Posts / NCAAW

---

Virginia was a First Four team. They had to win a play-in game just to get into the field of 64. They were seeded 10th. And on Sunday they beat Iowa, a 2-seed, 83-75 in the Round of 32.

That is the number that anchors this tournament right now. Not because Virginia is going to win the whole thing (BXS gives them 28% to beat TCU in the Sweet 16, and less than 1% to win the title). But because of what it reveals about a region that was already interesting: South Carolina's Sacramento 4 pod has more chaos per bracket slot than any other quadrant in the draw.

There was also this: Michigan, a 2-seed, put up 92 points and beat 3-seed Louisville by 23. The model had Michigan favored at 56-44 going in. Louisville backers who trusted the seed number over the ratings had a rough weekend.

Round 2 Results: Who Advanced

Before the Sweet 16 picks, here's the full Round 2 scorecard:

Fort Worth 1 (UConn's region)

North Carolina knocking out 5-seed Maryland is the cleanest result here. BXS had Maryland as the lower-rated team despite the higher seed (1797 vs 1822 for UNC), so calling this an "upset" overstates it. The model would have given North Carolina the edge. They advance to face UConn.

Notre Dame over Ohio St. is the more interesting result. Notre Dame came in rated 1828, Ohio St. at 1846. The Buckeyes had a 7-point rating edge. Close game, defensively framed, Notre Dame survived 83-73. They now get Vanderbilt.

Fort Worth 3 (Texas's region)

Texas was not messing around. 100 points against Oregon, a 42-point win. That is a dominant performance against a team that BXS had rated about 200 points below Texas (1755 for Oregon vs 1977 for Texas). The simulation results showed that outcome often; the bracket just needed to confirm it.

Kentucky over West Virginia by one point, BXS 1848 vs 1856 for WVU. The model had West Virginia as the slightly better team. A one-possession game went the other way. That is how it goes at the margins.

Louisville over Alabama, another one-possession game, 69-68. BXS had Louisville at 1845, Alabama at 1801. Louisville was the model's preferred team here, and they held on. They advance to face Michigan, where the model flips and strongly disagrees with seed-based expectations.

Sacramento 2 (UCLA's region)

LSU dropped 101 points. BXS has them rated 1933, tied for the highest 2-seed rating in the field. The model does not view LSU as a 2-seed in strength: they are closer to a 1-seed. The Sacramento 2 path to the championship runs through UCLA, and that bracket is shaping up as the most talent-dense in the draw.

Minnesota over Ole Miss (65-63) is an interesting result. BXS had Minnesota rated 1811, Ole Miss at 1807. Essentially a coin flip. The 4-seed won.

Sacramento 4 (South Carolina's region)

Virginia won going away. Iowa, despite the 2-seed, carried a BXS rating of 1842. Virginia was at 1730. That is a 112-point gap. In BXS terms, that should translate to Iowa winning roughly 80% of simulations. Virginia won anyway. They are the tournament's best story going into the Sweet 16.

The Full Sweet 16 Board

Here is the complete projection table for all eight Sweet 16 matchups, with BXS ratings and win probabilities from the model:

The Upset Pick: Michigan Over Louisville

Michigan comes in as a 2-seed facing 3-seed Louisville. On paper it looks like a standard seeding matchup. On the rating sheet it is not close.

Michigan's BXS rating is 1875. Louisville's is 1845. That 30-point gap is meaningful: BXS converts to a win probability of 56-44 in Michigan's favor. And Michigan just dropped 92 points on NC State, a 7-seed that came in rated higher than Louisville. This is not a mild lean. The model is telling you Michigan is the stronger team.

Louisville's path to this point, beating Alabama 69-68 on one possession, does not inspire confidence for a step up in competition. Michigan has the higher BXS rating and the cleaner overall profile.

The pick: Michigan over Louisville, Sweet 16.

The Other Angle: Vanderbilt Over Notre Dame

This one does not look like an upset on paper at all. Notre Dame is the 6-seed, Vanderbilt is the 2. Conventional wisdom says Vanderbilt wins going away.

BXS agrees, but the probabilities are worth examining: 60-40 Vanderbilt. That is a stronger lean than the raw seedings suggest, but Notre Dame at 40% is not dead. The Irish are rated 1828, only 49 points behind Vanderbilt's 1877. A 49-point gap in BXS terms is meaningful but not a blowout. Notre Dame just beat Ohio St., who came in rated above them at 1846. If they can stay disciplined against Vanderbilt, the 40% is real.

The Model's Blind Spot: Virginia's Momentum

The model gives Virginia 28% against TCU in the Sweet 16. TCU is rated 1856 to Virginia's 1730, a 126-point gap. In normal circumstances, this matchup has TCU winning about 75% of simulations.

But Virginia just upset Iowa, who was rated 112 points above them. The model does not adjust for recent performance or tournament momentum because those effects are real but hard to quantify without adding noise. Virginia won that Iowa game by playing disciplined defense and keeping the game closer than the ratings suggested it had any right to be. TCU is a better team than Iowa. That does not mean the 28% is a floor.

Is 28% right? Possibly. Is it understating a team that just outperformed their rating in the most important game of the season? Also possibly.

Championship Odds

Three things jump out here.

First, UCLA leads at 20% with Texas right behind at 19%. The model's projected championship game is UCLA vs Texas (51-49 UCLA), which is a coin flip. The difference of 7 BXS points between them is statistical noise.

Second, LSU at 9% is the most undervalued team in the draw relative to what casual bracket players will expect. The Tigers are rated better than every team in the field except the four 1-seeds, and they are on a collision course with UCLA in the Elite 8. That is the matchup to watch.

Third, UConn (15%) and South Carolina (14%) are running behind UCLA and Texas by 5-6 points despite also being 1-seeds. South Carolina's BXS rating (1957) is the weakest of the four 1-seeds. They are still very good. They are just not the model's pick to win the whole thing.

What to Watch

The Elite 8 picture is coming into focus. On the left side of the draw, the model projects an UConn-Vanderbilt Sweet 16 matchup (UConn 61-39 favorite) leading to a Texas-Michigan Elite 8 (Texas 65-35 favorite). On the right side, UCLA over Minnesota and LSU over Duke should set up an all-time Elite 8: UCLA vs LSU, with UCLA favored 59-41.

If the bracket holds to the model, you get a Texas vs UCLA final. If one upset breaks through, the most likely candidates are Michigan beating Louisville (56%), Notre Dame over Vanderbilt (40%), or Oklahoma upsetting South Carolina (33%).

Virginia beating TCU would be the headline. But after Iowa, nobody is ruling them out.