Chiefs vs Eagles: can AI predict Super Bowl winner?

Can AI score you a winning bet, or should sports predictions be done the old-fashioned way?

The Super Bowl LIX 2025 starts this week, and naturally, people are desperate to place their bets on the winning team.

But how do you know who’s going to win? Sports fans have tried various crazy tactics to predict who will win sporting events. Punters even went as far as consulting an octopus about which team would win the 2010 FIFA World Cup.

On this day in 2010: Paul the Octopus, famous for his accurate predictions of Germany's matches, sadly died. pic.twitter.com/mKN4wzMCue

undefined ESPN FC (@ESPNFC) October 26, 2015

While Paul the Octopus was 85% successful when predicting the outcome of games, we have now abandoned ‘aquatic game predicting wildlife’ for more advanced methods.

These days, we’re letting artificial intelligence models and AI enabled analytics influence our decisions.

So, with the upcoming Super Bowl event hitting our screens on Sunday, February 9th, I wanted to know whether an artificial intelligence model could predict the winning team.

Disclaimer: This article is for informational purposes only. Cybernews does not endorse, promote, or encourage any form of gambling, nor do we advise or encourage relying on outputs or predictions from artificial intelligence (AI) models when engaging in it. Readers assume full responsibility for their actions, and Cybernews shall not be liable for any losses incurred.

Chiefs favored to win

The Kansas City Chiefs and the Philadelphia Eagles will play against each other on Sunday, and I want to know where to place my bet.

So, I asked ChatGPT, “Who will win the 2025 Super Bowl?” and the chatbot responded with context from Reuters, information from Wikipedia, and information from Fox Sports.

Upon first inspection, ChatGPT provided a lot of context from various sources. However, it didn’t provide a strong enough answer to make me want to bet on a team.

So, I asked again. This time, I asked it to reason with the question and come up with an answer by evaluating potential outcomes.

While the chatbot said that it doesn’t have any personal opinions, it can collate information and make inferences based on what analysts have discussed online.

ChatGPT concluded that the Kansas City Chiefs are slightly favored to win the Super Bowl. But, this is not based on divine intervention or psychic abilities. This is purely based on data.

The AI model cites experts like Christian Gonzales, Maurice Jones-Drew, and Dante Koplowitz-Fleming, who are all affiliated with the NFL or have played American football professionally.

The chatbot cites experts who have thrown numbers out there, but all claim that the Chiefs will win over the Eagles. According to the chatbot, Gonzales predicts a 34–27 victory for the Chiefs, crediting lead quarter-back Patrick Mahomes’ connection with tight end Travis Kelce.

Jones-Drew has allegedly predicted a high-scoring game, with the Chiefs winning 45–42 against the Eagles, and Koplowitz-Fleming predicts a 33–30 win for the Chiefs that will go into overtime.

Furthermore, ChatGPT was able to evaluate team dynamics, the range of talent, dynamic strategies, prior performances, and even the team’s culture in a matter of seconds.

Over time, more and more people are integrating AI into their bets, but should we trust it? I consulted experts in the field that told me whether we can really rely on chatbots to predict sporting events.

Should we trust chatbots?

There are a lot of variables when it comes to sporting events. For example, the weather could be off, there could be unintended substitutions or injuries, along with a plethora of other outcomes that AI is unlikely to predict.

Well, that’s what senior analyst at CasinoReports Jeff Edelstein told Cybernews.

Edelstein told us that he has been feeding ChatGPT, Claude, and Gemini with expert opinions that include a bunch of statistics.

For example, the opinion that certain players have a predicted completion rate when pressured, or that the Chiefs blitz 8.3% of the time and so on.

“As the week has worn on, I’m giving the chatbots hundreds of data points to play with. Interestingly, they are all landing near the same ‘most plausible’ scenario, which is a 27-24 game,” Edelstein told Cybernews.

This signals that, over time and with the right data, AI models might be able to synthesize information and come to conclusions based on hundreds of data points. However, this still takes time and human intervention.

While Edelstein suggests that at this point we shouldn’t solely rely on AI for our betting tips, it can be a great tool to process information that humans would struggle to make sense of.

But there’s a few reasons why we shouldn’t put all our faith in AI predictions.

Don’t bet everything on AI

Now, we must take everything ChatGPT says with a pinch of salt.This is because chatbots' answers are predictions themselves, as they’re working off of data they are trained on and will generate the most likely answer, not the most probable, Sandi Besen applied artificial intelligence researcher at the technology company IBM told Cybernews.

Furthermore, certain chatbots or AI models might not be up to the job.

“Chatbots are not the right AI model to use for predicting anything other than text,” Besen told Cybernews, “since predicting the output of a game is a numerical based prediction (aka the odds) that is not the right type of predicting method for a chatbot.”

While it might feel like chatbots know everything, it's important to remember that they are trained on a specific data set and can only generate the most likely outcomes based on the information they have available.

So experts like Bensen believe that chatbots based on large language models should act “as your research buddy” that can inform your decisions and “not predict the outcome for you.”