PDA

View Full Version : Cumulative Plus/Minus, 2008-09



Jumbo
11-10-2008, 10:05 PM
Final 2008-09 Totals ...

Individuals
Kyle Singler 2,365-1,938 (+427)
Jon Scheyer 2,400-1,993 (+407)
Gerald Henderson 2,206-1,846 (+360)
Nolan Smith 1,482-1,152 (+330)
Brian Zoubek 863-620 (+243)
David McClure 1,071-935 (+136)
Greg Paulus 1,075-994 (+81)
Lance Thomas 1,272-1,194 (+78)
Elliot Williams 1,017-943 (+74)
Martynas Pocius 226-217 (+9)
Steve Johnson 14-13 (+1)
Miles Plumlee 270-271 (-1)
Olek Czyz 74-79 (-5)

Per 40 Minutes
Brian Zoubek +22.8
Nolan Smith +18.0
Kyle Singler +14.3
Jon Scheyer +13.4
Gerald Henderson +13.1
David McClure +9.6
Greg Paulus +5.6
Elliot Williams +5.3
Lance Thomas +4.5
Steve Johnson +4.0
Martynas Pocius +2.7
Miles Plumlee -0.2
Olek Czyz -3.9


Lineups
Smith-Scheyer-Henderson-Singler-Zoubek (x68) 368-232 (+136)
Smith-Scheyer-Henderson-Singler-Thomas (x71) 281-241 (+40)
Smith-Scheyer-Henderson-McClure-Singler (x41) 206-177 (+29)
Paulus-Smith-Scheyer-Henderson-Singler (x21) 69-45 (+24)
Scheyer-Williams-Henderson-Singler-Zoubek (x16) 62-40 (+22)
Smith-Scheyer-Williams-Singler-Thomas (x11) 41-20 (+21)
Paulus-Scheyer-Henderson-McClure-Singler (x28) 124-104 (+20)
Paulus-Smith-Henderson-McClure-Singler (x6) 34-17 (+17)
Scheyer-Williams-Henderson-McClure-Singler (x21) 79-63 (+16)
Scheyer-Henderson-McClure-Singler-Zoubek (x2) 19-4 (+15)
Scheyer-Pocius-Williams-Singler-Zoubek (x2) 15-3 (+12)
Smith-Henderson-McClure-Singler-Thomas (x6) 17-7 (+10)
Paulus-Scheyer-Henderson-Singler-Zoubek (x21) 50-41 (+9)
Paulus-Scheyer-McClure-Singler-Zoubek (x10) 27-18 (+9)
Scheyer-Williams-Henderson-McClure-Thomas (x7) 20-11 (+9)
Smith-Scheyer-Williams-Singler-Zoubek (x7) 19-10 (+9)
Paulus-Scheyer-Henderson-McClure-Zoubek (x7) 27-19 (+8)
Paulus-Scheyer-Henderson-Thomas-Plumlee (x4) 17-9 (+8)
Paulus-Scheyer-Henderson-McClure-Plumlee (x3) 11-3 (+8)
Paulus-Scheyer-Williams-Henderson-Singler (x14) 31-24 (+7)
Smith-Williams-Henderson-Singler-Thomas (x8) 25-18 (+7)
Smith-Scheyer-Williams-Thomas-Zoubek (x2) 9-2 (+7)
Paulus-Williams-Henderson-Singler-Thomas (x7) 20-14 (+6)
Paulus-Scheyer-Henderson-Thomas-Zoubek (x4) 10-4 (+6)
Paulus-Scheyer-Williams-Thomas-Zoubek 9-3 (+6)
Smith-Scheyer-Pocius-McClure-Singler 8-2 (+6)
Smith-Pocius-Williams-McClure-Zoubek 8-2 (+6)
Smith-Scheyer-Williams-McClure-Singler (x8) 13-8 (+5)
Paulus-Henderson-McClure-Singler-Thomas (x7) 14-9 (+5)
Paulus-Smith-Scheyer-Singler-Plumlee (x3) 12-7 (+5)
Paulus-Smith-Scheyer-Henderson-McClure (x3) 9-4 (+5)
Smith-Scheyer-Pocius-McClure-Thomas (x2) 7-2 (+5)
Paulus-Smith-Pocius-Singler-Thomas 5-0 (+5)
Smith-Scheyer-McClure-Singler-Thomas (x19) 44-40 (+4)
Paulus-Henderson-McClure-Thomas-Zoubek (x6) 21-17 (+4)
Paulus-Scheyer-Williams-McClure-Singler (x6) 14-10 (+4)
Paulus-Pocius-Williams-McClure-Zoubek (x3) 18-14 (+4)
Smith-Williams-Henderson-Singler-Zoubek (x3) 9-5 (+4)
Smith-Scheyer-Pocius-Singler-Thomas (x2) 4-0 (+4)
Paulus-Williams-McClure-Thomas-Zoubek (x2) 4-0 (+4)
Scheyer-Pocius-Henderson-Singler-Plumlee 6-2 (+4)
Smith-Pocius-Williams-Czyz-Plumlee 4-0 (+4)
Smith-Pocius-Henderson-Singler-Thomas 4-0 (+4)
Smith-Pocius-Williams-Singler-Thomas 4-0 (+4)
Paulus-Smith-Williams-Henderson-Thomas 4-0 (+4)
Smith-Scheyer-McClure-Singler-Zoubek (x9) 27-24 (+3)
Smith-Henderson-McClure-Singler-Zoubek (x4) 7-4 (+3)
Paulus-Williams-McClure-Thomas-Plumlee (x2) 11-8 (+3)
