It's February, so bracketologizing and bubble watching are getting into high gear. (If you're keeping track, it's the fourth-most wonderful time of the year according to my seasonal analytics.)
One of the more interesting résumés at the moment is North Carolina's. For the most part, UNC looks like a team that is directly on the bubble:
But there's a glaring goose egg sitting there: 0–7 in the almighty "Quadrant 1" games. Sure, they're 4–0 in the semi-mighty "Quadrant 2" games, but the punditry is unanimous: Carolina just can't dance without at least one Quad 1 win.
I've got some quibbles with this Quad 1 absolutism.
First, it's really not unprecedented for a team with this type of résumé to make the tournament. It's true that no sQuad Zero has made the tourney since the invention of explicit quadrants in the 2017-18 season. But that's just three tournament selections, and for the record there were three teams invited with just one Quad 1 win in even this short period:
Nevada even got a 7-seed with just one Q1!
Yet just as three is greater than two, one is greater than zero. Perhaps infinitely so. This gets me to the real meat of my Quibble 1: the quadrants existed in spirit looooong before 2018. If you were alive back then, you will remember that the selection committee focused on something different but really the same: top 50 wins and top 100 wins. I can tell you with absolute certainty that the committee used top 50 RPI wins the same way they use Quadrant 1 wins now. The only difference was that our beloved quadrants flex for home and road, as they should. The quadrants were a big improvement, but were just a tweak of the familiar paradigm. For example, here's the 2015 Selection Sunday "Nitty Gritty" sheet showing the proto-quads (with a spoiler alert):
This means we can safely look back before 2018 to get actionable intelligence about the plausibility of a 0Q1 team making the tournament. What we find is that there is indeed precedent for teams making the tournament with zero top line wins:
One other thing you might notice about each of these teams is that they all had multiple bad losses (outside the RPI top 100 in those days) and still got in. Anyhow, unsurprisingly, three of these four teams show up in the default list of ten most similar résumés for North Carolina 2022 on my site:
Georgia in 2015 is almost an exact match, particularly in the categories that matter most. There you have it. It's been done. It could happen again.
On to my second, more fundamental, quarrel with Quad 1 absolutism: Quadrants are kind of dumb! Which would be okay if we needed them, but we don't!
Long, long ago, in a time before the NET, I wrote on this very blog that:
Alas, the NCAA did not do this. First, they invented the home-road affected quadrants, and then they replaced RPI with NET. Both of these changes made the selection process better. I'm a fan of NET, relatively speaking. For judging the quality of a team's wins and losses, which is what they are basically using NET and the Quadrants to do it works pretty well.
But what they really should do is use the NET to calculate an official NCAA version of WAB or Strength of Record. I think they shouldn't even publish the NET at all—just use it behind the scenes as the backbone for the simple calculations required to turn it into a strength of record ranking.
They should do this because there are more than four levels of opponents in NCAA Division 1 basketball. In fact, there's probably at least 100 levels. Pretending like there are only four different kinds of wins and losses might make sense if we didn't have good ways to rate teams and didn't have computers to do calculations for us. But thankfully we have those things, and lots of them. We can do this. We should do this.
To bring this back to UNC, we know that the Tar Heels would be in much better shape under any system that more rationally quantified the impressiveness of their record. We can see that they're a lofty 26th in ESPN's Strength of Record, which is one attempt at the kind of system I'm arguing for. They're a solid 35th in my own superior T-Rank-based WAB rankings. They'd still face scrutiny because of their particularly heinous performances against the best teams they've faced, but at least they wouldn't have to face the (false, actually) argument that inviting them would be totally unprecedented.
Glad the blog lives. I wish I spent more time thinking and writing about sports. Some day.....
ReplyDeleteLook forward to your return to the blog whenever you find the time in between cruises and vacations and all that other stuff you do.
DeleteThanks for the post! I had the false assumption you address prior to reading - partially because I was listening to Norlander on the CBS podcast. It's nice to be shown wrong with actual evidence. Of course, there's a difference between "the committee *does* not invite teams with 0 Q1 wins" and "the committee *should* not invite teams with 0 Q1 wins." It should be important distinguishing between the two for bracketologists or commentors.
ReplyDeleteRegarding a theoretical transition of the committee to using WAB, what kinds of adjustments might you find reasonable if the committee still wanted control the process. Your WAB measure ranges from -1.4 to 6.8 for the top-75, so there is some room for semi-manual adjustments without drastically changing team positions. Would you shrink negative WAB from losses due to player unavailability? Maybe a constant negative for super weak non-con SOS? A best-win or worst-loss multiplier? Or would you go full Gasaway and dismiss the committee for an unadjusted WAB bracket?
Good questions. My ideal would be no committee, because committee's aren't sports. We accept the primacy of win/loss standings for post-season selection in most sports, I think we could get weaned off this idea that humans are a necessary part of the process in college basketball.
DeleteI also oppose adjustments for injuries. If you lose a game because someone is injured, or suspended, or whatever: thems the breaks. We don't think an NFL team should get an extra win in the standings because they lost games while their QB was injured. Why should the NCAA tourney be any different?
One thing you could do easily with a WAB-style system is basically zero out wins at some level of mismatch. So if a top 10 team is playing #310, instead of giving them credit for .03 wins like WAB might do, just say it's zero. Doesn't make much of difference, but provides a tool to clearly incentivize against playing terrible schedules if you want to be in consideration for an at-large. I actually don't think this is really necessary -- I think basic WAB does a perfectly fine job, but if people are dissatisfied with edge cases of teams that build a good WAB ranking by going undefeated against lesser teams, I'm sure there are simple but clever ways to reduce that possibility.
All that said, back to the real world. The committee is not going away. But I do think a plausible path forward is if they can be convinced to make a WAB or SOR type ranking the primary basis for evaluating teams, then they'd basically be put in the position of having to explain deviations from it. Right now they've got so much wiggle room.. "NET is just a sorting mechanism, but we consider it. We look at Kenpom etc. but just as a guidelines..." There's no one ranking of teams that is recognized as the default. There should be, and it should be WAB.