NIT Bracketology: Wednesday, March 6th

Our NIT Bracketology is updated, and Wake Forest is firmly in the field.

Wake was already in our bracket, albeit in a Bid Thief Seat, so that’s not the biggest change. More significantly, UNLV moves in rather emphatically after their win over San Diego State, leapfrogging the Last Four In. Moving out? VCU, because our model assumes the committee will not consider Max Shulga’s absence as it reacts to that home loss to Duquesne. VCU drops all the way past the First Four Out, and Duquesne becomes our first team out, but their position and UNLV’s are both subject to change as our model gets the real NET/KPI/SOR updates today and stops relying on its own estimates from last night.

Here’s the bracket:

Last Four In, First Four Out, Etc.

On the low side:

• Last Four In (from the cut line going upwards): Yale, Washington, Loyola Chicago, Maryland
• First Four Out (from the cut line going downwards): Duquesne, San Francisco, LSU, NC State
• Others in the mix (>10% NIT chance, below NIT field, not in our field or in the First Four Out): VCU, Rutgers, UMass, Samford, Miami (FL), Indiana, Georgia, Florida State

On the high side:

• Last Four In (from the cut line going downwards): Utah, Texas A&M, Seton Hall, Wake Forest
• First Four Out (from the cut line going upwards): Virginia, TCU, New Mexico, Colorado
• Others in the mix (>10% NIT chance, above NIT field, not in our field or in the First Four Out): Florida Atlantic, Indiana State, Oklahoma, St. John’s, Villanova, Saint Mary’s

As always, we account for bid thieves by looking at the median cut lines in our model’s simulations. Today, those remain 45^th and 82^nd, meaning if there was a master seed list covering both tournaments, the 45^th overall seed would be the last NCAAT at-large and the 82^nd seed would be the last NIT at-large.

Conference Tournament Favorites (Reverse Bid Thieves)

Given where those cut lines stand and how many teams sit between them, we’re including one conference tournament favorite in our NIT Bracketology out of the four who occupy NIT at-large territory. That conference tournament favorite is James Madison, the likeliest of the four to wind up in the NIT.

Including James Madison and Appalachian State at the same time is unrealistic. It’s roughly 70% likely one of the two makes the NCAA Tournament. But each is individually likelier to play in the NIT than the NCAAT, and it’s likely there’s at least one instance of a reverse bid thief—someone who loses their conference tournament and lands in the NIT. Again, JMU is the likeliest team to do this. Here are the four:

• Princeton
• Grand Canyon
• James Madison
• McNeese

Model Talk

We talked about TCU at length yesterday, and we discussed Seton Hall and Colorado on Monday. The bottom line on those is that we don’t think TCU is in as much danger of making the NIT as our model implies (thanks to their best wins), that we think Seton Hall is safer from the NIT as well (again, thanks to best wins), and that our model is probably right on Colorado simply because they have so many routes left to play themselves off the bubble (we will continue to revisit the estimate).

That leaves Texas A&M as our weirdest outlier. Right now, we have Texas A&M in a Bid Thief Seat. Yesterday, nobody on Bracket Matrix had them in the field, and the top-ranked bracketologist on Bracket Matrix told Stu this morning that the Aggies are “pretty close to an NIT lock.”

We aren’t sure exactly what our model sees when it looks at Texas A&M’s median outcome. This doesn’t mean we don’t know what our model is thinking. It means our model is seeing one of a few possible paths as A&M’s median, and we just don’t know which one it is.

Texas A&M is favored by a point tonight on kenpom against Mississippi State. Our model starts with present kenpom, so the Aggies are favored by a point in our model as well. They’re a one-point underdog against Mississippi on Saturday. Our model lets simulations run their course, but the average scenario is, again, in line with kenpom. Mississippi State is currently 33^rd in the NET, and that game’s in College Station. That means there’s maybe a 40/60 chance it could be a Q1 win, or perhaps 35/65 since Mississippi State’s NET will worsen more in scenarios in which A&M wins. Mississippi? 79^th in NET, and that game’s in Oxford. Again, maybe a 40/60 chance of a Q1 win, or possibly 35/65. With A&M roughly 50% likely to win each, we know the median scenario involves a 1–1 path, and our guess is that it involves another Q2 win. Whether the loss is a Q1 or Q2 loss doesn’t directly matter to our model. Our best guess is that the median scenario involves Texas A&M finishing the regular season with 6 Q1 wins, a Q1 win percentage better than .450, and 10 Q1/Q2 wins. In a bunch of categories our model’s seen be responsible for teams outperforming their baseline numbers, Texas A&M narrowly beats out the cut line in the median.

The season isn’t over after those two games, though.

