Our bracketology is updated this morning. With the last update coming Sunday, that means we’ve got three things affecting the movements:
1. Results Sunday and Yesterday
The games themselves are the biggest impacters, as one would hope.
2. A Clearer Picture of Saturday’s Impact on Ratings Systems
We use one-day proxies for the ratings systems so we can run the model overnight, something that gives us more simulations (good), an earlier posting time (good), and some day-later ironing out of results (not as good). Most of the time, the ironing out is minimal, but every now and then, it makes a difference.
3. Conference Scheduling
With the end of the season in sight, we’re getting clearer pictures from conferences about how they’ll actually handle postponed games and their conference tournaments. As of this morning, the following conference schedules and tournament formats are almost entirely finalized in our model, only lacking things like reseeding and the occasional home/away scenario adjustment:
Atlantic 10
America East
Atlantic Sun
Big South
Horizon League
Missouri Valley
Ohio Valley
West Coast
We’ll keep adding more to this list as the days go on. This, like number two, isn’t all that significant, but it occasionally does play a role, and if you’re a fan of a team in one of these specific conferences (or a fan of a team next to some teams from these leagues in the seed list), this does impact you to some extent.
***
Now, who moved:
Moving Up: Houston, Virginia Tech, UNC, St. Bonaventure
Houston bounced right back from the Wichita State loss, annihilating Cincinnati at home on Sunday. They’re still not where they’d probably like to be (were they more consistent, they’d be undefeated right now), but they’re back up to a three-seed, which is at least enough to guarantee oneself three rounds free of Gonzaga and Baylor.
Virginia Tech yo-yoed, not “despite” not playing so much as in conjunction with not playing. There are a few things to make of this: Virginia Tech’s in a tight place in the seed list, where little changes move teams a lot. Virginia Tech’s signature wins (Villanova on a neutral court, Virginia at home, Clemson at home) are of variable quality depending on the day. What the ACC does with scheduling (we make some adjustments within our model prior to conference finalization to keep schedules even in terms of game quantity) can make a big difference. But the biggest thing to take away here is the impact of a single win or loss. Virginia Tech, like many teams, is on the cusp between two projected median records. As their odds change in the next three weeks, so does their expected seeding. The importance of individual games isn’t growing. The leverage is.
UNC got a boost from that Louisville result on Sunday. They get another today. One flaw of our NET proxy that we’ll need to correct is that it doesn’t account for margin of victory as much as it should. It underestimated the strength of that UNC win in NET’s eyes.
Finally, St. Bonaventure. The Bonnies got a big home win over Davidson, and it was impressive, coming by double digits in one of the lowest-possession-count games of the year. They’re the A-10 favorite now, and they’re somewhat comfortably off the bubble, at least for the moment.
Moving Down: Virginia, UCLA
UVA drops today from being the last two-seed to the first four-seed. Noteworthy, but not enormous. The Cavaliers got passed by Iowa, who increased their projected final win total by one with the win over Penn State. They got passed by Houston, of whom we spoke above. Their loss to Duke was a little worse in the ratings’ eyes than we anticipated, and with that on top of those Sunday results, UVA got passed by three more teams as well. Not disastrous, but a big shift for Tony Bennett’s squad.
For UCLA, the underwhelming win over Arizona State was, like UNC’s overwhelming victory against Louisville, misestimated by our NET proxy. We’re hesitant to change that proxy, since these happenings are exceptions more than the norm and we don’t want to overcorrect, but we do want to minimize day-later movement like UCLA’s and UNC’s, so we’re beginning to update our bracketology more frequently in order to at least catch these misestimations earlier. A good way to think about our model in this area is to probably consider it half a day behind. In the future, we want better, but right now, that’s where we’re at.
Moving In: Xavier
The Musketeers played a solid game on Sunday and climbed back across the cut line. They’re right on the bubble with a treacherous three weeks ahead, but for right now, they’re projected to be safer than not.
Moving Out: Penn State
Three years in a row now, our model has been optimistic on Penn State relative to the norm. It was wrong in 2019 (projected PSU to make the NIT somewhat comfortably—they did not make it). We didn’t find out whether it was wrong last year (again projected PSU comfortably in the NIT—there was no NIT). On Sunday, it was still projecting a team that was 7-11 at the time to grab an 11-seed in the NCAA Tournament.
What’s going on?
Penn State has faced historically difficult schedules each of the last three years. The Big Ten is historically deep these days, and Penn State has routinely borne the brunt of that. Without more data upon which to tell our model, “If a team’s __ games under .500, knock them down this many pegs,” we let the model run wild in State College, and it likes what it sees a lot: A good team with respectable numbers in the ratings systems. It doesn’t know how much humans hate sub-.500 records. We probably need to find a way to tell it next year.
Anyway, I was a little relieved to see Penn State drop well into the NIT this morning after Sunday’s loss in Iowa City.
***
We’re currently set to do another bracketology update tomorrow morning, so check back at that time. The next conferences on our list to almost-finalize are the SoCon and the Sun Belt, but hopefully we’ll chomp off even more than those tonight so we can stop worrying about scheduling soon and narrow our focus to the games themselves. Thanks for your patience, and please ask questions if you’re curious about anything. Our model’s not infallible. But over the next few years, we’re trying to get it there.
