Old Spice Classic: Plus Minus Data

Dylan Burkhardt

I experimented with plus/minus data last year in an attempt to go a little more in-depth with the analysis around here. The numbers appeared to be a little too abstract to do much with but they are set to make their triumphant return, at least for right now.

Frequent reader and commenter, gpsimms, volunteered to attempt to track plus/minus data. Here is the data he came up with and some bulleted analysis after the jump. (Note: I know minutes are a little off but they should be close)

Columns explained: Minutes played, net points while on the court (scored minus allowed), points scored per 40, points allowed per 40, net per 40, normalized (net/sqrt(minutes)). Column headers are clickable to sort.

Player Min Net Score/40 Allow/40 Net/40 Norm.
Lucas-Perry 84 13 75.71 69.52 6.19 1.42
Novak 91 5 74.73 73.41 2.20 0.52
Douglass 63 3 77.46 75.56 1.90 0.38
Harris 113 -3 72.57 73.63 -1.06 -0.28
Gibson 23 -1 80.00 81.74 -1.74 -0.21
Sims 96 -6 70.83 73.33 -2.50 -0.61
Vogrich 18 -5 77.78 88.89 -11.11 -1.18
Morris 69 -28 67.25 83.48 -16.23 -3.37
Wright 25 -25 62.40 102.40 -40.00 -5.00
Akunne 1 -2 0.00 80.00 -80.00 -2.00

  • Laval Lucas-Perry: The first guy who stands out. Not only did Laval score the ball well this weekend, it appears that he played good defense. Kudos to Laval for the good weekend.
  • Darius Morris: Yikes. Not the finest weekend for the freshman, he’s obviously still adjusting but judging by these numbers, he’s got a ways to go.
  • Anthony Wright: This isn’t really surprising, defense goes way down hill when Anthony is in the game. The offense is brutal too, at this point I’d be all for Gibson and Vogrich replacing a lot of his minutes.
  • Zack Gibson: Gibson scores points when he’s in the game, that much is clear. He also isn’t scared to shoot, taking 28% of Michigan’s shots while he’s on the floor. For those of you who think that Gibson might be the answer on defense, the data doesn’t quite support that.
  • Zack Novak: He typically grades out well in the plus/minus data which goes along with the argument that he does all the little things.
  • Manny Harris: Harris’ numbers are obviously not going to show much because he played almost every minute of every game in Orlando.
  • Stu Douglass: Stu hasn’t been shooting the ball well and drew a lot of criticism in the comments. The plus/minus data explains a little bit why he might have been on the court late in games.
  • DeShawn Sims: Not his best weekend and he needs to be more assertive on offense. Curious to see how he comes out on Wednesday.

I’d love to hear your thoughts, comments, and criticisms about this data as well as plus/minus in general in the comments.

  • Eric

    I like the plus minus, but a players effectiveness depends more than the individual players. I feel that a better way to track than an individual’s stats (although much more time intensive) would be a rotation of who’s on the court at any given time.

    This way, we could see who works together well. For instance, the defense may suffer with Gibson instead of Sims in the lineup, but if both played at the same time, how does that effect the team?

    This would allow us to compare who works well together and determine what a rotation should be. We could also go head to head with players on who should play without downgrading them for playing with the second team etc.

  • gpsimms

    We do have that data, Eric. For the whole tournament, Gibson and DeShawn only logged about 3 minutes together. In those three minutes, we went +0. It’d be an interesting thing to tinker with, but with the sample size we have there is nothing to say.

    I can say I understand why Beilein would worry about that lineup, especially against Marquette. Gibson is a real liability against such a small, fast, team. Especially because he’s replacing Novak, who (above) was one of our best defenders in the tournament.

  • Eric

    I’m not saying that we should have Gibson in the lineup with Sims. I just meant that if we were to compare any 2 players, we should look at them by substituting one for one, ie look at LLP vs. Stu while playing with Harris, Sims, Novak and Morris.

    I also agree that the speed of Marquette would have hurt us with Gibson especially when the opponent was hitting threes.

  • With the lineup data, everything besides a couple lineups becomes very fractured… Dealing with a couple minutes for each lineup. Making it hard to do much of anything with.