Scheyer-Williams-Henderson-Thomas-Zoubek (x2) 3-0 (+3)
Smith-Henderson-McClure-Thomas-Plumlee 6-3 (+3)
Paulus-Scheyer-Pocius-Singler-Zoubek 6-3 (+3)
Paulus-Pocius-Henderson-Singler-Thomas 4-1 (+3)
Paulus-Smith-Henderson-Singler-Plumlee 3-0 (+3)
Paulus-Williams-Henderson-Singler-Plumlee 3-0 (+3)
Paulus-Smith-Scheyer-Henderson-Zoubek 3-0 (+3)
Smith-Scheyer-Williams-Henderson-Zoubek 3-0 (+3)
Paulus-Smith-Scheyer-Singler-Thomas (x4) 10-8 (+2)
Paulus-Pocius-Williams-Czyz-Thomas (x2) 6-4 (+2)
Scheyer-Williams-McClure-Singler-Plumlee (x2) 4-2 (+2)
Smith-Williams-McClure-Singler-Thomas 7-5 (+2)
Smith-Scheyer-McClure-Thomas-Plumlee 4-2 (+2)
Scheyer-Henderson-McClure-Singler-Plumlee 4-2 (+2)
Paulus-Smith-Scheyer-Henderson-Thomas 3-1 (+2)
Paulus-Pocius-McClure-Singler-Zoubek 2-0 (+2)
Paulus-Smith-Williams-Singler-Plumlee 2-0 (+2)
Smith-Pocius-Williams-Henderson-Singler 2-0 (+2)
Scheyer-Williams-Czyz-Thomas-Plumlee 2-0 (+2)
Williams-Pocius-Henderson-Singler-Zoubek 2-0 (+2)
Paulus-Henderson-McClure-Czyz-Thomas 2-0 (+2)
Smith-Scheyer-Williams-McClure-Zoubek 2-0 (+2)
Smith-Scheyer-Pocius-Singler-Zoubek 2-0 (+2)
Smith-Pocius-Henderson-Singler-Zoubek 2-0 (+2)
Smith-Scheyer-Williams-Henderson-McClure 2-0 (+2)
Smith-Scheyer-Henderson-Singler-Plumlee (x8) 29-28 (+1)
Paulus-Scheyer-Williams-Singler-Zoubek (x6) 16-15 (+1)
Paulus-Scheyer-Henderson-Singler-Plumlee (x4) 11-10 (+1)
Williams-Pocius-Johnson-Czyz-Plumlee (x3) 10-9 (+1)
Paulus-Smith-Scheyer-Singler-Zoubek (x2) 7-6 (+1)
Paulus-Smith-Henderson-Singler-Zoubek (x2) 6-5 (+1)
Paulus-Scheyer-Pocius-Singler-Thomas (x2) 5-4 (+1)
Paulus-Scheyer-Williams-Henderson-Thomas (x2) 5-4 (+1)
Paulus-Scheyer-McClure-Thomas-Plumlee (x2) 1-0 (+1)
Paulus-Henderson-McClure-Singler-Zoubek (x2) 1-0 (+1)
Paulus-Williams-McClure-Czyz-Plumlee 5-4 (+1)
Paulus-Smith-Scheyer-Thomas-Zoubek 4-3 (+1)
Scheyer-Williams-Singler-Czyz-Zoubek 4-3 (+1)
Smith-Scheyer-Pocius-Henderson-Singler 3-2 (+1)
Paulus-Smith-McClure-Singler-Thomas 3-2 (+1)
Scheyer-Pocius-McClure-Thomas-Plumlee 1-0 (+1)
Paulus-Scheyer-Henderson-McClure-Thomas (x10) 21-21 (0)
Paulus-Williams-Henderson-McClure-Singler (x3) 7-7 (0)
Smith-Williams-Henderson-McClure-Singler (x3) 5-5 (0)
Scheyer-Williams-McClure-Singler-Thomas (x3) 5-5 (0)
Smith-Scheyer-Singler-Thomas-Plumlee (x2) 10-10 (0)
Paulus-Henderson-McClure-Thomas-Plumlee (x2) 4-4 (0)
Paulus-Smith-Williams-McClure-Zoubek (x2) 2-2 (0)
Paulus-Pocius-Williams-Singler-Zoubek 7-7 (0)
Paulus-Smith-Williams-McClure-Thomas 3-3 (0)
Scheyer-Williams-McClure-Thomas-Zoubek 3-3 (0)
Paulus-Pocius-Williams-Johnson-Czyz 2-2 (0)
Smith-Scheyer-Williams-Thomas-Plumlee 2-2 (0)
Paulus-Pocius-McClure-Czyz-Thomas 2-2 (0)
Paulus-Smith-Scheyer-McClure-Thomas 2-2 (0)
Smith-Williams-Henderson-Thomas-Plumlee 2-2 (0)
Paulus-Williams-Johnson-Czyz-Plumlee 2-2 (0)
Paulus-Pocius-Williams-Henderson-Plumlee (x2) 0-0 (0)
Smith-Pocius-McClure-Singler-Thomas 0-0 (0)
Paulus-Henderson-McClure-Singler-Plumlee 0-0 (0)
Paulus-Pocius-Wiliams-Henderson-McClure 0-0 (0)
Paulus-Pocius-Wiliams-Henderson-Plumlee 0-0 (0)
Paulus-Smith-Scheyer-McClure-Zoubek 0-0 (0)
Paulus-Smith-Henderson-Thomas-Zoubek 0-0 (0)
Smith-Henderson-Williams-McClure-Singler 0-0 (0)
Paulus-Scheyer-Williams-Thomas-Plumlee 0-0 (0)
Paulus-Williams-McClure-Singler-Zoubek 0-0 (0)
Paulus-Scheyer-Williams-Singler-Thomas (x12) 29-30 (-1)
