Previewing the upcoming men's NCAA tournament.
Hello everyone! This is the first post I’m writing that’s actually intended for this site. I plan to soon have documentation up that explains how everything on the site works, but I’m running out of time to actually do any NCAA Tournament analysis before the tournament starts. For now, if you need a primer on how the models I’ll be discussing work, here’s the lightspeed rundown:
If you want a more in-depth breakdown of these models, I suggest taking a look at my first post, specifically the “The Contenders” section.
With that out of the way, let’s dive into some analysis.
There has been a lot of discourse throughout the season about the “Big Three” of men’s college basketball: Duke, Arizona, and Michigan. My models similarly see them as prohibitive favorites for the title. All three of the models have those three at the top of their national championship odds in some order, with Resumetric liking Arizona the most, and Minnow and Branchy both preferring Duke, Branchy overwhelmingly so, giving them a 52% chance to win the title. The dropoff to the next tier of teams is significant. Resumetric has Michigan with the third-best odds at the title at 18%. Houston, with the fourth-best odds, sits at 4% by comparison. Even Minnow, which is the least bullish on the big three, still favors them over the rest of the field combined.
| Minnow | Branchy | Resumetric | |
|---|---|---|---|
| Duke | 31% | 52% | 32% |
| Arizona | 12% | 23% | 36% |
| Michigan | 22% | 12% | 18% |
| The Field | 35% | 13% | 14% |
One possible complication: my models do not account for injuries, so Duke (Ngongba and Foster) and Michigan (LJ Cason) are both likely more vulnerable than my model suggests.
That aside, my models say there’s not much benefit to going against the conventional wisdom for your champion. Pick one of the big three.
How did the committee assign seeds this year? I’ve flagged and charted every instance of the committee’s actual seeding differing from the seed my models implied each team deserved by two or more seed lines so we can find out.
Teams the committee seeded better than the model thinks they deserve.
Teams the model thinks deserved a better seed than they received.
Teams the committee seeded better than the model thinks they deserve.
Teams the model thinks deserved a better seed than they received.
Teams the committee seeded better than the model thinks they deserve.
Teams the model thinks deserved a better seed than they received.
Overall, it seems like the committee did a pretty good job this year. Allegedly, a team’s resume metrics get them into the tournament, and their predictive metrics determine their seeding, but of my three models, it was actually Resumetric (the resume-based metric, as opposed to Branchy and Minnow, which are both predictive) that had the fewest qualms of any of them with the seeding, with only Miami OH and TCU significantly underseeded, and only Saint Louis significantly overseeded.
It might just be the case that those three teams are hard to pin down in terms of quality, though. All three of those teams appear in both the “overseeded” and “underseeded” categories of at least one model. Miami OH in particular takes the cake for being the hardest team for the models to pin down, laying claim to the widest range of implied seeds with the lowest being a 13-seed grade from Minnow, and the highest being a 6-seed grade from Resumetric.
In terms of the consensus, there were no teams that all three models thought were significantly overrated or underrated, but there were three teams that both Branchy and Minnow agreed were significantly over or underseeded. Those teams were:
According to Minnow, Tennessee vs. SMU would have been the “least fair” matchup of the first round, with the implied matchup of an effective 4-seed playing an effective 9-seed. However, SMU failed to uphold their end of that bargain, falling to Miami OH in the First Four, and Minnow is much more bearish on Miami’s effective seed, as we discussed earlier. Ironically, the least fair matchup according to Resumetric is Miami OH vs. Tennessee, with an implied matchup of 6 vs. 6.
While we’re on the topic of fairness, let’s take a moment to acknowledge the fallen: the eight most deserving teams left out of the tournament according to Minnow and Resumetric. (Branchy is a true matchup-predictor, not a rating system, so while we can form rough power ratings for teams in the tournament, it’s less informative to apply those ratings to teams that did not make the tournament.)
National rank according to the model in parentheses
Looking at this list, there aren’t really any big snubs. I know there was some controversy about Auburn leading up to the tournament because their schedule was very hard and their predictive metrics were better than their record suggested. The part about the predictive metrics certainly holds true. Auburn was the best team not to make the tournament according to Minnow (our predictive metric) by a wide margin. However, their resume was much less impressive, and remember, the committee maintains that your resume is what gets you in. They certainly made good on that this year according to Resumetric. Although Auburn was again the most deserving team not to make the tournament, only three teams with worse resumes made the tourney over Auburn: Texas (45), Saint Louis (46), and SMU (49). I think the committee would have started a riot if they had left out Saint Louis in favor of Auburn, but I think they could have left out Texas or SMU for Auburn and there would have only been mild griping. That said, this was hardly a clear snub, so on the whole, it seems like the committee did a good job with who they selected to the field.
Is midnight striking already for our tournament Cinderellas? Last year’s bracket was one of the chalkiest on record, with all four 1-seeds making the Final Four, and only a single double-digit seed (a high-major 10-seed at that) making the second weekend. Will we see more chaos and underdog stories this year?
To qualify as an “upset” there needs to be a gap of at least three seed lines between the teams. Otherwise, it isn’t much of an upset. To qualify for this section, the matchup must meet these rules of thumb:
First Round:
Beyond the first round, there are so many matchups that can occur, and generally the gap between seeds playing narrows. Instead of trying to pinpoint individual matchups of note, let’s see if we can spot Cinderella by looking at the general numbers. Which double-digit seed is most likely to make the second weekend according to the models?
While there could be intrigue in many of the 5-12 and 6-11 games, overall it’s looking pretty bleak for Cinderella. But hey, they call it “March Madness” for a reason, and I feel confident that a team will come out of nowhere and surprise us. It happens every year.
Thank you to everyone who is starting this journey with me. It feels a bit silly to make a big deal about something as small as starting a blog, but it’s been exciting to be able to share my passions with everyone, and it’s been so nice to hear the support and words of encouragement from those of you who have read and given feedback in the early days of this site. I still have a lot to learn about blogging, web development, and sports modeling, but here’s to first steps.
Wishing everyone a tournament that lives up to its moniker.
Happy March Madness,
Noah