efficient and/or inefficient pari-mutuel markets if one still loses one’s shirt?

    Keep your chin up, Daniel-san. It’s not so much what the stats tell us about these specific instances, it’s what they imply about handicapping in general. Let’s go back to the definition of market efficiency: all the relevant and available information affecting the outcome of a horse race is generally known and accounted for.

    To me, this suggests that “all the relevant and available information affecting the outcome of a horse race” may be overvalued or undervalued in races attracting more or less wagering dollars, respectively. In other words, rather than patterning one’s handicapping around specific race conditions — placing extra value on workouts in two-year-old races, stressing class in turf races, etc. — a player might be better served by using the straight wagering pools to emphasize or de-emphasize traditional factors.

    Take speed figures, for example. Using the database of races above, I compiled the following stats on horses possessing the best last-race Brisnet speed figure over today’s general track surface (AW/dirt or turf):

    Number – 6,353
    Winners – 1,835
    Win Rate – 28.9%
    $2 Net – $1.74

    Nothing to get the pulse racing, right? Well, if you’re standing up, grab a chair (you’ll want to be sitting) and look at what happens when the digits above are parsed based on the size of the win, place and show pools:

    STRAIGHT MUTUEL POOLS GREATER THAN OR EQUAL TO $25K

    Number – 4,299
    Winners – 1,209
    Win Rate – 28.1%
    $2 Net – $1.70

    STRAIGHT MUTUEL POOLS LESS THAN $25K

    Number – 2,054
    Winners – 626
    Win Rate – 30.5%
    $2 Net – $1.83

    In races with less than $25,000 in the win, place and show pools, the horse(s) with the top last-race speed figure produced a loss of just eight cents on the dollar — nearly half the loss produced in races with higher pool totals.

    Get the point? By gauging the relative efficiency of the market one is betting into — be it the first race at Arapahoe Park or the feature race at Del Mar — my research suggests that well-known predictive factors like speed and class can be upgraded or downgraded accordingly.

    And that, my friends, is what good handicapping is all about.

  • The Difference Between Probability & Profitability

    POSTED Jul 19, 2014 By Derek Simon
    If there is one mistake that I see both new and veteran handicappers make time and time again it is confusing probability with profitability — often in very inconsistent and haphazard ways.

    For instance, most bettors know that the post-time favorite wins approximately 1/3 of the time, making it a highly predictive factor. In fact, we can measure just how predictive by employing “impact values,” which were explained by Dr. William Quirin in his masterful work “Winning at the Races.”

    Impact values, or IVs, are calculated by “dividing the percentage of winners with a given characteristic by the percentage of starters with that characteristic,” Dr. Quirin explained.

    “An IV of 1.00 means that horses with a specific characteristic have won no more and no less than their fair share of races,” the good doctor concluded. Similarly, an IV greater than 1.00 denotes that a particular factor is producing more than its fair share of winners, while an IV below 1.00 means that it is producing less than its fair share.

    With that in mind, take a peek at the digits I obtained in an examination of 14,505 races featuring a sole betting favorite (no entries):

    Winners: 5,409
    Winners: 2,017
    Win Rate: 37.3%
    IV: 2.76

    What this means is that the post-time favorite can be expected to win 2.76 times more often than random chance would dictate — which is great.

    However, before we break out the top-shelf pork rinds and don our party hats, let me introduce another metric — one that I came up with several years ago called the odds-based impact value, or OBIV.

    The OBIV is based, not on field size, but on the average odds of the horses meeting the criteria of the study. The advantage of such an approach is that it more accurately assesses the factor being tested (provided the factor is not odds) by using an established and highly predictive methodology instead of random chance to determine the expected win rate.


    Note: The reason the “normal” range is 0.80-0.85 is to account for the various straight takeout rates and breakage points.

    So, harkening back to our study above, we find that post-time favorites produce an OBIV of 0.81 — which helps to explain why, despite a high IV, the ROI on such steeds is negative to the tune of about 16 cents on the dollar.

    The OBIV also explains why merely seeking high-IV, i.e. obvious, factors never makes money in the real world — although many handicapping gurus have advocated just that.

    Tim Maas, author of “Overlay Handicapping,” took it one step further: He used a variety of IV values to produce a fair odds line. Now, before I illustrate the folly of this, I want to credit Maas for at least attempting to use disconnected, or independent, variables in his method (this is another area that gets horse players into trouble — evaluating dependent variables as though they are independent, e.g. speed and form).

    Among the factors that Maas obtained IV value for were Quirin speed points and average earnings per start. To keep this demonstration simple, I will provide my own IVs for specific subsets of these factors — mainly, I will look at horses with at least eight Quirin speed points and horses with the highest average earnings per start in the field:

    * At least 8 Quirin speed points

    Number: 5,068
    Winners: 878
    Win Rate: 17.3%
    IV: 1.31
    OBIV: 0.83

    * Highest earnings per start in the field
    * (If the horse had fewer than five starts this year, the last two racing years were used)

    Number: 13,069
    Winners: 3,331
    Win Rate: 25.5%
    IV: 1.91
    OBIV: 0.83

    By combining these two factors in a makeshift system, we would expect an IV of approximately 2.50 (using Maas’ technique of multiplying the individual Ivs):

    * At least 8 Quirin speed points
    * Highest earnings per start in the field
    * (If the horse had fewer than five starts this year, the last two racing years were used)

    Number: 841
    Winners: 248
    Win Rate: 29.5%
    IV: 2.12
    OBIV: 0.87

    On the positive side, the numbers are vastly improved from those for each individual factor — even the OBIV is nominally better. However, they’re still not good enough to show a profit. In fact, the $2 net return of $1.67 (-16.5% ROI) is less than the $2 net return for post-time favorites ($1.68).