Paulus-Scheyer-McClure-Thomas-Zoubek (x7) 7-8 (-1)
Paulus-Pocius-Williams-McClure-Plumlee (x6) 25-26 (-1)
Smith-Scheyer-Henderson-McClure-Thomas (x6) 12-13 (-1)
Paulus-Smith-Scheyer-McClure-Singler (x5) 9-10 (-1)
Smith-Scheyer-Henderson-McClure-Zoubek (x4) 11-12 (-1)
Smith-Scheyer-Williams-Henderson-Thomas (x4) 8-9 (-1)
Paulus-Williams-Henderson-Singler-Zoubek (x4) 4-5 (-1)
Paulus-Scheyer-Williams-McClure-Plumlee (x2) 3-4 (-1)
Paulus-Williams-McClure-Singler-Thomas (x2) 1-2 (-1)
Smith-Scheyer-Williams-McClure-Thomas (x2) 1-2 (-1)
Smith-Williams-Henderson-Thomas-Zoubek 4-5 (-1)
Smith-Williams-Henderson-McClure-Plumlee 3-4 (-1)
Paulus-Pocius-McClure-Thomas-Zoubek 2-3 (-1)
Smith-Scheyer-Pocius-Singler-Plumlee 0-1 (-1)
Paulus-Scheyer-Williams-Henderson-McClure 0-1 (-1)
Paulus-Scheyer-McClure-Singler-Thomas (x19) 41-43 (-2)
Scheyer-Williams-Henderson-Singler-Plumlee (x6) 17-19 (-2)
Smith-Scheyer-Henderson-Thomas-Zoubek (x2) 7-9 (-2)
Scheyer-Williams-Henderson-McClure-Zoubek (x2) 3-5 (-2)
Paulus-Pocius-Henderson-Singler-Zoubek (x2) 2-4 (-2)
Paulus-Pocius-Henderson-McClure-Singler (x2) 2-4 (-2)
Paulus-Pocius-Williams-McClure-Singler (x2) 0-2 (-2)
Paulus-Scheyer-Pocius-Henderson-Singler (x2) 0-2 (-2)
Paulus-Scheyer-McClure-Singler-Plumlee (x2) 0-2 (-2)
Paulus-Smith-Scheyer-Thomas-Plumlee 4-6 (-2)
Paulus-Pocius-Williams-McClure-Thomas 3-5 (-2)
Paulus-Pocius-Williams-Singler-Thomas 2-4 (-2)
Paulus-Pocius-Williams-McClure-Czyz 2-4 (-2)
Paulus-Pocius-Williams-Czyz-Zoubek 0-2 (-2)
Smith-Scheyer-Williams-Singler-Plumlee 0-2 (-2)
Smith-Pocius-Williams-Thomas-Plumlee 0-2 (-2)
Paulus-Smith-Henderson-McClure-Plumlee 0-2 (-2)
Paulus-Scheyer-Pocius-McClure-Thomas 0-2 (-2)
Smith-Scheyer-Pocius-Williams-Thomas 0-2 (-2)
Paulus-Smith-Williams-McClure-Singler 0-2 (-2)
Smith-Pocius-Williams-McClure-Czyz 0-2 (-2)
Paulus-Smith-Pocius-Williams-McClure 0-2 (-2)
Scheyer-Williams-McClure-Singler-Zoubek (x4) 10-13 (-3)
Paulus-Smith-Williams-Singler-Thomas (x4) 8-11 (-3)
Paulus-Scheyer-Williams-Singler-Plumlee (x3) 4-7 (-3)
Paulus-Scheyer-Pocius-McClure-Zoubek (x2) 1-4 (-3)
Paulus-Scheyer-Williams-McClure-Zoubek (x2) 0-3 (-3)
Paulus-Pocius-Czyz-Plumlee-Zoubek 3-6 (-3)
Smith-Pocius-Williams-Henderson-McClure 2-5 (-3)
Paulus-Williams-Henderson-Thomas-Plumlee 2-5 (-3)
Paulus-Smith-Henderson-McClure-Zoubek 2-5 (-3)
Williams-Pocius-Czyz-Thomas-Plumlee 0-3 (-3)
Paulus-Smith-Scheyer-Pocius-Zoubek 0-3 (-3)
Paulus-Smith-Henderson-McClure-Thomas (x5) 5-9 (-4)
Smith-Williams-Henderson-McClure-Thomas (x3) 6-10 (-4)
Paulus-Scheyer-Williams-Henderson-Plumlee (x2) 2-6 (-4)
Paulus-Scheyer-Williams-McClure-Thomas (x6) 17-22 (-5)
Paulus-Williams-Henderson-McClure-Thomas (x4) 9-14 (-5)
Smith-Scheyer-Henderson-Thomas-Plumlee (x2) 0-5 (-5)
Scheyer-Pocius-Williams-Thomas-Plumlee 0-5 (-5)
Paulus-Pocius-Williams-Czyz-Plumlee (x8) 30-36 (-6)
Smith-Scheyer-Williams-Henderson-Singler (x7) 14-20 (-6)
Paulus-Smith-Pocius-Singler-Zoubek 2-8 (-6)
Scheyer-Henderson-McClure-Singler-Thomas (x4) 6-13 (-7)
Paulus-Williams-Henderson-McClure-Zoubek (x2) 9-16 (-7)
Paulus-Smith-Henderson-Singler-Thomas (x6) 5-13 (-8)
Smith-Henderson-McClure-Thomas-Zoubek (x4) 12-20 (-8)
Paulus-Scheyer-Henderson-Singler-Thomas (x35) 128-138 (-10)
Paulus-Pocius-Williams-Thomas-Plumlee (x4) 11-21 (-10)
Scheyer-Williams-Henderson-Singler-Thomas (x53) 251-264 (-13)


