Google has teamed up with the NCAA for this year’s March Madness (it’s apparently the “official cloud” of the tournament), bringing machine learning and statistical analysis to decades of game data. But now they’re going even further, with an audacious real-time experiment that could well blow up in their face.
As a blog post from the Google Cloud team explains, a group of basketball enthusiasts, data scientists, and computer techs set out to analyze the data from 80 years of NCAA games and unleash machine learning on this massive pile of data to uncover trends or facts that might not be readily apparent.
For example, if you’re picking an upset, go with the cats — teams with a feline mascot are more likely to bust your bracket.
If you’re a statistics nerd, this is all interesting stuff, but so far the “Wolfpack” (that’s what the team is calling themselves) has focused on historical data culled from thousands of minutes played. But what if they turned their efforts to anticipating what would happen next?
That’s exactly what they’ll be doing this weekend, and they’ll even be airing their predictions in a commercial set to air during halftime of the Final Four games.
Other sports have been using predictive models for quite some time, of course. For instance, baseball uses WAR (Wins Above Replacement) to estimate how much a player will add to a team’s win total as compared to a hypothetical average player. That’s over an entire season though, with hundreds of at-bats. The Google team is planning to estimate what will happen in the final 20 minutes of a single game, which is a tiny sample size.
During the first half, the team will use all the data they can gather and crunch the numbers up, down, and sideways against historical tournament performances, team tendencies, basic strategies, and anything else they can come up with to produce statistical predictions that they think are highly probable to occur.
On site at the Alamodome in San Antonio, they will then quickly turn their predictions into a television ad, which will be passed on to CBS and Turner to air on TBS just as the second half begins.
They aren’t planning to predict the game winner, so don’t get your bookie on the phone just yet. The predictions could be anything — number of three-point attempts, offensive rebounding percentage, even minutes played.
“One of the exciting things about running an experiment like this in real time is we don’t know for sure what will happen,” the team wrote on their blog. They could be right on the money or wildly off base. Either way, it will be fun for number-crunchers everywhere to watch the results of their experiment unfold in real time.
Related Posts
Your WhatsApp voice notes could help screen for early signs of depression
The study, led by researchers in Brazil including Victor H. O. Otani from the Santa Casa de São Paulo School of Medical Sciences, found that their AI could identify depression in female participants with 91.9% accuracy. All the AI needed was a simple recording of the person describing how their week went.
Talk to AI every day? New research says it might signal depression
This finding comes from a national survey of nearly 21,000 U.S. adults conducted in 2025, where participants detailed how often they interacted with generative AI tools and completed standard mental health questionnaires. Within that group, about 10% said they used AI daily, and 5% said they engaged with chatbots multiple times throughout the day. Those daily users showed higher rates of reported depressive symptoms and other negative emotional effects, such as anxiety and irritability.
You might actually be able to buy a Tesla robot in 2027
The comments follow a series of years-long development milestones. Optimus, which was originally unveiled as the Tesla Bot in 2021, has undergone multiple prototype iterations and has already been pressed into service handling simple tasks in Tesla factories. According to Musk, those internal deployments will expand in complexity later this year, helping prepare the robotics platform for broader use.