NC Central Plus/Minus

Dylan Burkhardt
on

I have been playing around with ways to do some more in depth statistical post-game analysis. Something along the lines of MGoBlog‘s Under Further Review or 82games.com.  An entire play by play system I think would just be far too boring for the amount of effort it would take (but gives props to those who put in the work). There are usually 2-3 games every week and I wouldn’t have time to do an entire UFR for every game. I started to do some kind of UFR and then decided it would make a lot more sense to just read a play-by-play if someone wanted that kind of data.

In the end I decided the first thing I wanted to do was analyze what lineups worked and who had the biggest effect on the team’s success. What I came up with was a very limited +/- analysis similar to 82games.com. I am definitely planning to expand this feature but I really wanted to see what you guys thought of the idea and what I have so far. The result is simply the number of points scored when the player/unit is on the floor minus the number of points scored by the other team while they are on the floor.

Player +/- Minutes
Novak 31 15
Harris 23 36
Sims 21 34
Douglass 20 17
Grady 18 27
Merritt 2 13
LLP 0 23
Puls 0 1
Wright -1 3
Lee -3 4
Shepherd -3 13
Gibson -8 14

Obviously the first thing that stands out is Novak’s +31 in only 15 minutes. Novak was on the floor during both of Michigan’s big runs but he only had 3 points, 1 assist, 1 rebound, 1 steal, and 1 turnover. I think that Novak at the four definitely helps to stretch the opposing defense and really opens up the offense but I was still a bit shocked to see him at the top of the list.

Manny and Peedi as expected are next in line, they both played substantial minutes in a 20 point win so they naturally have a high plus/minus.

Douglass and Grady are right behind and they combined for only 7 points but also 5 assists and no turnovers. McCormick also pointed out a couple times when Grady had what would have been an assist but led to free throws.  Point guard minutes from here on out should be a lot closer to 30 minutes for Grady and 10 for Merritt. Merritt basically has no effect on the game (+2) while Kelvin posted a very solid +18.

The troubling part of this list are the three other guys at power forward besides Zack Novak: Gibson -8, Shepherd -3, and Wright -1. These three just are not producing. Gibson was abused on defense and missed three wide open three pointers. Shepherd had one good stretch with some follow ups but other than that didn’t do much. Michigan needs one of these guys to step up because this was a game where DeShawn Sims had to play 34 minutes. If Sims gets injured on into foul trouble that means Gibson will play significant minutes.

I was also surprised at Laval Lucas-Perry’s goose egg. Laval came out and rained three triples but after that was very unproductive. He missed a couple layups, a free throw, and another three point attempt. It is only his third game so I’m willing to accept the fact that this has to do with his transition at this point.

Lineup +/-
Grady-Douglass-Harris-Novak-Sims 20
Grady-LucasPerry-Harris-Novak-Sims 8
Merritt-Douglass-Harris-Shepherd-Sims 4
Grady-LucasPerry-Harris-Novak-Gibson 3
Merritt-LucasPerry-Harris-Shepherd-Sims 0
Merritt-LucasPerry-Lee-Wright-Puls 0
Grady-LucasPerry-Harris-Wright-Sims -1
Grady-LucasPerry-Lee-Shepherd-Sims -3
Merritt-Douglass-Harris-Shepherd-Gibson -4
Grady-LucasPerry-Harris-Gibson-Sims -7

Beyond individual performances I looked at which lineups clicked. The one thing I need to do next time is track the number of minutes each line up plays. This will be a little more work but looking back at it, it is definitely necessary. The three lineups that got the most playing time in my estimation are:

  • Merritt, Douglass, Harris, Shepherd, Sims
  • Grady, Lucas-Perry, Harris, Gibson, Sims
  • Grady, Douglass, Harris, Novak, Sims

The best by a large margin was Grady, Douglass, Harris, Novak, and Sims. The lineup numbers confirm that when Novak is on the floor the team produces. The Grady, LLP, Harris, Novak, Sims lineup was only on the court the first time Laval came on the court and when he hit his first two threes. I would like to see Beilein go with this group a little more but understand that Novak was limited to 15 minutes because of fouls.

The two worst lineups have the common denominator of Zack Gibson being on the floor. The lineups with Gibson at the five were only used for short stretches when DeShawn was on the bench (6 minutes total). Wright also got limited playing time but he wasn’t on the floor with any lineups that posted a positive +/-.

Looking at these lineups you can see why Beilein started Novak in the second half and pulled Merritt after only 30 seconds or so. Grady, Douglass, Harris, Novak, Sims is definitely our strongest lineup at this point even if it does present a size problem at the PF.

It’s hard to take too much from this data after only one game because it’s just a small sample size and after a couple games in the Big Ten it will be easier to identify which lineups work and which don’t. I am very curious whether Novak continues to post huge +/- numbers as well or if this was more of a fluke.

I’d love some feedback on this so I can continue to try to improve it. 82games is the inspiration but they have a lot more resources and data at their finger tips to track this kind of stuff so I am going to have to pick and choose what is reasonable.

On another note, be sure to check out Dana O’Neill’s feature on Manny Harris and John Beilein.

  • Robert 04

    I think that this will sufice as a sort of evaluation/post-game tool. As you noted, I think you want to track the number of minutes each respective lineup plays.

  • Robert 04

    Question for the readers/bloggers: Should LLP ultimately start? If so, in replace of douglass or intandem with him (displacing Merritt)? Or should he stay as firepower off the bench?

  • LLP is playing strictly the 2 so he won’t replace Merritt, Grady would be the one to do that.

  • Andy

    Very interesting and good to see. I think factoring time played would definitely be a positive addition. When I calcuate value for fantasy baseball (nerd, I know) I divide stats (H, HR, whatever) by the square root of ABs. It’s the best way I’ve come up with to make time played a factor, but to also realize that the value of someone who is 3 for 9 is much less than the value of someone who is 200 for 600.

    So, for example, say Wright plays 2 minutes and is a plus 1. At the same time, LLP plays 16 minutes and is a plus 8. If you just do value per minute, you’d say that both are worth 0.5 pts a minute. But, if you divide by the square root, LLP gets a value of 2 (weird units, just call it a value factor or something) and Wright gets a value of 0.7 which I think is more representative.

    Thanks for allowing my nerdy love of sports and math to come together. :)

  • Ryan

    Great stuff Dylan, a very nice addition.

    One question and then a couple comments…are you calculating +/- by taking a player’s +/- while on the floor AND subtracting +/- while off the floor? If you are, I’m not saying it’s the wrong way to go, but using this method on one individual game will reinforce positive or negative numbers (tend to make positive numbers more positive and vice versa), as evidenced by the results at the PF position.

    I know you mention it towards the end, but it is dangerous to make any real conclusions from individual +/- from one game…definite sample size problem with a lot of noise. If Novak were to continue to put up those kinds of individual numbers and still have a great +/- over the next 10 games, then I think you can conclude the “threat” of Novak’s shooting (as opposed to the actual results) was resulting in opening up the floor for other guys to score more effectively. But in one game it’s entirely possible that he was just lucky enough to happen to be on the floor when Michigan was making a run…which I happen to think was mostly the case last night, especially when you consider when Novak is out of the game Michigan will generally have other people on the floor that can shoot and provide the same threat. I think individual +/- is much more valuable as an analytical tool after 10-15 games. Having said that, it is still interesting to look at on individual game basis.

    Another thing complicating matters in Michigan’s case is that several guys are playing multiple positions. For example, you could conclude from last night’s results that Novak is clearly our best option at PF…but maybe he got his best results as a SF or SG? I don’t know whether that is the case or not, just using as an example.

    I really like the lineup analysis, and think it carries a little more weight on an individual game basis. It will be interesting to see if rotation tightens up as Beilein says it will, if/how the various lineups’ minutes converge to the best performing lineups on a +/- basis.

  • Good data. I look forward to the expanded minutes of each lineup like you mentioned.

    I would also give a “second” to Andy’s suggestion.

    A third suggestion: I can’t imagine the work that would go into this, but it would be amazing to see how one UofM lineup compares against opponent lineups (+/- only I’m sure).

    It might provide some transparency as to how certain lineups perform against heavy front court lineups or small, fast lineups. But again, I’m not sure how much work that would take. And my once-promising math brain seemingly retired sometime around 2005.

  • JimC

    The +/- is interesting, and it seems to confirm what we’ve all been seeing the past several weeks. It’ll be cool to see that after many game.
    Like Ryan, I’m wondering what the ‘measure’ means. Is it point-margin change while the player is in?

  • Andy

    I’m assuming the +/- is like hockey, right? If I come on the floor and it’s 0-0 and when I leave, it’s 10-5. I am +5, right?

  • Dylan

    Sorry I should have been more clear. It’s just points scored by the team when the player is on the court minus points by nccwhile they are on the court.

  • Tom

    I was watching the NCCC game again today and I really think JB really needs to trim the player rotation down to 7-8 players and give them time on the floor. With this 10 man rotation, he his getting players that come in and think, “I gotta make something happen or I’ll losing more playing time.” I think that 7 players getting solid minutes (Harris, Sims, Novak, LLP, Grady, Douglass, Shepard….with a little Gibson) would settle the team down a bit and give the players more time to get comfortable on the floor. As a former Beilein player, I know that when guys get in and know they are coming out after 4/5 minutes, they tend to force things…especially younger players (Douglass will hoist 5 threes a game no matter what the defense gives him). JB needs to get people in there that will run his offense for 25-35 seconds each possession….not three passes and a 22 footer…how many really good backdoors have we seen the last two games?

  • AC

    I also assumed like Andy that the +/- was calculated just like hockey.

    I will also “third” Andy’s suggestion that you normalize the data against minutes played. I’m not sure the best way to do this, but I think it makes the data the most meaningful. Maybe instead of minutes you could do possessions, but that takes a lot more work.

    My real interest is whether this will be something you do every game. I agree that the UFR is not really practical. I tried to digest the one on Varsity Blue and lost interet really quick. This makes sense to me. But I think the only way you can trust the results is to compile it over multiple games. Combine multiple games with a minutes/possessions normalization and you’re now on to something.

    Keep it up!

  • ToBlav

    Nice feature. I think it indicates that LLP is requiring a little time to find his stride, which is to be expected. However, it is easy to forget that when I got off to the good start he had.

  • Also RE: positions…

    They are pretty much set in stone:
    PG: Grady 27, Merritt 13
    SG: Lucas-Perry 23, Douglass 17
    SF: Harris 36, Lee 4
    PF: Shep 13, Novak 15, Wright 3, Gibson (9)
    C: Sims 34, Gibson (5), Puls 1

    The only guy who played 2 positions was Gibson. When I wrote the lineups I had everyone in positions as well, so you can see LLP never played with Stu, Novak never played without Stu or LLP etc.

  • Ken in Vegas

    I like it, Dylan. I have problems putting the stats hat on sometimes, but Lord knows I try. I’ve got some well needed training from reading the DetroitTigersWeblog all the time. Baseball is a whole other ballgame, pun intended. I’m really going to enjoy watching last nights game for the first time this evening now that I know some things to keep an eye on.

    How pumped are you guys for the Wisconsin game? Man am I juiced. The Badgers have badgered (lame) us for so many years that I am just foaming at the mouth to win that game.

  • JRose

    Dylan, great stat breakdown from the NCC game…yet again, great work! Here is a link to a great article about Coach B and Harris. Go Blue!
    http://sports.espn.go.com/ncb/columns/story?columnist=oneil_dana&id=3799411

  • Bluebufoon

    Dylan have you heard anything about Will Regan growing (somewhat considerably) ? I know, I know I’m taking this stuff way too seriously but last week one of the sites might have been Rivals listed Regan at 6-10, not 6-8 which is what has been the more consistent listing for him, but that extra two inches can make all the difference in the world when Michigan takes on the North Carolina’s, Ohio State’s and Michigan State’s of the world.

    Anyway Regan had another stellar game on Monday night 18 pts and 19 rebounds — can’t argue with those numbers.

    From Buffalo paper,

    NHSCA FESTIVAL: Will Regan had 18 points and 19 rebounds as Nichols edged Gwynn Park, 62-57, in overtime. Ben Meyer hit the game-tying shot to force OT. Ron Canestro led the Vikings with 19 points and Andy MacKinnon added 11. Nichols, ranked No. 2 in the state in Class A, improves to 7-1.

  • NVE

    I love the lineup +/-. Great way to see which combos of players are working best.

    The player +/- is really useful for guys getting mid to low # of minutes. As you pointed out, it’s less insightful for players that are on the floor most of the game since the final score pretty much dictates their +/-.

    Thanks for doing this!

  • Not sure about his height but he has been producing and he’s still only a junior so I wouldn’t be surprised if he was still growing. I’m not going to argue with his numbers and it seems like he’s playing pretty legit prep competition.

  • NVE

    I also noticed that one could argue that the following combination of players is the most talented quintet U-M could put on the floor at the same time:

    Grady-LucasPerry-Harris-Gibson-Sims -7

    Worst combo of the bunch.

  • Robert 04

    NVE I thought the same thing. That’s why I was advocating for maybe LLP at the point, with Douglass at the 2.
    BUT, as we all know, LLP is not being used as a 1.

    I would love to see LLP, Douglass, Novak, Harris and Sims when we can get away with small ball.

    Gibson is only really an asset against big frontlines.
    Though, Grady has been a good point I like the idea of an all 6 foot or taller lineup with fewer missed layups.

  • ryan

    Illinois Beats Purdue in OT at Purdue.

  • Yea just watched the game, Big Ten is gonna be tough every night. We better bring it tomorrow because you know Bo Ryan’s boys will be fired up.

  • FL Wolve

    I love the +/- numbers. That is really good info. I bet if you went back and did the same thing for the Oakland and FGCU games you would get similar results. Just a couple of points that I have noticed over the past few weeks which are confirmed by this analysis

    1. The team is playing considerably better when Grady is in the game. Merritt is really starting to struggle on both ends but probably more so on the defensive side.
    2. LLP is able to get his when he’s in the game but he’s not facilitating for his teammates. He has zero assists in 3 games. The offense really seems to bog down when he’s in the game. He’s either hitting a shot or they’re not scoring at all.
    3. I think the Novak number is obviously inflated by this one game. It seems like in the prior 2 games where he played PF the team struggled with him at the PF.
    4. I understand Shepherd’s minutes were cut for Novak but he hasn’t really been playing poorly. I still think we need Shepherd playing more PF minutes for us (I just wish he could make those wide open 3 pointers).

    Great job on the analysis Dylan. I’m looking forward to this kind of stuff after every game. You’ve set the bar pretty high now.

  • Erik

    I like the +/- but will hope that you can expand it to include other stats.

    Ultimately points is what matters, and +/- is the best indicator for the score, but I think things like assist to turnover ratio, FG%, etc are very telling stats and it would be interesting to see those broken down by unit to see how efficiently they are operating.

    If you wanted to stick with the +/- thing, then maybe +1 for an assist -1 for a TO. For FG percentage maybe +1 for a FG, +1.5 for a 3 pointer and then -1 and -1.5 for the misses?

    I’m no math major, so maybe those need tweaking. I don’t think +/- is the best way to capture the efficiency for those stats, but having some way to break down those key stats by unit would be very interesting to me.

  • Giddings

    GREAT stuff Dylan… I remember craving a +/- breakdown after the Duke game and although it took a little while it’s finally here!

    I also agree with Andy on the normalization method, and agree with everyone on compiling the info over several games. I’m not sure how much time it took to do the breakdown for this particular game, but if we had this after every game it would be amazing.

    For individuals:
    raw +/-, minutes, adjusted +/- (for that game only)
    raw +/-, minutes, adjusted +/- (cumulative / season)
    For lineups:
    raw +/-, minutes, adjusted +/- (for that game only)
    raw +/-, minutes, adjusted +/- (cumulative / season)

    Heck, if it’s too much work for you I’d be glad to help out. These kind of stats, in my opinion, are right up there with Ken Pomeroy’s stuff (albeit a little more simple) as far as revealing the truths about a team and its players.

    Also, I agree that this is 10x better than the play-by-play breakdown. Please please stick to this.

    Let’s pack Crisler tomorrow, folks!

  • Giddings

    Forgot to add one thing – does anyone think that Beilein does a similar sort of analysis for his teams? It would be interesting to find out…

  • Adam

    I’d be willing to bet that he does. He may only present them with the bigger +/- differences but we had something similar when I played. It’s not that difficult to figure out and you know coaches watch that tape at least 3 times of every game.

  • 90 M varsity

    +/- is interesting info, but I don’t know how much you can take from it because the majority of the guys who are a +20 line up are also in the -7 line up.

    I like the individual +/- numbers, but there are so many variables that I don’t know what you can take from them unless you just have a guy or two who is always a minus player. Fun to read though.

  • stinky

    I like the addition of the + – lineup stats!

  • Evan

    I like the number-based stuff, but as long as we’re putting up our wish lists, I wouldn’t mind seeing a more qualitative analysis. Example: X lineup was on the floor when opponent goes on a 10 point run, Beilein makes subsititutions so Y lineup is on the floor and those substitutions work or they don’t work. Maybe go a little deeper – 10 point run by hitting threes, by pushing the ball, by pounding it inside, etc. – so we can see if Beilein is getting the results he expects from his substitutions. The best line-up from a statistical standpoint might not show us the best line-up in a specific situation, unless you got way too complicated, but some analysis of what is going on in the game and how the line-up changes to adjust to the game would give us insight into how well Beilein understands his players and can maximize their effectiveness.

    Evan

  • Tim

    Dylan, I noted the times for each lineup in the NC Central UFR if you care to add them up and see how often each set of 5 played in the game.