I Battled ChatGPT in College Football Picks—The Unexpected Outcome Will Shock You

Ever wonder if artificial intelligence could outsmart us human sports bettors, or if it’s just another flashy gimmick? Well, I’ve been slugging it out against the Vegas spreads all season, and frankly, the machines were starting to sound like a promising scapegoat. So, I threw down the gauntlet and handed over 35 college football games to ChatGPT—asking it to crunch numbers, factor in injuries, and spit out picks against the spread. After a few minutes of digital pondering, it confidently rolled out its choices… but did the ‘intelligent’ bot actually beat my gut feelings and years of experience? Spoiler alert: the human vibe edged out the machine this time, proving that sometimes, betting on instinct beats betting on algorithms. Curious how it all shook out? LEARN MORE.

Picking winners against the spread in college football is not easy. As the saying goes, they don’t build those buildings in Las Vegas because the house consistently loses. Gambling is hard, and as I’ve been getting beat up a bit this season, I decided to let ChatGPT make its own picks against the spread to see if artificial intelligence was any better at this than I was.

Here is a breakdown of what exactly I told ChatGPT to do, and how many games I went head-to-head against the machine.

In total, I selected 35 games for it to pick. I kept the prompt with ChatGPT straightforward, telling it to make picks against the spread, which I provided, and to take into consideration key injuries. I was also sure to make note of which team was the home team.

After about three minutes of ‘thinking,’ which was ChatGPT scouring the Internet for injury updates and statistics, the picks were delivered in a simple table format. Unprompted to do so, ChatGPT also ranked each game with a confidence level between 1 and 5, with 5 being the most confident, while also giving the number against the spread it suggested to play to.

ChatGPT didn’t rank a single game with a confidence level of 5, which was smart on its part because you can never be completely confident when it comes to gambling, but it did circle four games with a confidence level of 4:

Michigan -11.5 vs Northwestern (neutral site)

Iowa +6.5 @ USC

Texas +6.5 @ Georgia

Virginia Tech +13.5 @ Florida State

ChatGPT went 1-3 with those picks, with Iowa +6.5 being the lone winner, and AI’s bad luck with its most confident picks was a preview of what was to come.

All in all, ChatGPT went just 14-for-35 with its picks against the spread on Saturday for a success rate of 40%. Heading into this little experiment, I predicted the AI picks to be right around 50%, but even the ‘intelligent’ machine didn’t even come close to breaking even on Saturday.

As for how I did with my own picks against the spread, it wasn’t a great week by any means, but I did manage to break .500 and pick 19 winners. A success rate of 54% isn’t going to bankrupt the sportsbooks, but getting the best of ChatGPT was a nice W for the weekend.

The five games ChatGPT and I didn’t see eye-to-eye on the most were:

Air Force +7.5 @ UConn

North Texas -18.5 @ UAB

NC State +14.5 @ Miami

UCF +23.5 @ Texas Tech

Mississippi State +7.5 @ Missouri

(The bolded team was ChatGPT’s pick)

Unfortunately, I was completely on board with the four games that ChatGPT marked with a confidence level of 4. If I had faded the computer with those four games, we would have been looking at a very nice Saturday.

The takeaway here is that ChatGPT isn’t anywhere close to predicting games against the spread, at least not yet, which is a phenomenal conclusion for all of us sports betters out there who love to throw a few bucks on games based on nothing but vibes and gut feelings.

The human vibe outdid the computer vibe this week.