Net +/-

Name Net+/- +/-On +/-Off +/- +/- +/-
****** per 40 per 40 per 40 On Tot Off Tot Net Tot
Zoubek 15.77 22.76 6.99 243 185 58
Singler 14.18 14.32 0.14 427 1 426
Smith 12.76 17.98 5.22 330 98 232
Scheyer 10.31 13.41 3.10 407 21 386
Henderson 6.09 13.11 7.03 360 68 292
McClure -3.16 9.58 12.74 136 292 -156
Johnson -7.58 4.00 11.58 1 427 -426
Paulus -9.70 5.61 15.30 81 347 -266
Pocius -9.75 2.67 12.41 9 419 -410
Williams -10.10 5.26 15.36 74 354 -280
Thomas -13.06 4.53 17.59 78 350 -272
Plumlee -13.24 -0.24 13.00 -1 429 -430
Czyz -16.00 -3.92 12.08 -5 433 -438

Lulu
11-11-2008, 03:50 AM
Are we allowed to post in this thread? If not please just delete this.

Sorry for being new to this, but I'm guessing the numbers are:
"Duke pts-Opponent pts (difference)"
thus giving the scoring disparity for each player and lineup. Correct?

If so, then I have a very humble suggestion... What if we took the ratio DukePts/OppPts and reported that, perhaps in place of the difference? Then we know how many points Duke scores for every point the opponent scores with that player/lineup. This not only removes the time bias, but also the pace bias, because when it comes to winning I'd think the ratio at which we outscore the opponent is an even better measure than the rate. (That said, if we want purely offensive or defensive measures, per time or preferably per possession would take precedence.)

For the present numbers, the ratio would look like this:
PLAYER Duke Opp Ratio
Nolan Smith 47 19 2.474
Jon Scheyer 56 24 2.333
Lance Thomas 41 19 2.158
Brian Zoubek 14 7 2.000
David McClure 36 19 1.895
Marty Pocius 19 11 1.727
Kyle Singler 51 30 1.700
Elliot Williams 36 26 1.385
Greg Paulus 43 36 1.194
Miles Plumlee 14 12 1.167
Gerald Henderson 37 34 1.088
Olek Czyz 6 8 0.750
DUKE TEAM 80 49 1.63

Note that instead of above we could also take similar ratios:
DukePts/(DukePts+OppPts); or
(DukePts-OppPts)/OppPts; or
(DukePts-OppPts)/(DukePts+OppPts); etc etc etc.
All with their own meanings. But I think I'd personally just prefer the first.

