📉 📊 📈 Stats Nerdery

Men's Basketball
41

It’s a really complicated problem to address relying on public-facing data. This article gets into it:

I like this:

For each player, the evaluation is really a three-step process:

  1. What is he being asked to do?
  2. How well does he do it?
  3. How much value does he provide relative to role?

None of these are easy questions, and we’ll start with the last one first. At this point it seems reasonably well-settled that players on the larger end of the position spectrum provide more defensive value than smaller, because being large and in the way of the basket has proven to be a fairly major part of modern defense. Big surprise there. Beyond that, it’s hard to contextualize value across roles at this point without resorting to catchall metrics, which we’ll get to.

Article goes on to try to answer “what’s is he being asked to do?” by identifying how frequently / player guards a team’s primary option (defined by usage rate). And then gets into the “how well does he do it” by looking at pros and cons of different metrics.

Honestly, my takeaway is that a coach watching film and knowing what his scheme is supposed to accomplish will probably have a noticeable edge on any of the existing defensive impact models when it comes to accurately evaluating his players’ defensive abilities. And especially if they can also marry some selective stats-y things like percent allowed with [player] as closest defender with their scheme knowledge.

3 Likes
42

Will check that out

43

This has a lot of good quotes too:

No one likes the defensive metrics:

To get a general sense of the state of these stats, we surveyed more than a dozen NBA analysts, comprising both team employees and public writers, most of whom used to work for teams. On a scale from 1 to 10, they rated the quality of public defensive stats as just a 3.6, on average. Their perception of teams’ defensive metrics behind the scenes wasn’t much better.

Defensive rebounds have issues analytically:

Defensive rebounds are “noisy to the point of being very nearly useless,” Myers says. That’s because they’re largely contingent on a player’s role rather than skill; only 24 percent of defensive rebounds leaguewide are contested, per tracking data, and there isn’t a great relationship between a player’s individual rebounding stats and his effect on his team’s rebounding stats. Andre Drummond, for instance, is the best defensive rebounder in NBA history, by percentage. But Drummond’s teams have often allowed fewer offensive rebounds, and fewer points, with him on the bench.

Steals and blocks don’t happen often enough:

This season, a block or steal occurs once every eight possessions leaguewide, meaning each player averages just one block or steal out of every 40 possessions. That leaves 39 out of 40 possessions in which the defender in question doesn’t record either statistic, but he’s still contributing, or not contributing, somehow.

The data can’t capture everything relevant:

Analytics can capture what happens when a shot goes up, but more often falter when considering every other part of a possession. “It’s all the times,” the NBA personnel executive says, “where, did my center show effectively and then recover, and that bought us an extra half-second for the weakside corner guy to get back out to the shooter, which then led to a good defensive possession?”

And then there’s scheme:

“Even if we had all the data we wanted,” says ESPN’s Kevin Pelton, “I don’t know if we’d ever be able to isolate an individual’s impact as easily on defense as on offense, because so much of it is scheme-dependent.”

Until Second Spectrum’s tech can figure out a way to measure intent—as in, what was the defender supposed to do?—this massive facet of defense will remain missing from all the numbers.

1 Like
44

I mean, not all is lost. You could also just pair all of this with a deeper qualitative study at a few intervals and get what you need to make choices. You won’t have a precise measurement but it’ll be enough to go on to make decisions.

Sincerely,
qual guy

2 Likes
45

Almost a week old now, but interesting stats-y rumination on how good the Big 12 is this year.

2 Likes
46

This could probably go in a lot of threads, but I missed the selection committee naming their top 16 seeds as of right now. Note, this was released yesterday, so I assume it accounts only games up through 2/16. Purdue gets the kiss of death, apparently.

This should both somewhat satisfy the people who dislike the NET (there are some interesting divergences) but also raise questions like “wait, what exactly are their seeding criteria?”

1 Like
47

But it’s encouraging for ACC teams, particularly UVa and Wake, that UNC is a #2, Duke a #3 seed and Clemson was one of the teams that was considered for the last #4 seed. I don’t think that is consistent with the often expressed (by writers) idea that the ACC is a weak conference and only 3 teams will get bids.

2 Likes
48

An interesting read. I didn’t find his arguments in defense of the Big 12 being the best conference particularly compelling. It leans heavily on the efficiency metrics, which are predicated on the assumption that, for example, a home win against Miss. Valley State can be compared to a road game against Purdue, apples-to-apples. They’re about the best that we can do, but I think it’s silly to use efficiency metrics as evidence that a mid-Big 12 team would be anything other than mid in one of the other Power conferences.

4 Likes
49

The committee rankings actually line up well with the Strength of Record metric that’s on the team sheets, which is a resume metric (results-based in the NCAA’s parlance). That explains Purdue at #1, Houston at #3, Marquette and Duke being higher than they are in the NET, BYU not being in the top 16. Doesn’t line up 100%, but perhaps it’s a bigger factor for the committee than people commonly talk about.

