Designing a computer program to “beat the races” is probably still wishful thinking for most casual fans, though the tools that are needed are much easier to come by in the A.I. revolution.
Whereas before it was a massive undertaking to gather all the data and finance expert software programmers to build decision models, that sort of project can now be achieved super-fast using A.I, and at virtually no cost.
There are many factors to consider when trying to predict race outcomes and once done, a largely efficient market, and cunning rival bettors still make finding positive expectation plays/” overlays” very challenging.
Making sharp selections is a vital foundation but not absolutely the key to the mint.
Racing also has an inherent unpredictability because we are dealing with flesh and blood creatures, and people. That leads to tremendous variation and “randomness”. However, most performances can still be accounted for much of the time, and sound predictions made.
Cleverly compiled selection models which create rankings, probability estimates and comparisons with the actual betting odds are increasingly high-tech and sophisticated.
It can be fun these days, even for recreational players, to tinker with A.I. programs and come up with their own “systems” to help make horse racing predictions.
Just appreciate though – even if you build a good model, the real-world problem is that low percentage things still happen, which are, by definition, extremely hard to foresee. A 3% slim possibility is not zero chance, so just because something has a measly percentage of occurring, does not mean it cannot happen at all.
Some lateral thinking is also required. Many frustrated Jackpot and P6 players should realize that, added together, four outsiders at 15/1 (6% chance of winning x 4 = 24%) have about the same probability as their rock solid 3/1 banker.
When putting together a model, lots of complex minutia can be included. Some of the early computer assisted wagering teams in Hong Kong would consider several hundred data points, even attempting to quantify subtle effects of precipitation and humidity on horse’s individual ratings.
They would harness computer power to crunch the numbers, come up with prediction rankings and odds estimates, before batching and submitting huge bets once they knew how the actual market was forming.
But there is also an argument to keep things simple and home in just a few fundamental influences on race results.
How about focusing on four principal factors: the ability of the racehorse, its form/physical condition, the circumstances of the upcoming race (distance, draw, going, pace) and the skills of the trainer/jockey.
That may be an over-simplification, yet if you have accurate data in these four categories you could probably create a surprisingly powerful selection system.
If you have the time and inclination, rev up your preferred A.I. ally of choice and give it a try…