Racing and Data Science

There’s a well-known Bloomberg piece The Gambler Who Cracked the Horse-Racing Code that has attracted renewed interest recently. The world of algorithmic betting has come a long way since then - data is more widely available and advances in machine learning have made many traditional quantitative models redundant.

Access to detailed data is often the difference between making a profit and not. Accurate, clean data is expensive and valuable. Without it, you are unlikely to stumble upon a system that works, because all such systems have been (or soon will be) made redundant by data-driven machine learning techniques.

If you’re serious about making money, don’t bet with a bookmaker than has a financial interest in you losing (Bet365, Paddy Power, William Hill, Ladbrokes, and most others). As soon as you start winning, or placing bets indicating you are taking things seriously, they will ban you permanently. If you want to make a profit in the long term, use a peer-to-peer exchange like Betfair. This way, the more you win, the more money the company makes, and so they have a strong incentive to keep you as a customer.

Teams like Bentor’s still exist, but run more like well-organised quant funds. These days they are generally composed of data scientists and machine learning experts, often with past experience in the financial domain. Making a profit is still possible, but you will have to work hard.