Sports betting has changed a lot in the internet era. While the same goals will drive the industry, including the emotional draw of backing your favorite team or player to succeed, the integration of statistics into sports betting has changed the strategy for many experienced bettors, many of whom will look to wager on advanced team and player props rather than the outcome of a game.
In the simplest sense, it means if you aren’t using statistics to bet, you are putting yourself at a disadvantage.
If the above holds water, you can imagine, then, the drive to use artificial intelligence to help with sports betting.
Indeed, the web is littered with specialist AI tools and services offering ‘winning’ sports betting strategies, though we would be highly skeptical about anything that claims to guarantee success.
Nevertheless, consumer-facing AI is here, and it can, under certain conditions, be a useful aid for sports betting.
AI has limitations in sports predictions
The first thing to bear in mind is that AI is only going to be as adept as the rest of us when it comes to broad sports predictions.
If you fed the latest NBA Playoff odds into an AI model and asked it to predict the championship winner, it wouldn’t likely deviate from the human answer, probably landing on the Oklahoma City Thunder as the red-hot favorite.
Yet, it’s not about broad predictions; it’s arguably about spotting value in the odds. To explain, consider what you are doing when picking a wager: A sportsbook says the odds of X happening are Y, and it is your job to weigh up whether X will happen and whether Y offers good value.
As humans, we are not really adept at gauging the latter beyond our own intuition, but AI can help run the numbers, weighing up the probabilities based on data and comparing that to the odds on offer. That’s the value prospect.
The AI will need quality data from professional sources
The question, however, is where to get the data to ‘feed’ the AI model. If you are using consumer-facing models like Gemini or ChatGPT, there are limits to what they can retrieve from the live web.
Sure, it can find rudimentary stats like form guides and player statistics, but the kind of high-value proprietary data that is produced by companies like Opta, Sportradar and Genius Sports is guarded jealously by those firms.
And why not? They make a living by selling the data, including to sportsbooks, so they aren’t going to let themselves be usurped by OpenAI et al.
You can, of course, get that data yourself, though it is something that must be factored into your profit margins. A subscription to one of these platforms might yield information that will be useful to your AI for spotting value in the market.
Arguably, it must be targeted. You should maybe start with a specific market, like over/under corners in soccer games or PPG with a specific collection of NBA players, then start fueling AI with the specific data on that market.
Then it’s time to give the AI odds from a variety of sportsbooks, helping you decide where the sportsbooks have been overgenerous in setting prices.
As you can see, none of this is a panacea. All of those “AI predicts with 100% accuracy” claims are nonsense. If you know the history of sports, you know that they are inherently impossible to predict with 100% accuracy. But AI with useful data can give you a better edge.
In a sense, it’s a bit like card counting in blackjack: You don’t win every hand, but the system can turn the odds a little bit in favor, potentially overturning the fine margin between success and failure.
