The prevalence of AI technology right now is filtering through to every aspect of commerce and social life. From AI therapists to vibe coding assistants, AI customer service to AI marketing. With the trillions of dollars being spent on the industry by Big Tech companies, it’s only going to become more influential.
Yet, it is interesting to note that AI has an unclear future with sports predictions. That goes for all types of sports predictions, from making fantasy sports predictions in soccer to betting on the NBA Playoffs. It’s not completely the case that it will be omitted from sports predictions, but there are certain caveats that must be understood.
The first thing to discuss is that the current iteration of AI based on large language models has limits in how it understands sports, certainly compared to human observation. An LLM can look at the numbers – sports data, for instance – and run calculations that far outweigh what any human could do. But that only tells half the story.
AI Must Have Strong Data to Help
You need to ask yourself – what is sports data? Let’s say you’re asking AI, will the Los Angeles Lakers win the NBA game tomorrow? What can AI do to inform you? It can look at the form of the Lakers – how many games it has won, how many at home, how many on the road, the average margin of victory, the winning percentage when certain players are starting, and so on. But where does it get that data? The live web? Not really, and it’s certainly not trained on it.
AI must be fed proprietary data to have any real use beyond the basic form guide, and it is not always easy to get that data. Indeed, as a bettor, you will likely have to pay for it, and then that would have to be factored into your calculations of betting profits. Yet, you might argue that, given the right data, AI can make better predictions than humans.
AI Can Help With Spotting Value-Linked Errors
Secondly, there is the idea of spotting value in the odds or fantasy selection. By that, we mean AI’s usefulness in spotting a mismatch between the premise – your bet and the odds – and the chances of a successful outcome. It’s a delicate process, one that assumes a mathematical error on the part of the sportsbook. Those “mistakes” aren’t made too often, and, perhaps ironically, it is also becoming increasingly likely that odds-setting will use AI to create the market in the first place.
Another important area is what we as humans can see and what AI can’t see, and by that, we mean things that cannot be quantified by data. You might see a soccer player scoring a couple of flukey goals, yet looking out of sorts throughout the game. As a human, you question the data, trusting your intuition and your own eyes, but to AI, it will still look like two ticks in the box for goals scored. We could provide a long list of athletes who have had bursts of success despite not performing well. We can put it down to blind luck, and that cannot be modeled.
A danger, perhaps, is that people will begin to look at AI as some sort of prophet, when sports are inherently unpredictable. It can only go so far, and the distance it can go is intrinsically linked to analyzing data and then making predictions based on the data. There are always, always instances in sports where data sets and modeling is left behind, rendered useless. It’s part of the reason sports are intrinsically human. There is plenty of reason to believe AI has a future in ‘helping’ analyze data for sports predictions, but as things stand, it will still need a human’s unique eye for detail to make it useful.
