This hit the feed today and caused a spiral of google searching and squinting at dense technical stats articles:

Introducing Bayesian Adjusted RAPM for NCAA! Norvell/Hunter top the list in 2019 - see more @tothemean https://t.co/z15ybWpr7Z pic.twitter.com/WtslRbKqXz

— Jesse Fischer (@jessefischer33) May 1, 2019

Naturally, you are thinking "what the hell is that?" if you didn't immediately log off. If this stat (BARAPM) adds any more components it's going to sound like you're firing a machine gun whenever you mention it.

[AFTER THE JUMP: wonkery and then an Isaiah Livers takeaway]

After the squinting this is my take:

• It starts with plus/minus. That's the PM.
• It adjusts for who is on and off the court with you. That's the A.
• These results are then "regularized," which is a mumble something regression uh… ok. It's a mathematical technique that's supposed to reduce the problems induced by substitution patterns* and make the overall data less noisy. This is the R.
• The BA takes the regression in the previous bullet and changes the target from a league average player to a combination of tempo-free stats and overall team quality.

*[ie, if two bench guys share most of their minutes it's tough to determine which of them is causing the difference between their play and that of other lineups]

After going through all that and looking at BARAPM stats on O and D for several years I have the following conclusion: it's still a plus-minus stat that is dependent on your teammates. The regression does throw players on good teams up high, but that's kind of defining good as good.

For instance: Texas Tech lands five guys in the top 50 of D-BARAPM this year. None of them are named Tariq Owens. They're Matt Mooney, Davide Moretti, Jarrett Culver, bench wing Brandone Francis, and Owens's backup Norense Odiase. This makes some sense since Texas Tech's defense is actually a bit worse with Owens on the court:

But a big part of that is Odiase, the 6'9" 250-pound senior who split minutes with Owens at the 5. If Owens, the starting C on the best D in the country, was backed up by  a numpty he'd rocket up the list as the team got worse.

The best examples of this in the top 50 who are familiar to Big Ten fans are Kaleb Wesson and Isaac Copeland. Neither is a standout defender. Wesson has a block rate of 2.9% as a ground-bound center; Copeland was an all right switchable four. They're on this list because of what happens when they go out: for much of the year Wesson got replaced by his 6'6" brother, and when Copeland got hurt Nebraska put in Tanner Borchardt and lost their switch-everything ability.

So I think this stat is potentially useful, especially on defense where stats are hard to come by, but it can't be used as a straight up ranking of players. Team context is too important in any plus/minus stat, and this is still one. That said:

SOME MICHIGAN STUFF

Your #1 most dude in D-BARAPM last year:

Teske was also 9th last year. Context here is important, obviously. Last year Teske was backing up Moe Wagner. This year he was backed up by a rotating cast of short guys, guys who can't jump, and skinny guys. We want Teske to fall back down the list because Colin Castleton becomes an excellent sidekick.

There are two more Michigan players in the top 50, but you'll need three guesses to get them. X: yes. Matthews: no. Livers: yes.

This is impressive, and very good news as Michigan seeks to keep their defensive efficiency up. Almost a third of Livers's possessions came as a small-ball center, where he was worse than the other options. (On average.)

For Livers to make the machine gun stat list of top defenders he had to annihilate it on the other two thirds of his possessions. The vast majority of these (all but ~100) came next to Jon Teske, and yup:

Livers is enough of an upgrade on Iggy defensively to expect he'll be able to more or less replace Matthews. In fact, lineups with Teske/Iggy and one of Matthews/Livers gave up an identical 0.84 PPP. Matthews is a better defender, in all likelihood. Livers is on this list because the guy he replaced was a meh defender. Matthews is not because the guy who replaced him much of the time was Livers. But there's every reason to expect Michigan will have three elite defenders in the starting lineup again next year.

Unfortunately only 10-15 of those minutes will be new Livers minutes; Michigan will have to find a bench player to step up if they're going to maintain their level from last year.

TEAM CONTEXT

Michigan's six-man-rotation and cliff is easily observable in the team data:

A few points of interest:

• Charles Matthews might not be that hard to replace given the Livers items above and his overall negative impact on  Michigan's offense.
• Eli Brooks will be a solid bench player if he can shoot all right.
• Castleton scraping above zero on defense despite playing as Teske's backup is impressive. Limited time, obviously, but early returns are encouraging. He should be a plus defender.

This has been "can we make this stat weirder." Thank you for getting to the end.