I’m a pretty big sports fan – particularly around the NBA and MLB. With that said, I’ve never been into college sports. I don’t know what’s wrong with college sports exactly, but I just couldn’t care less. That said, every March I still fill out an NCAA tournament bracket nonetheless. 

This year, I’m even less connected to college basketball than usual. I don’t know any of the top players/prospects, and don’t have a clue who’s doing what, and I haven’t even casually watched a few minutes of a game while flipping through channels. Given my completely disconnected state, I thought this would be a good time to test my intuition against that of AI. 

I fired up ESPN’s Tournament Challenge, filled out my own bracket, then asked ChatGPT to do the same. Here’s what that looked like:

Daniel’s Bracket

Again, I was totally disconnected from the NCAA season this year. So I defaulted to a handful of criteria:

• My familiarity with a school’s past success
• Whether or not I had ever heard of the school
• The location of the school based on my own or my loved ones’ geographical preferences
• Whether or not the school was named after my 10 year old’s favorite president

Maybe not the most scientific approach, but I like my chances. Here’s where I landed:

ChatGPT’s Bracket

On the ChatGPT front, getting answers was an arduous process that involved a whole lot of coaxing. 

First off, it pretended it couldn’t use the internet to see who was in the tournament. After a couple rounds of back and forth, it finally admitted that it did know who was playing in a handful of first round matchups. 

Secondly, it didn’t want to identify the schools when giving me the analysis – it would only tell me which seed from which conference would win each matchup. This means a whole lot of decoding for me. 

Finally, as we got deeper into the rounds of the tournament, it became a whole lot squirrelier about providing an actual answer – it began dancing around making a pick, relying on narratives about how it’s going to be a close fought battle, etc. I ultimately had to be a bit firmer and just demand that it pick a winner for those final contests. 

After all the struggles, here’s where ChatGPT arrived on its bracket predictions:

So Who Wins? 

Well there’s the tricky part – we won’t have a final answer until early-mid April. Although it’s possible we both get completely blown out in the first round. I’ll check back periodically and see how we’re looking, then recap once the tournament is over. Until then, who do you think wins?