Winning AI Sports Predictions? We Simply Aren’t There Yet

Fri, Jun 6, 2025
by CapperTek

As we trundle towards the conclusion of memorable seasons in the NBA and the NHL and welcome in a long summer of baseball before the NFL gets underway in September, plenty of sporting narratives have gone wayward.

Consider the following: a year ago, everyone talked about the Boston Celtics as a potential dominant force for years to come. That narrative has quickly flipped. Also, in the NBA twelve months ago, the 76ers were predicted to have a stellar 2024/25 season. In the NFL, the 49ers were beaten in the Super Bowl, and many thought they would go one better in the 2024/25 season. Manchester City was tipped to win the UEFA Champions League. For various reasons, All of those teams failed to live up to the consensus built up around them.

AI has largely followed betting odds.

Yet, you can ask – how predictable was all this? And could artificial intelligence have predicted it? The answers are “not very” and “probably not.” The introduction of AI as a sports predicting tool – one that could surpass humans – is an interesting concept with a lot of merit, but there are a lot of wrinkles to iron out, as well as some misconceptions about how AI works.

First, let’s consider some of the predictions made by AI, focusing on basketball. Before the 2024/25 season, numerous articles were published with the tagline “using AI to predict the NBA season,” with most focusing on the outcome of the NBA Championship. Some predicted an OKC vs. Boston Celtics showdown in the 2025 Finals. Well, the prediction was half-right.

And yet, this is hardly going out on a limb. The Celtics and OKC were the favorites in the 2025 NBA Playoff betting right through the season, at least until the former’s exit. How many ‘human’ fans – even casual fans – would have picked the reigning champions and a team that every pundit was saying would dominate the Western Conference to go all the way with some good odds to win the Finals?

AI has a long way to go.

The point, as such, is that we are a long way from AI telling us that the Cleveland Cavaliers would surprise everyone by topping the Eastern Conference standings at the end of the regular season or that, despite the Celtics and Cavs having the best record in the East, it would be the Pistons and Knicks fighting it out for a shot at OKC in the Finals.

Today, numerous AI tools, often touted on social media, purportedly offer sports predictions. Yet, it is debatable whether they have the data to make more of a fist of it than human tipsters. AI can analyze data and needs good data; it is not clairvoyant. Getting that good data is laborious and costly, and you can be sure that standard models offered by ChatGPT, Gemini, etc., do not have it.

On the other hand, we aren’t far from being in a scenario where AI can analyze in real-time, reacting to every aspect of a season, changing its predictions, and informing bettors as a season unfolds or giving live betting tips in-game. But we are certainly not there yet. You should be skeptical or combine AI analysis with human experience. If and when the AI revolution comes to sports betting, it will be much more explosive than the endless, unimaginative “supercomputer” predictions we see churned out in the sports media today.