  • Acedeuce

    Brilliant work as always with the plus/minus stuff. I find this sort of in depth quantitative examination interesting (at least compared to MSU getting mauled by UNC).

    It will be interesting to see how Morris improves over the season because if he doesn’t improve soon, it may be in the best interest of the team to limit his time. What about this lineup, LLP, Novak, Harris, Sims, Gibson. This would negate the loss of Novak when Sims and Gibson play together while giving us seemingly the best offensive/defensive team. Stu backs up the two guards to end up with a bunch of minutes, Wright backs up Harris leaving him near 0 minutes, and Cronin(!?) backs up the bigs. Morris I guess would get some minutes both at the Guard and Manny’s “3” position.

  • gpsimms

    I’ll add some thoughts as they come to me:

    Darius: You can see he had a rough weekend and the game tracking statistics really bear this out. Actually, you can see how rough his weekend was just in the box scores: 5 ast, 8 turnovers in 3 games. He struggled finishing. The end of the Bama game, all that. But, the kid’s a freshman, and I think careful analysis of the game flow show that this team’s ceiling is VERY dependent on Darius Morris’ progress.

    My favorite 4 minute stretch in tournament was the first 4 minutes of the second half of the Marquette game. Marquette is a fast athletic team, with which we can only keep up if we have our best athletes on the floor. For that stretch, the 1-3-1 was clicking and caused 3 turnovers in a 7 possession stretch. Michigan logged a +8 during those 7 possessions.

    The opening 4 minutes of the Bama game was pretty similar (6 turnovers in the opening 7 minutes!!!). Like any basketball game, there are hot stretches all over the place. But Michigan’s best combination of intense turnover causing defense with speed on offense to score in transition is when Morris was on the floor.

    The numbers look ugly, but I’m excited about what level this team can rise to as he gets his head around the half court sets a little more.

  • UMIndy

    Great work on these numbers. The economatrician in me would respond to the line up criticism by saying that in large enough numbers (i.e., over the course of the season) I would expect that who the guys are on the court with will randomize itself out of the equation to some degree and that even with these individual numbers we can meaningfully compare players’ impact on the game.

    One could also criticize these numbers because they do not account for who the player is playing against. That is, was he on the court against the other team’s best players or their 2nd and 3rd string. But again, with large enough numbers, that shouldn’t be a problem.

    Is this a perfect system? No, but few measures are.

    Again, good work guys!

  • oldblue

    I love the +/- stats! I do not understand the use of the square root in the last column, though. Can you explain why that was done? The net/40 seems like the best stat, especially when more games are added. Please keep it us. I was quite surprised by some of these, and I hope to watch them as the year goes on.

  • Someone suggested it last year. The idea is to normalize for the players who play very few minutes.

  • gpsimms


    The main thing about LLP’s plus minus can also be found in the standard box score. He was helping the team as much as he was because he was the only one who was shooting from deep at an acceptable clip. (Apparently ‘acceptable’ around these parts is 50% from downtown. Alright: call it nasty)

    The team needs a scorer to help out Manny, and LLP could be the guy. The problem is that he is out of control as a ball handler (he kind of reminds me of Manny as a freshman. Head down, run faaaast, turnover.) There was a play right around the 12 minute mark of the Marquette game, in which Laval had a lazy pass which led to a Marquette breakaway and then he took a dumb foul.

    Beilein promptly sat him for the rest of the half. Beilein is a teacher, the benching probably hurt the team in the short term, but as we know from the Iowa game, if Beilein sees a teachable moment he will use it.

    Being benched for a long chunk (12 minutes) of the Marquette game is best move for a guy’s individual +/- a man can make, so that is part of the reason for his inflated numbers. Of course, we can argue his benching was the reason for the -7 point stretch over the 12 minutes.

  • I believe that was his second foul around the 12 minute mark right?

  • gpsimms

    DeShawn: Peedi had a rough weekend as well. And compared to expectations coming in, he may have had the most disappointing on the team. He was outrebounded by shorter, more agressive big men all weekend (except the Marquette game). I have to remind myself that DeShawn has become more of a “skill” big man type than a “dominant athlete” big man type. I can usually tell the type of game to expect from Peedi early: if he asserts himself with his back to the basket and gets blocked, then he will shy away from the post the rest of the game.

