Statistical Edge: Chart Your Biases
Fans of Brad Pitt are familiar with the 2011 film Moneyball, where Billy Beane, general manager of baseball’s Oakland Athletics, uses enhanced statistical analysis to predict the performance of prospects. Theo Epstein is a big advocate of “Moneyball,” too, and he ended epic World Series droughts for both the Boston Red Sox and Chicago Cubs.
Applying a “Moneyball” approach to harness handicapping can be used as an enhancement, but not a substitute, for traditional tools such as past performances and replays. I personally use an Excel spreadsheet to track this.
There are few tasks that require as much objectivity as handicapping. But everyone has built-in biases. The “Moneyball” approach will point out your handicapping biases – both positive and negative – thus improving your performance over time.
The measurement tool you should use is the index. Index is the same as return on investment (ROI), except the number 1 is added. An index of 1.00 is break-even, which I call the “Mendoza Line” (another baseball term). The analysis is based on win price, and most people bet exotics. However, the “Moneyball” approach is applicable to exotic bets, too, as exotics invariably involve picking a winner.
The key variable is the $2 win price, since the index is based on it. You can use as many predictive variables as you see fit. Some are numeric, some are names, and some are “Boolean,” where yes=1 and no=0. Over the past eight years, I have tracked about 15 different variables. Sometimes, I realized that a variable made no difference, so I stopped tracking it.
Some variables you may want to track are Racetrack, Gait, Post, Driver, Positive Driver Change (1=yes, 0=no), etc. After the race, enter the $2 win price, or $0. The columns are the variables. The first row is the header, and the races (observations) start in the third row.
Due to the takeout, the index for all win bets combined is about .82 — well below the break-even point. This is why your Aunt Millie always warned that “you can’t beat the track.” However, the .82 average includes not just handicappers’ money, but also bets from people who bet numbers and horse names. Even a mediocre handicapper should average about .94. If you are a skilled bettor, you should be able to reach the “Mendoza Line.”
Proper application of “Moneyball” techniques should help to increase your index a few points over time. Note that total profit is a function of index, bet size, number of bets, and betting service rewards.
In addition, if you are able to obtain an average of 1.00 on win bets, you should fare better on exotics. This is because exotics involve multiple decisions spread out over just one pari-mutuel takeout.
The collection of hundreds or 1,000-1,500 observations (races) is a good first step, but here comes the difficult part – interpreting the data. The basic premise of “Moneyball” is that if a particular variable, after a sufficient number of observations, yields an index well below 1.00, then you should reduce or eliminate such bets. Conversely, if it yields an index well above 1.00, you should make such bets more often. There are some cautions:
Sample size: What is a sufficient number of observations? Two or three is far too small of a sample and 250 should be more than enough. This is largely a judgment call. You should analyze data on races only within the past 18 months.
Outliers: Say you hit a $50 horse. That may skew your analyses for months. For analysis purposes only, enter a maximum win price of $25-$30.
Cause vs. correlation: This is the trickiest part of analyzing data. Over time, you will spot numerous trends. However, unless you can pinpoint a reason for that trend, it may be merely a statistical aberration. In such instances, you should not alter your betting strategy. I noticed that post 2 on half-mile tracks yielded a poor index. I initially considered it merely a fluke. Then I realized that those horses run a real risk of ending up with the undesirable first-over trip. I no longer make such bets, unless the post 1 horse is slow out of the gate and the post 2 horse motors out of the gate.
After eight years of “Moneyball,” I learned quite a bit, and made numerous adjustments, leading to better results over time. Here are just a few: first-time Lasix horses were a phenomenal bet until sometime in 2010. Since then, the index for this variable has hovered about 1.00, and I eventually stopped tracking it. Positive driver changes and horses off good qualifiers are good bets. Posts 7 and 8 on half-mile tracks are horrible bets, which we covered in the January issue of Hoof Beats.
http://xwebapp.ustrotting.com/absolutenm/articlefiles/73626-moneyball_sample.xls
by Bob Gardner