(SOR is good for us too)

SOR standings here: 2025-26 Resume Men's College Basketball Power Index - ESPN

They helpful have a column for “SOR Seed” which is basically what would happen if you used SOR as the only selection and seeding criteria. A fun look at alternative universe where we put the selection committee out to pasture (not the bad metaphorical kind, but an actual pasture where they can have a picnic instead of a meeting on Selection Sunday):

Things of note in the SOR tournament alternative reality:

  • Big 12 still has 10 in by SOR, but with a couple of teams hanging on by a thread. And BYU a little closer to the edge than one might think just looking at their strong predictive metrics.
  • MWC only has 4 in, and their 4th team (Nevada is just hanging on). In general, SOR handles resumes mostly built off of in-conference win-trading better, I think (this applies to B12 too).
  • Strong midmajors that have done little wrong but just don’t have Q1 games do well in SOR (and similar results based metrics): Grand Canyon, Princeton, JMU. This is one of the things I like best about using SOR as the selection criteria. Feels fairer.
  • We are a 7-seed (actually one spot above Clemson on the S-Curve). This is the selfish reason for preferring SOR.
  • ACC only has 4 in, but I’m guessing Wake would have potential to improve enough to get in given their remaining schedule. Not too different from the current reality, besides Syracuse being closer to the bubble than they are commonly discussed now.
3 Likes
50

I believe David Worlock told Eamonn Brennan that SOR is much more important to the committee than SOS, so Eamonn’s swapped SOR into his Bubble Watch columns. It may be that SOR has an even higher priority than that.

2 Likes
51

Yeah, I saw that in a past Bubble Watch; I think that would be a significantly positive development!

1 Like
52

Maybe … depends on next 4 game results.

53
  1. I didn’t know ESPN’s predictive metric (BPI) adjusts for altitude (essentially factoring it as extra home court advantage that translates into discounting a team’s rating some).
  2. I didn’t know they adjusted this much for it.
54

You could also say that ESPN isn’t sold on the MWC for reasons other than elevation (a lot of those teams with home arenas <= 1,500ft are from the Mountain West). Said otherwise, BPI might be biased towards P6 conferences, but it’s another one of those black box metrics, so maybe elevation is strongly weighted for whatever reason?

2 Likes
55

Same old same old with NET and non-conference scheduling:

https://richmond.com/sports/net-rankings-acc-big-12/article_eba150b6-d0d7-11ee-abe7-d7f0618df6c9.html

I do think the coaches should probably try to not give off the air that they want to get special bonus points for being the ACC. The critique of how well the strength-of-schedule adjustments in the NET are or aren’t working can stand alone and is a stronger point, in my opinion.

There’s also talk from some coaches about reevaluating how they schedule. I think there’s a collective action problem, in that everyone scheduling (and getting blowout wins in) more Q4 games would make it more possible for conference games to be a resume builder for the teams, but the benefit is mostly in terms of making yourself a more valuable win for other conference teams to pick up. It’s probably better for an individual team’s resume to play a harder schedule with more Q1 games and that benefit is all going to you.

3 Likes
56

My issue with the NET (and most of the efficiency metrics) are that they’re really sensitive to blowouts (outliers). Depending on how literally the NET other efficiency metrics are used, this incentivizes teams to schedule games against crap competition and keep their foot on the gas pedal when they’re up big. So basically we’re at risk of creating a system that “encourages” basketball games that no one really wants to watch. I don’t care how Top 25 teams perform in buy-games (unless they lose), and I don’t particularly care to watch one team run up the score on another so the winning team can pad their NET ranking.

4 Likes
57

Agreed with the slight modification. You don’t want to play the absolute dregs and blow them out. But a healthy dose of teams ranked between 150 and 250 with 20 point wins is the path to a lofty NET/KENPOM/ETC. Just look what the big 12 and mountain west (who famously regulated schedules to play the rpi game a decade ago) are doing.

1 Like
58

Certainly true of the BIG12, but the MWC collectively has had a good OOC strength of schedule this year, at least according to Torvik. So, there is the extent to which you could game the system and your success at doing so further obfuscates the original gaming and this whole enterprise becomes a self-licking ice cream cone but :person_shrugging:

59

I was thinking about this the other day. Efficiency metrics are both “Team X is 3 points better than Team Y on a neutral court for the various reasons” AND “Team X is 25% likely to exceed expectations by 7 or more points, while Team Y is only 20% likely to exceed expectations by 5 or more points”. And absolutely annihilating a team helps with both aspects.

60

I’m not sure that that’s quite the correct. Sorry in advance if this is super pedantic, and here are the links that I’m using to draw my conclusions: KP and Torvik.

Starting with a single game:

  1. Calculate the points per possession (PPP), both offensive and defensive, for that game.
  2. Scale the raw PPP up or down based on how your opponent’s efficiency compares to the average D1 team.
  3. Scale the outputs from Step 2 based on game location.

That gives you the efficiency contribution for a single game, the Offensive and Defensive Efficiency Metrics just take the averages of all of those games. It reads to me like KenPom just subtracts your average Offensive Efficiency from you average Defensive Efficiency, and then ranks teams based on that net figure. Torvik applies some exponents and does some division to calculate the score that he uses to rank teams.

I think KenPom and Torvik are simpler than people think (it’s basically all about how the results of a game or series of games affects your average efficiencies), and the calculations for Evan Miya’s rankings are probably more complex than you’d expect.

1 Like