clock menu more-arrow no yes

Filed under:

Michigan State Spartans Basketball: 2014-15 Box Plus-Minus

New, 18 comments

Joe reviews some Box Plus-Minus numbers and shares a couple scatterplots!

Mark Konezny-USA TODAY Sports

Plus-Minus

Over the course of the year I'd noticed some chatter in the comments about plus-minus data and I thought it'd be interesting to revisit the topic again, this time for the 2014-15 season.

About a year ago, I wrote a primer on Adjusted Statistical Plus-Minus and gave some caveats for this analysis. It's important to note: adjusted statistical plus-minus (ASPM) and box plus-minus (BPM) are the same things. The bullet points:

  • Plus-minus is the Holy Grail of stats, because it captures everything
  • Because plus-minus catches everything, the data is noisy
  • So noisy, in fact, that viewing raw plus-minus doesn't say much
  • In the NBA, it is possible to do regression analyses on raw plus-minus data, eliminating noise
  • In the NCAA, there are too many teams and too few games to regress upon
  • We can use an equation to guess what a player's regressed plus-minus score is, based upon box score stats
As you might expect, using only box score stats to predict players' value does yield some unexpected results. Basically, you're guessing at the hidden value a player creates based upon what the scorer is able to record.

Another note here. BPM is representing points per 100 possessions for which the player was on the court. So it's minutes independent.

B1G Plus-Minus in 2014-15

Awesomely, Sports Reference now has BPM for all seasons going back to 2010-11. Below, I've grabbed the BPM for all Big Ten players who appeared in at least 500 minutes this season. That roughly corresponds to 40% of available minutes on a team which did not make the postseason. I've charted Offensive BPM (OBPM) on the x-axis vs. Defensive BPM (DBPM) on the y-axis. I've also moved the axes to reflect the average OBPM and DBPM for the Big Ten. Those values are 2.72 and 2.29, respectively (which makes Kendrick Nunn the most average B1G player). I'm guessing the differences are due to the fact that good offensive players are more likely to reach the 500 minute barrier. Anyways, without further adieu:

2014-15 B1G OBPM vs. DBPM

There are a few takeaways here. First, there's essentially zero correlation between offensive and defensive value as measured by BPM (it's actually very slightly negative). Second, Rutgers is bad. They had 4 of the bottom 9 players per BPM. Third, Jordan Dickerson for Penn State ends up being an outlier here, mostly because of his high block rate while he was in the game for Penn State. Despite that BPM still mostly account for this overall value.

Here's what you really want to know:

Michigan State BPM for 2014-15

[Editor's Note: Javon Bess is in the top left quadrant, he's the light blue square. His entry in the legend was cut off for some reason]

And for the data in table form head here; this page has a bunch of great information and is more interactive.

There are some surprises here.

Matt Costello is valued highly by BPM, and I can see why. First, he was the most efficient offensive player for MSU, tallying a 119.5 offensive rating. Second, he generated a high number of offensive rebounds, while not turning the ball over very often. These are all good signs, pointing to a player who is playing within himself and making teammates better. Defensively, BPM likes a high block rate (as shown by Dickerson above) and Costello also rebounded the ball well defensively. Overall, there's a lot to like, and Costello is pretty much the perfect veteran big man to mentor the freshmen coming in, while picking his own spots.

Tum Tum Nairn, predictably, does poorly by this metric. True adjusted plus-minus I believe would be much more kind to Tum, but because BPM is limited by the box score, he looks pretty bad. I'm afraid there's truly no way to accurately measure his value to the team.

Travis Trice doesn't look great defensively, but I believe that's simply because he was asked to do so much offensively. It's tough to be a high usage rate guy and also give maximum effort on defense.

...except Denzel Valentine looks quite good on both ends. I believe that BPM likely places additional value on offensive rebounds, which makes Denzel more valuable than Trice offensively. He also shot the ball better than Trice, who was forced to create for himself more often than was ideal. Still, Denzel's effectiveness on both sides of the ball, despite an increased workload, bodes well for next season.

Marvin Clark looks solid offensively with some work to do on defense. I believe that Marvin's lack of playing time for stretches during the season were mostly due to his defense, and not his penchant for taking long jumpers. I feel like Marvin is going to play a role on next year's team, but I'm not totally sure how just yet.

#SWollenman rated as the team's fourth-best defender. That checks out. Alvin Ellis and Gavin Schilling were also plus defenders, though Ellis was a disaster offensively. That also checks out. Bryn Forbes was not much of a defender for much of the year, but his offensive prowess was hugely helpful. That also checks out.

I think BPM also gives us a little bit better look at what Eron Harris might bring to the table next year, as well (from S-R):

Eron Harris

The glaring issue here is defense, and that makes sense with what I've heard about Eron's time in Morgantown. Harris' OBPM from his final year at West Virginia would make him the third best offensive player for MSU last year, and he's had a year to sit. But the difference between Eron and Bryn might be a little bit less pronounced than I expected prior to this analysis. Given Izzo's penchant for playing guys who will give defensive effort, Harris' improvement on that end over the past year will be crucial. That said, Harris can clearly soak up possessions in a way that Forbes has not been able to do, and that might be just what this MSU team needs.