We are firmly in the era of artificial intelligence. Technology is changing every facet of every industry, although it’s unclear to what extent. As you might imagine, its influence has extended to sports analytics. Again, it’s to varying degrees, and we don’t know how things will look in terms of AI & sports, say, ten years from now, but it’s enough to say that AI is creeping into the system, albeit gradually.
Looking across the web, you’ll see no end of articles with headlines like “Using AI to predict the 2025 season.” It’s interesting to view the pieces as an exercise, but there are many flaws, both in reporting and perhaps in how we perceive the accuracy of the technology.
The article titles should give fans pause for thought, for a start. AI is not a singular entity but framed as such in these pieces. Each AI tool – ChatGPT, Gemini, Llama, Grok – is trained on different data. Moreover, each tool has a subset – there are several different versions of ChatGPT, for example, each based on a unique model with varying degrees of accuracy. So, you aren’t “asking AI” who will do well next season; you are asking a specific model.
AI Mock Drafts Can Offer Insight
That said, there is plenty of merit in parsing out answers from a technology that can instantly scrutinize vast amounts of data. Predicting an entire season – something we will get to a bit later – can be difficult, but AI simulations are perfect in areas of contained, structured data like the NFL Draft. Some brilliant examples of a mock draft using AI are available for free online.
The draft works for AI much in the same way that human-conducted mock drafts do: the AI tool will match the team’s needs with the availability of players in each position. It can analyze millions of data points, all while removing the emotional element that can color our human decision-making. And here’s the kicker: the analytics department at NFL teams is almost certainly using AI now to help them arrive at some decisions.
Yet, the NFL Draft is a short-term, finite event. Predicting an entire season can be messy. 17 games per team – injuries, inexplicable loss of form, locker room dynamics, coaching changes – there are numerous ways that human-made predictions end up looking terrible, and AI is broadly no different.
Season Predictions Are Difficult
Moreover, there is sometimes a misconception about how AI “thinks,” it can be presented as some oracle. AI doesn’t think like humans do – it can’t take leaps of faith. AI models are “next-word predictors.” When you say “dog,” the model can judge that the next word you want to hear is “cat” or “bone” or “man’s best friend.” That’s a crude way of describing it, and it is, of course, a bit more sophisticated than that, but you can imagine that basic concept a million times over so it can come up with “Rams” and “Super Bowl LX” or more fanciful links like “Matthew Stafford” and “MVP.”
But here’s the rub: AI looks at data, and the data is historic. Numerous articles predicting the 2025 NFL season have listed the most likely teams to win: the Eagles, Chiefs, Bills, and Lions. It’s not very imaginative to put the teams with the best records in 2024 as the most likely to succeed in 2025. It rarely pans out that way. Look at what happened to the San Francisco 49ers last season, coming off a Super Bowl appearance. Numerous prediction articles – based on AI or otherwise – predicted the 49ers would do well (they entered the 2024 season as second favorite for the Super Bowl with sportsbooks) – and the season ended up being a disaster.
That, as such, should tell Rams fans not to worry too much or to be too overconfident. You will see a flood of AI prediction articles over the next four to five months as we approach the start of the 2025 season. Some will have interesting insights; some based on flawed data sets. But just like human predictions, you should take them with a pinch of salt.
