[imagesource:flickr]

Don’t tell us you didn’t think of horse racing and the Lotto when everyone began salivating over the seemingly endless possibilities of AI.

It’s a consideration that has its logical appeal. If you input every detail about a race, from the history of the horses, to track conditions and weather on the day, would AI be able to tell you who would win? That’s a R53,5 billion question.

As it stands, one of the most common applications of AI in horse racing is the analysis of racing statistics.

Researchers and trainers can examine enormous quantities of data using machine learning algorithms and other AI approaches to detect patterns and make predictions about which horses are most likely to do well in a specific race. Some trainers use AI-powered sensors to track a horse’s performance throughout a race, adjusting their strategies on the fly based on the data.

The use of AI by trainers and racing managers is a given, seeing as how specific the feedback can be. But what about using AI to predict a race, and then betting on it? A recent article in Financial Times gave this some serious thought, and lucky for us, a senior data journalist at the FT even tried it. And he is not alone, as a large number of people take the AI approach when deciding which horse is going to make them rich.

These betting nerds are ‘collectively known as computer-assisted wagerers, or CAWs, and they are largely anonymous’.

Can you imagine adding a large pile of data about 1000 horse races for AI to analyze them? The end result will be a prediction for an upcoming race that can be very accurate, which sounds scary for the betting industry.

It’s been tried before, and in 2018 a start-up correctly predicted the Kentucky Derby exacta pick. Exacta bets also referred to as perfecta or exactor bets, involve choosing two horses for the top two positions. It is a way of making big bucks from little bets. Their AI helped them to collect R1 200 on a R60 bet.

The system produces a forecast called AI Fair Odds™ that includes odds for first, second, third, fourth, place, show and top exacta and trifecta picks. The win odds are calibrated to within a 0.5 percent accuracy.

According to the Paulick Report, a 20-horse race can have a result of more than 2,4 x 10¹⁸combinations, and unless you have a billion years and a lot of pencils, AI can compute these odds in seconds, giving you a fairly good idea of how things will play out. With the evolution of AI, we might even see horses fitted with real-time sensors, giving us immediate info on their performance.

The Financial Times journalist doesn’t seem to be sold on the idea. After winning a few races, they ended up losing most of their money in the second set before reaching for the old-fashioned dailies.

AI might be able to give you a good idea of where to bet, but as with most things AI, it can never accurately account for the human, and animal, elements. If the jockey is on fire, and the horse finds a little extra gallop in its stride, the carefully calculated odds can go out the window.

AI is after all just a machine.

[sources:paulickreport&fintimes&technology.org]