    This is not a terrible thing: Peedi is shooting .367* from three on the year. But he helps his team most when he is being a versatile inside-outside threat, and, most importantly, helping Manny on the boards.

    I have to worry that DeShawn will struggle against teams with uber-athletic bigs (Sparty, OSU) and really shine against teams with “skill” type bigs (Illinois, Purdue, Minnesota, etc.)

    *-.367 for the season, but 1/9 in the Florida tournament. He’ll shoot somewhere between the two, obviously.

  • gpsimms

    Yeah, you’re right. That was a rough 18 second stretch. So nix the ‘teachable moment’ stuff.

  • ZRL

    Great stuff!

    Even though I am not the biggest Anthony Wright fan, I will defend his numbers by pointing out that he is mostly in the game as a sub for Manny. Of course the team is going to be worse when Manny is on the bench. If anything, Wright’s +/- numbers are just more proof as to how valuable Manny is to this team.

  • Tweeter

    Good work guys! Its tough to take much of anything from such a small sample size. LLPs numbers look good because he scored well when he was in the game and he didnt seem to be in the game that much when a lot of the other starters were sitting. Whereas Wright was in the game a lot of times with other subs. That said you dont need +/- data to tell you LLP is playing far better than Wright.

    I have been a little befuddled by the substitution patterns Beilein has been using so far. I have noticed that several times over the course of these games that we end up with a lineup combo including Douglass, Vogrich, Wright, Gibson and then someone else or Manny. I dont think we have enough quality depth right now to play our subs together (although Douglass is basically equal to a starter). I would rather see a pattern where only on or two subs come in, then at the next sub insert the starters back in with the next two subs. Therefore eliminating lineups where you have mainly subs out there.

    I dont know if it actually cost us that much in points in these games, especially considering it was usually only for a short stretch, but it certainly kills any momentum we might have had. I imagine Beilein is largely experimenting right now to find out what works and I trust by the end of the year our rotation will be more sorted out. But as of right now I dont think you can afford to have Vogrich and Wright out there together.

    Obviously it is a long season and you have to keep guys fresh, but in big games I would much rather see a short bench with the top 7 getting all the minutes save for a couple in each half to Vogrich and Wright. Preferably at separate times.

  • mark

    Wright numbers don’t surprise me at all. He is usless on defense and it seems the other team scores every time he is on the floor.

  • TomAce

    Hey, how ’bout that ACC Challenge…now a best of 5…
    And check out Northwestern! Beating undefeated NC State on the road! How good is this team?!

  • i am bill

    Love the +/- Dylan. Keep it coming.

  • maxwell’s demon

    On the plus side are assts factored in?

  • No. It’s just net points scored while the player is on the court.

  • maxwell’s demon

    Thanks. I was envisioning more of a mgoblog ufr type deal but I like this better. Far more objective.

  • Nick

    Good data; thanks for the effort. The thing I think many people aren’t talking about with Morris is his length. His ability to play at the top of the 1-3-1 while playing the pg position is critical for the long-term success of this team this year. That allows LLP or Stu to play at the bottom of the zone. Remember how many times last year the pg (someone like Grady or Merritt) played the bottom and they would give up put-backs and multiple rebounds. He’s young and didn’t shoot well but his length is important in Beilein’s zone.

  • Andy


    I was the one that suggested the normalized data last year. It’s something I use when I prep for fantasy baseball and I’ve found it to be pretty helpful. Basically, it corrects for people that don’t have a lot of ABs (or minutes in this case). It’s kind of just an emperical thing that works for me. Quick example…

    Think of a player who was on the floor for four minutes, during which time his team made a quick 4-0 run. Consider a second player who played the entire 40 minute battle and the team pulled out a 66-62 win. Now look at their +/- numbers in different ways.

    Using a standard +/-, they’re both +4.

    If you divide by minutes, player one is +1, player two is +0.1.

    If you divide by sqrt minutes, player one is +0.632, player two is +2.

    There are advantages and disadvantages to each method. I don’t think they should be considered equal as far as impact on the game, and I don’t think that player two had ten times more impact in the game. So basically, the normalized thing is a compromise, for what it’s worth.

  • Andy

    EDIT: I got the sqrt one backwards. Player one is +2 and player two is +0.632. My bad, making a confusing thing more confusing.