Technically, the best numerical indicator for the effectiveness of a player/lineup and cross-game analysis is probably to compare the ratios reported above to that of the team as a whole in the last row. Probably something like (Player Ratio/Duke Ratio)-1, which for the above yields:

Nolan Smith 0.5151
Jon Scheyer 0.4292
Lance Thomas 0.3217
Brian Zoubek 0.2250
David McClure 0.1605
Marty Pocius 0.0580
Kyle Singler 0.0413
Elliot Williams -0.1519
Greg Paulus -0.2684
Miles Plumlee -0.2854
Gerald Henderson -0.3335
Olek Czyz -0.5406

This basically says that Duke's scoring ratio increased 52% with Smith in the game, and fell 33% from average with Henderson... at least in today's game. [EDIT: Fine, I guess it isn't technically correct to say "increased" since we are comparing Smith's presence to an average that was also computed with him partially present, instead of when he wasn't playing at all... but the idea is the same...]

I know there are a lot of smart people here, so this has probably already been discussed...

Over the course of the season I also think it would be interesting to add weight to these numbers depending on the ease of the win. This would account for numbers getting overblown in easy wins, and the fact that easy wins are when some players see the most playing time. It's a little arbitrary trying to decide how much more an even game is worth as opposed to a game with an overmatched opponent, but one example would be to use the inverse of the ratio given for the Duke Team above for the weight (in other words OppPts/DukePts for the entire game). Games we lost would actually have the highest weight then, presumably against the toughest opponents. We should also add a weight for the number of minutes the player played each game, since a 5:1 ratio in 1 minute of playing time isn't worth a 2:1 ratio in 30 minutes. Then we take a weighted average of the totals over all games played.

However, that's getting far more complex than I intended when I began this comment. A cumulative season-long ratio is far more simple.

Out of curiosity, how do you get your numbers Jumbo? Just from watching the game and lineup when a score occurs? Must be a bit of work if that's the case. I was just wondering because I was going to play with a code to make all of this easier if the numbers were readily available somewhere.

Jumbo
11-11-2008, 11:48 PM
If so, then I have a very humble suggestion... What if we took the ratio DukePts/OppPts and reported that, perhaps in place of the difference? Then we know how many points Duke scores for every point the opponent scores with that player/lineup. This not only removes the time bias, but also the pace bias, because when it comes to winning I'd think the ratio at which we outscore the opponent is an even better measure than the rate. (That said, if we want purely offensive or defensive measures, per time or preferably per possession would take precedence.)
I've never really seen a site do this as a ratio. No reason you can't add it on, but I think end-of-the-bench guys will be overvalued. The way I do it is the most common format, and it tends to normalize as the season goes on. Check out last season's numbers when you get a chance. I don't think you could weight it by strength of opponent -- way too tough.




Out of curiosity, how do you get your numbers Jumbo? Just from watching the game and lineup when a score occurs? Must be a bit of work if that's the case. I was just wondering because I was going to play with a code to make all of this easier if the numbers were readily available somewhere.

It's just a simple game log. All you have to do is track substitutions and note the score.

Jumbo
11-11-2008, 11:48 PM
Now updated through the Georgia Southern game ...

Lulu
11-12-2008, 01:41 AM
I've never really seen a site do this as a ratio. No reason you can't add it on, but I think end-of-the-bench guys will be overvalued. The way I do it is the most common format, and it tends to normalize as the season goes on. Check out last season's numbers when you get a chance. I don't think you could weight it by strength of opponent -- way too tough.

I didn't realize it was a standard format. This definitely makes sense if you are looking for the MVP, but it's difficult to try to identify an undervalued player or one that is making intangible contributions to the team's performance. I thought the ratio helped in that regard. The raw scoring difference means little without considering minutes played, if the intent is to compare player effectiveness. I just wanted to remove the minutes played and pace factors.

I asked about the source of your stats because I was going to explore things such as guard combinations and look for correlations. However, I think there's enough info in the numbers you do post to do this. Also, once I realized there were "On" and "Off" columns and what that meant it helped. Thanks!

ACCBBallFan
11-12-2008, 02:08 AM
Thanks for continuing to do this, Jumbo.

It always adds some insight, for example most observers would say Kyle Singler was clearly player of the game both games, but is not the plus/minus leader.

I use a more simplistic attempt to factor in all the box score stats both raw score and on a per 40 minutes basis and for GA southern came up with:

Rating Name (min)

81 Singler (22)
65 Henderson (20)
54 Scheyer (22)
53 Smith (18)
48 Paulus (21)
48 Zoubek (14)
44 McClure (11)
43 Pocius (17)
35 Thomas (13)
34 Williams (14)
32 Plumlee (14)

I draw an arbitrary distinction for at least 10 mintes played vs. < 10 minutes

57 Czyz (9)
06 Johnson (5)

Ratings can get skewed in a single game but tend to normalize over course of the season.

CDu
11-12-2008, 06:41 AM
I didn't realize it was a standard format. This definitely makes sense if you are looking for the MVP, but it's difficult to try to identify an undervalued player or one that is making intangible contributions to the team's performance. I thought the ratio helped in that regard. The raw scoring difference means little without considering minutes played, if the intent is to compare player effectiveness. I just wanted to remove the minutes played and pace factors.

I asked about the source of your stats because I was going to explore things such as guard combinations and look for correlations. However, I think there's enough info in the numbers you do post to do this. Also, once I realized there were "On" and "Off" columns and what that meant it helped. Thanks!

The problem with the ratio approach, as Jumbo alluded to, is that it explodes for small numbers. In other words, say you have a guy who only plays a few minutes during which Duke outscores its opponent 6-2. The ratio there is 3. Now, say you've got a guy who is there when we outscore the opponent 50-30. The ratio there is 1.67. Now, how do you compare those two? I'd say it's very misleading to suggest that the guy with the 3 ratio did better, because it's just too small a sample size with too much room for random variation. But simply going by the ratio one would say that the bench guy did better.

Jumbo gives a +/- per 40 minutes, which sort of does what you want (takes the time out of it). It is subject to some of the same concerns for small-minute players. He also presents the overall totals. Between the two, you can discern who is productive in more limited time and who is productive in a lot of time.

CDu
11-12-2008, 06:47 AM
Ratings can get skewed in a single game but tend to normalize over course of the season.

Yup. This is the small sample size issue. Too much can happen in a single game to skew the stats and make a player's individual +/- look much better (or worse) than his actual performance would suggest. This is especially true for a team game in which a player may have little influence at all on many (if not most) of the baskets during any given stretch of the game.

Last year, we saw the +/- for the key players eventually cluster right around the team's average margin of victory by the end of the season when the sample size got a bit bigger (still small, but bigger). Some players were a bit above, some were a bit below. I'd expect more of the same this year. Of course, it may vary a bit more if Coach K goes into a deeper rotation, or if he goes to more distinct team units.

Lulu
11-12-2008, 09:09 AM
You'd think I had suggested rewriting the Bible... I was just suggesting an idea for an additional informative number we could look at. It could just be a 4th column. Sorry for using the word "replace" earlier; the difference column seemed redundant but it's nice to have there if you don't feel like doing the math. The point totals alone indicate how much Duke is using a particular player.

I also already made note of the issues regarding players with fewer minutes and their potential for skewed results in my initial post.

The problem with the raw scoring numbers is that it becomes little more than a measure of how many minutes a guy plays. Fine, we've acknowledged this. And yes, per minute (or per 40 minute) numbers are given as well. Fine, that's very nice too. I only thought that looking at a per opponent point ratio would be nice as well, for the reasons mentioned. Wasn't supposed to be a big deal.

The other, and probably bigger, problem with any cumulative and unweighted difference (or ratio) is that the easy games carry by far the most weight. A player might stand to earn +40 against an easy opponent, but will be lucky to earn even a couple points, if not lose a few, in a close win against a good opponent. This is why comparing to a team average on a game-by-game basis was an idea.

If you look at last year's numbers, the top 3 vary by only 11 pts on the season. The next 2, Henderson and Paulus, vary by only 2 points from each other. It's virtually the same as total minutes played, seeing as these point differences are insignificant. (We all know a single 3 minute stretch could account for 11 points, not to mention the effect if someone has to sit out a 30-point blowout.) Perhaps Scheyer's number should be strengthened by the fact he played fewer minutes, but it's not. Perhaps Nelson should be penalized for playing almost 2 full games more than anyone else, but you also can't forget that any player who plays 40 minutes in a tough game that's decided by 1 point has earned virtually nothing in the raw scoring difference for that game. It's the big wins (insignificant games) that contribute the most to these numbers.

The only real outlier seemed to be Taylor King, but it's pretty clear that can probably be attributed to his performance early on, especially considering the fact he saw considerable time in the early blowouts. His numbers will therefore be "artificially" high no matter what type of score in computed; barring some measure of opponent strength which we all agree is unnecessary.

So... I guess I'm sorry for even having an idea. I'll just look at my own numbers if I get curious. Thought it was all for fun anyway.

CDu
11-12-2008, 10:08 AM
So... I guess I'm sorry for even having an idea. I'll just look at my own numbers if I get curious. Thought it was all for fun anyway.

I think you're wildly overreacting. You're certainly entitled to your opinion. The way you presented it made it sound like the information Jumbo was presenting was somewhat useless and the ratio thing was the solution.

If Jumbo is willing to add your ratio as an additional measure, I'm certainly not opposed to the idea. I was just stating the flaws in your suggested metric as you were stating the flaws in his metrics. I don't see the need to get huffy about my (or his) response.

Both approaches have flaws, that's all. Ultimately, creating an individual stat for a team-oriented result is going to have flaws. This is especially true when you have a game with lots of substitutions, lots of points scored under random circumstances, and not a lot of games played.

Jumbo
11-14-2008, 01:25 PM
I think last year I did ACC-only plus/minus, too. No reason why I can't do that again.

