Can AI help you win your March Madness bracket? One disruptor bets $1 million on ‘yes’ (and Houston)

By Canadian Press on March 19, 2025.

DENVER (AP) — Perhaps the surest sign that artificial intelligence really is taking over the world will come the day it wins your favorite March Madness bracket pool.

The day could be coming soon.

In an experiment that a) was bound to happen, b) might actually make us all look smarter and c) should probably also scare the daylights out of everyone, a successful CEO-turned-disruptor is running a $1 million March Madness bracket challenge that pits his AI programmers’ picks against those belonging to one of the world’s best-known sports gamblers.

“We’re not a crystal ball,” says Alan Levy, whose platform, 4C Predictions, is running this challenge. “But it’s going to start to get very, very creepy. In 2025, we’re making a million-dollar bet with a professional sports bettor, and the reason we feel confident to do that is because data, we feel, will beat humans.”

Levy isn’t the only one leveraging AI to help people succeed in America’s favorite pick ’em pool — one that’s become even more lucrative over the past seven years, after a Supreme Court ruling led to the spread of legalized sports betting to 38 states.

ChatGPT, a chatbot developed by OpenAI, is hawking its services to help bracket fillers more easily find stats and identify trends. Not surprisingly, it makes no promises.

“With upsets, momentum shifts, and basketball’s inherent unpredictability, consistently creating a perfect bracket may still come down to luck,” said Leah Anise, a spokesperson for OpenAI.

Also making no promises, but trying his hardest, is Sheldon Jacobson, the computer science professor at Illinois who has been trying to build a better bracket through science for years; he might have been AI before AI.

“Nobody predicts the weather,” he explained in an interview back in 2018. “They forecast it using chances and odds.”

$1 million on the line in AI vs. Sean Perry showdown

Levy’s angle is he’s willing to wager $1 million that the AI bracket his company produces can beat that of professional gambler Sean Perry.

Among Perry’s claims to fame was his refusal to accept a four-way split in a pot worth $9.3 million in an NFL survivor pool two years ago. The next week, his pick, the Broncos, lost to New England and he ended up with nothing.

But Perry has wagered and won millions over his career, using heaps of analytics, data and insider information to try to find an edge that, for decades, has been proprietary to casinos and legal sports books, giving them an advantage that allows them to build all those massive hotels.

Levy says his ultimate goal is bring that advantage to the average Joe — either the weekly football bettor who doesn’t have access to reams of data, or the March Madness bracket filler who goes by feel or what team’s mascot he likes best.

“The massive thesis is that the average person are playing games that they can never win, they’re trading stocks where they can never win, they’re trading crypto where they can never win,” Levy said. “4C gives people the chance to empower themselves. It’s a great equalizer. It’s going to level the playing field for everyone.”

But can AI predict the completely unexpected?

It’s one thing to find an edge, quite another to take out every element of chance — every halfcourt game-winner, every 4-point-a-game scorer who goes off for 25, every questionable call by a ref, every St. Peter’s, Yale, FAU or UMBC that rises up and wins for reasons nobody quite understands.

For those who fear AI is leading the world to bad places, Levy reassures us that when it comes to sports, at least, the human element is always the final decider — and humans can do funny and unexpected things.

That’s one of many reasons that, according to the NCAA, there’s a 1 in 120.2 billion chance of a fan with good knowledge of college basketball going 63 for 63 in picking the games. It’s one of many reasons that almost everyone has a story about their 8-year-old niece walking away with the pot because she was the only one who picked George Mason, or North Carolina State, or VCU, to make the Final Four.

“You can’t take the element of fun and luck out of it,” Levy said. “Having said that, as AI develops, it’s going to get creepier and creepier and the predictions are going to get more and more accurate, and it’s all around data sets.”

Levy suggests AI is no three-headed monster, but rather, an advanced version of “Moneyball” — the classic book-turned-movie that followed Oakland A’s GM Billy Beane’s groundbreaking quest to leverage data to build a winning team.

Now, it’s all about putting all that data on steroids, trying to minimize the impact of luck and glass slippers, and building a winning bracket.

“We’ve got to understand that this technology is meant to augment us,” Levy said. “It’s meant to make our lives better. So, let’s encourage people to use it, and even if it’s creepy, at least it’s creepy on our side.”

The AI’s side in this one: Houston to win it all. Perry, the gambler, is going with Duke.

___

AP March Madness bracket: https://apnews.com/hub/ncaa-mens-bracket and coverage: https://apnews.com/hub/march-madness. Get poll alerts and updates on the AP Top 25 throughout the season. Sign up here.

Eddie Pells, The Associated Press