From that median scenario, A&M goes to the SEC Tournament. They’re 8–10, seeded 9^th or 10^th, playing either the 8^th or the 7^th-seed, a 9–9 Mississippi State or LSU. Which they play matters a lot to our model, because if it’s LSU, A&M’s favored to win (and finish 18–15, three games over .500 and therefore only receiving a small knock for their raw W–L number), and if it’s Mississippi State, A&M’s favored to lose (and finish 17–15, two games over .500 and therefore receiving a much larger knock). Which is A&M likelier to play? In the real world, it’s Mississippi State. A&M’s likelier to finish 0–2 against Ole Miss than 1–1, while Mississippi State beat LSU head to head. Our model, though, doesn’t look at conference tiebreakers. This is a known flaw, and we don’t have any excuse. Like the best wins thing, we need to figure this out this offseason.

What’s happening overall, then, is that we think our model’s looking at the 18–15 scenario as Texas A&M’s median: We think our model’s seeing the Aggies beat Mississippi State, lose to Ole Miss, beat LSU, and lose in the SEC quarterfinals and basing its final seed list off of that. It’s not all that unlikely of a path. It is nearly the real median scenario. But that real median, in the end, is worse for the Aggies. It’s 17–15. And with our model also not considering the effects a late-season five-game losing streak and a 2–7 end to the season will have on subjective perceptions of the worth of the team in question, that’s why our model has Texas A&M in a Bid Thief Seat while the industry doesn’t have them sniffing the field. (Also, as Texas A&M fans can tell us from 2022, the committee sometimes makes up its mind too early in the week for the SEC Tournament to have much impact. The quarterfinals don’t happen until Friday.)

The bright spot for Texas A&M is that our model sees a viable path for them to get back to the bubble. It’s overestimating their chances, we’re pretty sure, but it sees a 41.4% chance of an NCAAT bid, and that’s a long way up from 0.0%. There’s a 25%-ish chance A&M sweeps the Mississippi schools this week. If that happens, there’s a pretty good chance at least one ends up being a Q1 win, a seventh of the category. If it happens, A&M will likely be the 7-seed, most likely opening the SEC Tournament against Mississippi, a pretty good draw. Add in the scenario where they do go 1–1 but they maybe draw South Carolina in the SEC quarters or upset one of the SEC’s other top seeds, and by the back of the envelope, I’d give the Aggies close to a 30% chance of breaking into the territory where it’s hard to leave them out. That median scenario our model sees would only get them to the bubble. The 30% we’re outlining here could get them past it.

**

Shifting gears, we wanted to compare our lower bubble to John Templon’s. We’ve been spending a lot of our words on the upper bubble. But the NIT field is a two-front war.

Part of why we don’t discuss the lower bubble as much is that our model’s NIT selection criteria is much simpler than its NCAAT selection criteria. It isn’t worried about Q1/Q2 wins, nonconference SOS, or how close a team is to .500 (provided they aren’t below it). It’s a lot simpler, and it still performs a good bit better than its NCAAT counterpart in backtesting. The NIT committee operates with much less time than the NCAAT committee. I think this leads to a less complex, more consistent, more predictable rubric. Famous last words, I know.

I say all that to say: John had Rutgers, San Francisco, and LSU in his projected field on Monday instead of Loyola, Yale, and Maryland, who were/are in ours, and I don’t necessarily disagree. It’s a difference of our formula versus his judgment, and his judgment’s usually really good. Two teams worth talking about here, though:

He mentions Maryland as facing the risk of a sub-.500 record. We see this risk too. They’re the teensiest underdog on Sunday up at Penn State. By sitting at 49.6% likely to win, though, or something like that, they still get a median scenario of finishing above .500, because while the median leaves them starting the Big Ten Tournament at 15–16, there’s better than a 0.81% chance they win two games next week. There is also, importantly, still a chance Iowa or Ohio State manages to escape the NIT. If that happens, Maryland currently leads Minnesota by three spots for the Big Ten’s second NIT automatic bid.

In short, I see why our model likes Maryland, and I see why John doesn’t, and I think it’s going to work itself out.

Yale is interesting, too, because despite sitting in NIT at-large territory in their median scenario, our model only has the Bulldogs 1-in-4 likely to end up in the NIT. The issue here is that when our model looks at “median scenario,” what it’s really looking at is median NIT selection rank and median NCAAT selection rank, and then the same thing again but with seeding scores rather than selection scores. Importantly, when it looks at median NIT selection rank it doesn’t isolate for instances in which Yale loses during the Ivy League Tournament. So, the top 35.8% of Yale’s selection ranks in our model’s readouts are instances in which Yale’s already made the NCAA Tournament, leaving the median skewed higher. This is intentional on our part—Yale is aimed at being the 81^st team nationally in NIT selection rank, and we want to show that—but it creates some confusion. Ultimately, we don’t think Yale will make the NIT. We do, however, see that in most situations in which they miss it, they or Cornell or Brown knock Princeton in.

In other words? If you’re a Rutgers, San Francisco, or LSU fan, I think you should be encouraged by John’s bracketology. If you’re a Loyola fan, you’re around 50/50 in both John’s eyes and the estimations of our model. If you’re a Maryland fan, you need to finish at or above .500 (or catch a break via Iowa) for us to even have the discussion. If you’re banking on Yale not making it, don’t, because if they don’t it’ll probably mean Princeton does.

For the Archives