UPDATE: The Excel file has been updated as of 11:50 PM on March 20 including First Four results and updated injury info affecting Kansas. For data integrity reasons — so you can check my work — I’ve preserved the screenshots “as is” in this post before any games were played. But there’s a new version of the spreadsheet for paying subscribers.
UPDATE: Injury info refreshed at 6:30 PM on March 19. Odds and tables have been changed accordingly.
Dating back to when FiveThirtyEight was at the New York Times, I’ve run March Madness projections every year. And it would be a shame to skip out on the fun this time just because of all the carnage in the media industry. So I dusted off the old model, which is basically the same forecast l I ran at FiveThirtyEight every year from 2011 to 2023. It took longer than I was expecting — basically all day instead of “a few hours” — but the good news is the numbers are ready.
Other than perhaps the summer stretch run of the World Series of Poker, March is my favorite time of year. If you live in the Northeast or the Midwest, the weather becomes almost halfway bearable. The presidential primaries are usually exciting at this stage — even if they weren’t this time around. There’s spring training down in Florida and Arizona. And then there’s the NCAA basketball tournament. It’s probably not the best sporting event in the world in terms of the quality of play; basketball isn’t a very forgiving sport, and if you’re mostly an NBA guy like I am, you’ll notice that skill level of guys who are just 18 to 21 years old is lacking compared to the pros (Victor Wembanyama aside).
But it’s probably the most fun event on the sports calendar. That’s because the tournament is nearly always a wide-open affair, and this year is no exception. The model has Connecticut as the favorite with a 23 percent chance of repeating as champs, followed by Purdue at 15 percent, Arizona at 13 percent, and Houston at 10 percent, but that means there’s a 39 percent chance the winner won’t be one of those four.
In case you’re wondering, this is partly intended as a proof-of-concept for (hint, hint) what it might be like to run a more low-key version of the presidential election forecast for paid subscribers. And it’s partly to give the sports models their due. (Figuring out what to do with hasn’t been my top priority with everything else going on, but I own the IP and am still looking for a long-term home if I don’t find some way to host them on Substack — feel free to drop me a line about that.)
But it’s mostly because — well, I wanted these numbers for myself. You think I’m not going to be sweating the action?
So here’s the drill. What follows are projections for each team’s possibility of advancing to each round of the men’s tournament. (I’d love to run the women’s numbers too and I think the newfound excitement in women’s basketball is super cool, but that’s another full day of work and I’m not sure how likely that is.) I’ll provide the numbers the first men’s region, the East, for everybody, and the rest for paying subscribers. Paid subscribers will also get the following:
XLS files with more precise numbers for each team, as well as a compilation of the various power ratings the model uses, including my own Elo ratings which aren’t published anywhere else.
Updates after the first and second rounds. From that point onward, there aren’t that many more games to play so I’ll probably just post updates for everyone unless the Elite Eight matchups are particularly compelling.
Plus, the opportunity to permanently lock in current pricing to all other paid content at Silver Bulletin.
Methodologically, the process is essentially the same as the version of the numbers I ran at FiveThirtyEight. These forecasts are based on a composite of power ratings, with a small amount of weight also assigned to preseason rankings — yes, they actually help to improve the predictive power of the model — plus adjustments for injuries and travel distance.
There are a couple of other very minor changes, but the only one that is more than a rounding error is that the longstanding Sagarin ratings now seem to be defunct, so I’ve had to cycle them out. Instead of adding a new rating system, I’ve given the ratings I trust the most, KenPom and my own Elo ratings, 1.5x weight.
Ready? Let’s start in the East, where defending champion Connecticut is the #1 overall seed — and also the best bet for this year, according to the Silver Bulletin forecast.
What’s the catch? It’s got nothing to do with the Huskies. The model does regard them as the best team, in part because they rank #1 in its most favored ratings, Elo and KenPom, and their path takes them through friendly territory in Brooklyn and then Boston. Rather, it’s just that as is often the case, no team in the tournament has more than about a 20 percent chance of winning. The market knows that; Connecticut is also the favorite in Vegas, but with +400 odds at FanDuel, that still implies only a 20 percent chance of winning, similar to the model’s 23 percent.
Otherwise, this region is pretty chalky. The model does like the 7-10 upset in the Drake-Washington State game and it has BYU rated relatively well for a 6-seed. And Florida Atlantic is a bigger favorite than you might think in the 8-9 game over Northwestern, in part because Northwestern has injury issues. It seems like every sportswriter I follow on Twitter went to either Northwestern or Michigan, so this might be a place where you can go against the conventional wisdom just by picking the favorite, although FAU is also favored in Vegas.
The East winner will play the winner of the West in the Final Four. The West is projected to be livelier, with more potential for upsets; here are the odds there:
