A week and a half ago, I wrote that we’d gotten our model through our transition from the early-season proxy for NET to the late-season proxy for NET.
I was mistaken.
Thanks to my own oversight error, the transition was only half-completed. It was completed yesterday, and the shifts from it weren’t meaningless, especially in the case of Colgate, whom we’ll discuss in a moment. So as you look at our updated bracketology today, know that there are three causes for possible movement:
- Results since Sunday morning, when we last updated the model.
- Clearer views of results from Saturday, as our model’s proxies incorporate data from the real ratings systems’ reactions to those games (this is always something that happens—not unique to today).
- The NET proxy transition (unique to today).
Apologies for the misinformation on the 7th. Here’s who moved:
Moving Up: Colgate
The Raiders are a projected 10-seed, buoyed by their 11th-place ranking in NET.
I’ve tended to be a defender of NET. I don’t think it’s the ideal design, but I think the NCAA could do worse, and I think its impact is overstated. The problem with it is that it’s not robustly designed to handle a season in which some teams, like Colgate (ranked 11th) and UC-Santa Barbara (ranked 45th), play either no nonconference games (as in Colgate’s case) or play an extremely limited nonconference slate (UCSB played only three nonconference D-I games, and played just two opponents in those games). If you’re blowing teams out and you’re not losing, NET will estimate you as better than you are. This year.
With NET comprising roughly a third of our model’s eventual “selection score,” this really helps Colgate in our model’s eyes. The question is: Does it actually help Colgate? Or rather, how much does it help Colgate? Because we know it doesn’t hurt.
Apologies for the overused word, but this season’s unprecedented. We don’t know what the committee’s going to do. We don’t know how they’re going to interpret the data on hand. They could say, “Colgate’s best opponent was Army, and they went 3-1 against them,” and leave them in firm automatic-bid territory. They could say, “You know, we can’t blame Colgate for not playing nonconference games, and they sure dominated the Patriot League,” and make them a 10-seed.
My best guess? In the median simulation, in which Colgate really does just win out (they’re that much better than the rest of the league), they’ll end up in between where our model has them and where Bracket Matrix will have them, which will likely still be a 13-seed considering where cut lines should fall and who should win the OVC/MAC/Big South/Conference USA auto-bids. In other words, if I were forced to make a firm prediction, I’d guess that Colgate’s true projection should be for an 11-seed or a 12-seed.
Do I not trust our model? I wouldn’t say that. I trust it with every team in its current projected tournament field, with the exception of Colgate, a team the model isn’t really built to handle, given it was built on seasons much more similar to one another than this season is to those seasons. And I still think it’s useful. In a normal season, we’d expect a team with Colgate’s numbers in the ratings systems to be a 10-seed, albeit one for whom one loss would be catastrophic for their season (expect a two or three seed-line drop if Colgate loses at some point). That’s good to know. It might even end up being correct.
Moving Down: Wisconsin, Colorado, Minnesota
Wisconsin’s drop is mostly driven by their result on Sunday. Our NET proxy wasn’t that far off on them. The Badgers are in a tight spot in the field in terms of selection score, so little things make a big difference, and coughing up a home game, even against a team as good as Michigan (a projected 1-seed), is a little-to-medium thing.
For Colorado, our model knew the loss to Cal on Saturday was bad, but it didn’t realize quite how bad it would be for the Buffs. Combined with some other teams moving up a step, Colorado’s now solidly in the middle of the field, close enough to the bubble to be at risk.
Minnesota’s demise is half-result half-NET proxy. Our proxy overestimated NET’s estimation of the Gophers. Couple that with a rather thorough loss in College Park, and the Gophs have the bubble in their sights.
Moving In: Colorado State, Drake, Boise State, Norfolk State (auto-bid)
NET loves the Mountain West. This, we underestimated. Welcome back to Boise State and Colorado State, and best of luck to the league as they attempt to get Utah State in there and nab four bids.
Drake’s win Sunday over Loyola vaulted them back into things. Not a shock, but less than 50% likely, so not present in the median simulation.
Norfolk State finished off a road sweep of Morgan State and is firmly the team to beat in the MEAC Tournament, which is still evidently slated to happen.
Moving Out: Penn State, SMU, Stanford, Morgan State (auto-bid)
SMU and Stanford were the last two teams in our projected field Sunday morning, so chalk their exits up to the MWC/NET romance. Penn State, on the other hand, lost to Nebraska in State College. Not the move you want to make.
***
This should really be the last ratings proxy-related shuffle of the deck. Apologies again, and thanks for bearing with us. Expect new shuffles soon, though, as we start figuring out what the hell’s going on with conference tournaments and postponed games in every league. Next update is planned for Friday morning.