Can AI Really Predict The Outcome Of Sports?

With artificial intelligence (AI) being the latest buzzword, it seemed like only a matter of time before the tech would be turned to sports. Or, to be more specific, sports betting. AI is most useful in crunching large amounts of data, then giving results based on that data. So wouldn’t this make AI perfect for predicting the outcomes of sports matches?

For example, couldn’t someone input all the data for the NCAA, then have an outcome predicted on who will win well in advance? It’s an interesting prospect, especially for a sports betting enthusiast. In fact, it almost seems too good to be true.  

Close But No Cigar

The problem with sports franchises, such as the NCAA, is that there are an incredible number of variables to take into account. Assumptions might be that it would simply be a matter of inputting critical information regarding the team, after which the AI could easily predict which the better team is. It is far more complicated than that.

An NCAA favourite team to win might be suffering for low morale, or a star player might not be at their best. These factors can change at a moment’s notice, meaning that an AI simply can’t take into account all of the important factors. Heck, underdog upsets are well known to occur, and often for reasons that are very unclear. How can an AI predict that?

The truth is that while AI can often predict reasonably accurate results, it still simply doesn’t have enough information to work with. The upside is that betting enthusiasts still have a lot to work with, including NCAA picks from industry experts. 

The Technology Is Getting Better

But that doesn’t mean that AI isn’t getting better, and won’t soon be able to make increasingly better predictions. It may not be that long at all before the ultimate NCAA predicting AI is created.

In a recent experiment students used a machine learning (ML) model to predict a match while it was underway. During the match the team continually updated available information, allowing the system to learn, and thereby update its own prediction. Not surprisingly, the predicted outcome was rapidly readjusted, and by the time the match was over the AI had made a fairly good guess.

Fairly good may not be what most sports betting enthusiast are looking for, but AI tends to get better with time. It isn’t unfeasible to think that with enough time, and after being fed enough live matches, that the predictions could get extremely accurate.

It Won’t Be Perfect

So will there be a perfect NCAA predicting AI in the future? As with all things related to AI technology, almost certainly not. Where predictions are concerned, no matter how good an automated system gets it will never be perfect.

Although, the very real question needs to be asked; would anyone ever want predictions to be perfect? Wouldn’t that rob sports of a great deal of their entertainment value, if the results were always known in advance?