    Ouch. Two highly predictive factors and they produce more red ink than simply watching the tote board and playing the post-time favorite.

    And the situation doesn’t get any better when one asks for minimum odds (as Maas did by insisting on a “fair” price) — in fact it gets worse:

    * At least 8 Quirin speed points
    * Highest earnings per start in the field
    * (If the horse had fewer than five starts this year, the last two racing years were used)
    * Odds of 3-1 or greater.

    Number: 372
    Winners: 56
    Win Rate: 15.1%
    IV: 1.15
    OBIV: 0.87

    Of course, what all this tells us is that, in order to make money as opposed to just cashing tickets at the racetrack, one must look for unique factors and/or use known factors in unique ways.
    There is a difference between what is predictable and what is profitable.

  • Breaking Down the Brisnet Prime Power Rating

    POSTED Jun 19, 2014 By Derek Simon
    It’s no secret that, for many moons now, I have been trying to beat the game with mechanical methods of play. In large part, because I simply don’t have time to handicap like I did 5-6 years ago… nor do I enjoy the process like I used to.

    It is also no secret that I have been largely unsuccessful in this endeavor. 

    Don’t get me wrong: I’ve found some great systems and angles — ones that, through testing, I am confident will continue to offer solid ROIs for months and years to come. But, alas, they all suffer from “Tony Romo Disease” — they don’t show up often enough to produce consistent, meaningful returns.

    Still, I love the idea of a mechanical method that can, at the very least, help guide one’s wagering decisions. So, with that in mind, I decided to slice and dice the Brisnet Prime Power Rating and see if I could find any helpful wagering angles.


    (Click on image to enlarge)

    First, using a database of random races run from July 2012 to February 2013 at racetracks across the country, I examined the horse with the top (highest) Prime Power rating in the field. (In the Brisnet Ultimate Past Performances, the Prime Power rating and field rank is found right before the horse’s lifetime record — see California Chrome above.)

    Number: 11,953
    Winners: 3,972
    Win Rate: 33.2%
    $2 Return:  $1.75
    ROI: -15.09%
    IV: 2.48

    Not bad. These numbers are similar to what one would expect from the post-time favorite, which gives us a good starting point. Now let’s see what happens when the top Prime Power horse goes to post as a morning-line overlay (odds greater than or equal to the morning line) or a morning-line underlay (odds less than the morning line):

    OVERLAYS

    Number: 2,186
    Winners: 428
    Win Rate: 19.6%
    $2 Return:  $1.64
    ROI: -17.81%
    IV: 1.49

    UNDERLAYS

    Number: 9,767
    Winners: 3,544
    Win Rate: 36.3%
    $2 Return:  $1.71
    ROI: -14.48%
    IV: 2.70

    While these numbers may surprise some readers, who assume that higher prices naturally mean higher profits, they are, I have found, very indicative of what happens when one looks for overlay opportunities with “obvious” contenders.

    Caveat emptor applies at the racetrack too.

    A horse that looks like a legitimate 2-1 shot does not necessarily offer “value” at 20-1. Rather than rushing to the window to empty one’s 401(k) on such horses, players should ask themselves a simple question: Why? Why is a horse with so many positive attributes (remember, the Prime Power rating is a comprehensive assessment of ability) being dismissed in the wagering? What is the crowd considering that the Prime Power rating is not?

    Granted, some of these horses are, in fact, genuinely overlooked. I’ve witnessed numerous animals that I thought were fairly-priced at 2-1 or less pay double digits — but that is the exception, not the rule.

    With this in mind, it should come as no surprise that horses with the top Prime Power rating in a race actually perform better when that rating is lower, rather than higher:

     * Top prime power rating of 100 or greater.

    Number: 11,294
    Winners: 3,752
    Win Rate: 33.2%
    $2 Return:  $1.69
    ROI: -15.36%
    IV: 2.49

    * Top prime power rating of less than 100.

    Number: 659
    Winners: 220
    Win Rate: 33.4%
    $2 Return:  $1.79
    ROI: -10.55%
    IV: 2.42

    Moreover, horses with the top Prime Power rating that finished poorly last time out (fifth or worse) return a profit — further proof that factors esteemed by the public, i.e. a good last race, are best avoided when looking for overlays:

    * Top prime power rating of less than 100.
    * Finished fifth or worse in last race.

    Number: 225
    Winners: 74
    Win Rate: 32.9%
    $2 Return:  $2.03
    ROI: +1.69%
    IV: 2.36

    Of course, there is nothing to say that these numbers will continue to hold, but I hope they illustrate both the pros and cons of a mechanical approach.

    On the plus side, using the Brisnet Prime Power rating is quick and easy and, in some instances, profitable. On the negative side, all mechanical approaches suffer from one major drawback: they can only measure what was… not what is or what will be.

    Nonetheless, handicappers willing to put in the time to validate or invalidate a mechanical rating might very well profit from the experience.

    (Click on image to enlarge)
    NOTE: Rue Saint Honore had a race-best 95.6 Prime Power rating and finished 7th in her last race on July 29, 2012.