Jumbo
11-16-2008, 11:44 PM
Now updated through the Rhode Island game ...

Abraxas
11-17-2008, 08:15 AM
A valid point of Lulu was pace. There are levels of intensity at different times of the game that current metrics are not capturing. +/- per possession might be able to give you a better measure of the effectiveness of a player.

In regards to the strength of the opponent, one could use the "rating factor" of Sagarin (divided by 100) at the time of the game, or use an average of several computer rankings.

Lastly, a further "unreachable level" is to factor opponent players (who was your defensive assignment?). It is doable in K's system because our traditional man-to-man defense, but that metric would break down with different zone defenses or when we attack zones (as compared to M-to-M).

Like others, I truly appreciate Jumbo keeping these stats. They do add to the analysis of the game.

Jumbo
11-22-2008, 09:48 PM
Now updated through the Michigan game ...

ACCBBallFan
11-22-2008, 11:06 PM
I guess what surprises me the most is Lance and Dave having very favorable +/- given their lack of Offense, a testament to their defense.

So far Duke has not played any really big teams where IMO Zoubek and Miles are key. I expect Greg will improve as well once he heals a bit.

Hopefully these Montana and Duquesne games give Duke a chance to not play anybody more than about 25 minutes and to get the frosh more oriented, before coach K has to revert back to 8 or 9 man rotation for Purdue game.

Jumbo
11-23-2008, 09:31 PM
Now updated through the Montana game ...

BlueintheFace
11-23-2008, 09:35 PM
The Zoubek numbers are very very encouraging...

bdh21
11-24-2008, 11:09 PM
Can we pin this thread to the top of the EK forum?

Jumbo
11-24-2008, 11:22 PM
Oddly, I find that when things are stickied, they are noticed less. That's why I just bounce it to the top after each game. But if people want me to sticky it, I will.

Jumbo
11-28-2008, 09:22 PM
Now updated through the Duquesne game ...

PSurprise
11-29-2008, 12:28 PM
Yeah for statistics!! Job security for math majors!

Jumbo
12-03-2008, 12:32 AM
Now updated through the Purdue game ...

bird
12-04-2008, 04:15 PM
I am gratified at how often objective measures conform to my subjective beliefs. However, I would have said that our best three players are Singler-Scheyer-Henderson in that order, with a clear gap between that group and Smith and Thomas. The plus minus argues that Smith is right up there. Hmmm. Maybe this K dude knows what he is doing.

coachbob
12-05-2008, 02:06 PM
I find it interesting that our top 4 line ups all include
Smith-Scheyer-Henderson-Singler and one other player.

oso diablo
12-06-2008, 07:23 PM
first, Jumbo, thanks for doing these stats. very nice.

second, had a (stray) thought today about +/-. may be pointless, and may already be accounted for in the methodology, but...

1. is there any adjustment made for subs coming in on free throws?

2. should there be any adjustment for subs coming in when we have the ball vs. when opponent has the ball? maybe it all comes out in the wash or perhaps it's not material, but there would be an elevated expected + if you sub in when we have the ball vs. otherwise.

have any uber-stats-folks looked at that?

Jumbo
12-06-2008, 09:41 PM
1. is there any adjustment made for subs coming in on free throws?
The FTs count toward the group that was on the floor when the foul happened. So, even if we sub while a guy is on the line, the five guys who were on the court and created the FT situation receive the plus/minus, which is only fair.


2. should there be any adjustment for subs coming in when we have the ball vs. when opponent has the ball? maybe it all comes out in the wash or perhaps it's not material, but there would be an elevated expected + if you sub in when we have the ball vs. otherwise.

The only time this should really matter is at the end of the game, when teams are playing offense/defense on every possession. (And, no, there's no way to account for it.) Otherwise, subs play long enough that yes, it ends up being a wash.

I'll try to add the Michigan stats tonight.

Jumbo
12-07-2008, 02:48 PM
Now updated through the second Michigan game ...

ACCBBallFan
12-11-2008, 05:59 PM
I tried to analyze all the combinations that resulted in a negative cumulative (32 combinations @ 5 players = 160).

Based on actual minutes played, I developed how many occurrences across the 1800 minutes would be expected. Not surprisingly, the players with the best plus/minus in Jumbo's analysis had the best variation from expectation:

Sorted by minutes played which is interesting in and of itself:

Min Expect Actual Plus/Minus Player
265 23.6 13 10.6 Scheyer
261 23.2 14 9.2 Singler
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
171 15.2 13 2.2 Thomas
149 13.2 20 -6.8 Paulus
129 11.5 15 -3.5 McClure
122 10.8 23 -12.2 Williams
120 10.7 06 4.7 Zoubek
071 06.3 12 -5.7 Plumlee
065 05.8 16 -10.2 Pocius
026 02.3 05 -2.7 Czyz
006 00.5 01 -0.5 Johnson
1800 160 160 0.0 Total Duke

