Basketball players in action | Andreyuu on 123rf
Drama, suspense, and nail biting upsets, it is that time of year again. March Marchness! Can artificial intelligence lessen the stress by predicting March Madness winning brackets? If so, can AI pick the NCAA men’s basketball champion or finish with a higher score than my predictions? Before we go any further, I must confess that I have only been participating in the March Madness mayhem for three years and that is considered below novice status in “hoophead” speak.
We use Chatbots for just about everything today, so it doesn’t hurt to ask. I’m not embarrassed to say that if AI finishes with the winner or higher bracket points, game on next year!
A basketball Aarena | JC Gellidon on Unsplash
“The Big Dance,” one of the names coined for March Madness, is the annual NCAA men’s college basketball tournament. It is one of the most anticipated and watched events in sports. The men’s division has been playing since 1939 according to NCAA.com. The tournament is single-elimination consisting of 68 teams which is quickly reduced to 64 teams after the opening round, “first four.” The remaining 64 teams are divided into 4 regions (South, East, West, and Midwest) of 16 teams that compete in seven rounds until the “final four.” The final two spots to compete for the national championship. The women’s division shares the structure.
Georgia Bulldogs basketball game | Ryan Sepulveda on Unsplash
Selection Sunday held on March 17, 2024, was the unofficial beginning of March Madness when the Selection Committee revealed the full tournament brackets, including all teams and all seeds (based on stats and rankings, but no set formula). Our very own Horned Frogs made the tournament bid for the third year in a row at No. 9 seed and played No. 8 Utah State in the first round on Friday, March 22, 2024. The tournament games begin with the first four on Tuesday, March 19 and First Round on March 21, 2024. The final NCAA men’s championship game will be held on Monday, April 8 at the State Farm Stadium in Glendale, Arizona (time not posted yet). The women’s championship game will be held on Sunday, April 7 at 3:00 p.m. ET.
Before we get into bracket selections, you should know that no one has ever correctly picked all 67 games, and only a small handful have even made it out of the first weekend without missing a pick. In other words, no one has ever gotten a perfect bracket in the official NCAA Bracket Challenge Game. The stat quoted on their website states roughly 1 in 9,223,372,036,854,775,808 (my novice status) could pick winning brackets. They even sponsor the “Quest for the Perfect Bracket Tracker” show with sports commentators discussing the latest basketball happenings, predictions and superstitions supposedly to help you build a “perfect” bracket. ESPN also runs the number 1 “Men’s Tournament Challenge ” bracket game where you can compete with friends or family, join other groups, watch celebrity picks, and more.
We all want bragging rights even though my selections are just for the madness excitement. So, this year I decided to experiment, if possible, comparing my vs. AI chatbot’s bracket predictions. Can AI predict a winning TCU March Madness bracket?
Basketball player slam-dunk shot | August Phlieger on Unsplash
My bracket selections are based on the basics, how many games each team has won,
comparing divisions (SEC vs Big 12), review the top ten teams, look at a few player stats if can’t decide between two teams, and if all else fails then pick the color of the team uniform I like best. I’m not a “hoophead,” I don’t watch every game, I don’t know player injuries, or coaching stats, etc. As I said before, my bracket selections are for fun, but I wouldn’t mind a few bragging rights. Keep posted for updates!
New York Times building in New York City | Marco Lenti on Unsplash
After doing a little research, I discovered that the New York Times (NYT) ran a similar experiment last year. I’m pretty sure we will be hearing a lot more about AI assisted March Madness bracket selections in the future. The NYT experiment used Microsoft’s Bing chatbot and reported errors in the first round. After a long exchange, the bot wouldn’t respond with the same answer each time and the phrasing of the question mattered. Additionally, the Bing chatbot isn’t exactly designed to compete with expert forecasts or mathematical tournament prediction models. Microsoft has said that its system has struggled to keep up with live sports information. The chatbot frequently cited outdated or incorrect details about teams, even if its overall impressions seemed valid. There is still a lot of technical jargon that needs to happen before chatbots can be used for live sporting events, such as finding quality data sources with valued input, clean and plot the data, program and write the algorithm. My conclusion after reading the article, AI is NOT ready to pick a winning bracket just yet. I soon discovered the long tedious process which I will not bore you with. Yes, I did have to help Google’s Gemini bot.
Schollmaier Arena | Paris Smith
Horned Frog fans can follow my unconventional predictions at Fans of TCU, PsmithNY
for my personal bracket selections and national champion compared with Google’s Gemini bot selections at PsmithNYAI.
It’s ME vs. Gemini! Any predictions?