After selecting a scorching hot 27 out of 32 winners for the Round of 64 picks, my model choose just 10 out of 16 at predicting games in the Round of 32, and only six of the model’s initial Sweet Sixteen predictions have ended up reaching this round.
In order to evaluate the model’s performance and hopefully win back over some people who might think my model simply got lucky in the Round of 64, I’d like to dig into why some of these predictions were off.
As I’ve made clear, each prediction also includes confidence, which does a good job of explaining why some predictions were not correct. In the Round of 64, the model went 14/16 on games in the upper half of confidence, and 13/16 on games in the lower half, with two incorrect predictions coming on the four least confident games.
In the Round of 32, the model was 7/8 in games in the upper half of confidence, while going just 2/8 on the rest of the round, including getting four of the five predictions with the least confidence wrong.
Now, to dive more into some individual predictions and why they were wrong. First of all, the model picking TCU to beat 1-seed Arizona was surprising, but after their performance last week I can say I’m proud of that choice even though it didn’t work out. The two teams looked incredibly evenly matched and the game even went to OT while many predicted an easy Wildcats win. Meanwhile, Miami over Auburn and UNC over Baylor can potentially be explained away by underrated strength of the ACC. Conversely, Iowa State, Baylor, and Ohio State might have been a tad bit overrated based on the strength of the Big 12 and B1G. While that may sound like a cop-out explanation, I plan on looking into this more as we look ahead to next season. Finally, no one but Shaheen Holloway had Saint Peter’s in their Sweet Sixteen.
Sometimes in a single elimination tournament random, unpredictable things happen and this is likely a good example of that phenomenon (not to imply that the Peacocks are undeserving after pulling an overtime game out against Kentucky and thoroughly handling Murray State).
Overall, the model is 37 for 48, good for 77% on our predictions.
This is still a strong performance in spite of a weak Round of 32, and bodes well for this weekend’s predictions.
The second weekend kicks off with the Sweet Sixteen, wherein we will usually see a Cinderella or two take form and high seeds take losses that have them heading home early.
The first weekend weeds out the good from the great. The second weekend weeds out the pretenders from the contenders. At this point, if you’ve made it this far in the tournament, your seed really doesn’t matter, as is made clear in the plot below. In the first round or two, you can pick up a lot of points by going mostly chalk. This round will feature nothing but teams with top talent going head to head. You have to be sure to make the right picks, and you have to do so with seed playing only a very small role in your selection process.
Here is the only matchup featuring two teams my model knew would be here. I was initially surprised by the degree of confidence at which my model makes this prediction, but after digging into the data I can see why the Blue Devils are seen as an easy pick in this one. This is a matchup of two great teams, but Duke rebounds better and runs one of the most efficient offenses in the country.
Gonzaga scores and rebounds at a higher clip than any other team in this tournament. Arkansas won’t be able to fill in the talent gap against a team that features tournament experience in players like superstar Drew Timme and one of the top NBA draft prospects in college basketball in Chet Holmgren.
One of these two Cinderellas is guaranteed to head to the Elite Eight, and my model believes in the Hurricanes over the Cyclones. Miami has a marquee win on the schedule in their win over Duke that proves that they can hang with anyone, while Iowa State is battle-tested after a grueling Big 12 schedule.
Saint Peter’s run has moved from noteworthy to historic. With a win over Purdue, it would become legendary. No 15 seed has ever reached the Elite Eight. That said, it shouldn’t happen this year. The talent gap between these teams is just too wide, and Saint Peter’s has no way of handling a player like the 7’4 Zach Edey. Add in Jaden Ivey, who is one of the best scorers in the country, and this one could get out of hand early for the Peacocks.
The Jayhawks have the offense to go score for score with Providence and the defense to limit the Friars more than most of their Big East opponents have this season.
This Battle of the Blue Bloods will see last year’s Final Four squad knock off the Tar Heels. While UNC is a scary team, especially after knocking off a Baylor squad many expected to see in the Final Four, they blew a 25 point lead last week that showed just how beatable they are when they don’t have something like Brady Manek going supernova.
I made my doubt clear before each of the last two blogs that predicted Michigan victories: my computer doesn’t know what happened to this team down the stretch. Still, though, the Wolverines are predicted to head to the Elite Eight even after Villanova’s win over the model’s beloved Buckeyes. Now I’m willing to admit that there might be something to that after Michigan showed their mettle gutting out a win against Tennessee.
Here are two teams my model really does not believe in duking it out with a spot in the Elite Eight on the line. My model had each of these teams losing in the first weekend, making this one of only three matchups in the round that the model incorrectly identified both teams for, but this is the only matchup wherein both teams were expected to lose to a lower-seeded team. Because of that, I’m willing to admit that there is something about these teams that my model clearly does not understand. I think the model knows that too, classifying this as its least confident prediction. The Wildcats should beat the Cougars as things get catty in San Antonio, but don’t be surprised to see a semi-home court advantage turn the tides towards Houston.