Sorted by best to worst, it negates some conventional wisdom that Elliott, or even McClure, is a lock down defender, and also provides some validation why certain players play and others do not:

Min Expect Actual Plus/Minus Player
265 23.6 13 10.6 Scheyer
261 23.2 14 9.2 Singler
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
120 10.7 06 4.7 Zoubek
171 15.2 13 2.2 Thomas
006 00.5 01 -0.5 Johnson
026 02.3 05 -2.7 Czyz
129 11.5 15 -3.5 McClure
071 06.3 12 -5.7 Plumlee
149 13.2 20 -6.8 Paulus
065 05.8 16 -10.2 Pocius
122 10.8 23 -12.2 Williams
1800 160 160 0.0 Total Duke

What I found most interesting though is how few times Zoubek appeared out of 32 (expected that Johnson and Czyz would have minimal) and one of these was a gimmick lineup with Scheyer at PG:

Paulus-Pocius-McClure-Thomas-Zoubek 2-3 (-1)
Paulus-Scheyer-Williams-Singler-Zoubek 2-3 (-1)
Paulus-Pocius-Williams-McClure-Zoubek (x2) 5-7 (-2)
Paulus-Pocius-Williams-Czyz-Zoubek 0-2 (-2)
Scheyer-Williams-Henderson-Singler-Zoubek 0-2 (-2)
Smith-Scheyer-McClure-Singler-Zoubek (x4) 7-10 (-3)

Only once do the names Smith, Henderson, Thomas or Czyz appear with Zoubek in a combo with negative plus/minus. The other names all appear 3-4 times, whcih would be expectd more for guys who play a lot of minutes like Singler and Scheyer and perhaps indicate McClure should not be paired with Z much.

Paulus and Zoubek combo is slow at two critical positions and having inexperienced Williams or non sterling Pocius as co-defender just makes it worse, but still it only happened in 6 of 32 worst combos.

Min Expect Actual Plus/Minus Player
006 00.5 01 -0.5 Johnson
026 02.3 05 -2.7 Czyz
120 10.7 06 4.7 Zoubek
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
071 06.3 12 -5.7 Plumlee
265 23.6 13 10.6 Scheyer
171 15.2 13 2.2 Thomas
261 23.2 14 9.2 Singler
129 11.5 15 -3.5 McClure
065 05.8 16 -10.2 Pocius
149 13.2 20 -6.8 Paulus
122 10.8 23 -12.2 Williams
1800 160 160 0.0 Total Duke

Other observations, not unexpectedly, Paulus does worse paired with Scheyer or Pocius than Smith or Williams:

10 Paulus-Pocius
7 Paulus-Scheyer
2 Paulus-Smith
1 Paulus-Williams

Pocius-Williams is a bad combination on defensive end:

12 Pocius-Williams
1 Pocius-Czyz
1 Pocius-Henderson
1 Pocius-McClure
1 Scheyer-Pocius

11 Pocius-Williams
7 Scheyer-Williams
4 Williams-Henderson

Hard to track Elliott since he has played 1-2-3, so I excluded certain combos.

If I get some time, I will do a similar analysis on the top several combos that resulted in favorable cumulative plus/minus. Too much work to aggregate them by total minutes in each combo, so just doing occurrences.

Jumbo
12-11-2008, 07:10 PM
I tried to analyze all the combinations that resulted in a negative cumulative (32 combinations @ 5 players = 160).

Based on actual minutes played, I developed how many occurrences across the 1800 minutes would be expected. Not surprisingly, the players with the best plus/minus in Jumbo's analysis had the best variation from expectation:

Sorted by minutes played which is interesting in and of itself:

Min Expect Actual Plus/Minus Player
265 23.6 13 10.6 Scheyer
261 23.2 14 9.2 Singler
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
171 15.2 13 2.2 Thomas
149 13.2 20 -6.8 Paulus
129 11.5 15 -3.5 McClure
122 10.8 23 -12.2 Williams
120 10.7 06 4.7 Zoubek
071 06.3 12 -5.7 Plumlee
065 05.8 16 -10.2 Pocius
026 02.3 05 -2.7 Czyz
006 00.5 01 -0.5 Johnson
1800 160 160 0.0 Total Duke

Sorted by best to worst, it negates some conventional wisdom that Elliott, or even McClure, is a lock down defender, and also provides some validation why certain players play and others do not:

Min Expect Actual Plus/Minus Player
265 23.6 13 10.6 Scheyer
261 23.2 14 9.2 Singler
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
120 10.7 06 4.7 Zoubek
171 15.2 13 2.2 Thomas
006 00.5 01 -0.5 Johnson
026 02.3 05 -2.7 Czyz
129 11.5 15 -3.5 McClure
071 06.3 12 -5.7 Plumlee
149 13.2 20 -6.8 Paulus
065 05.8 16 -10.2 Pocius
122 10.8 23 -12.2 Williams
1800 160 160 0.0 Total Duke

What I found most interesting though is how few times Zoubek appeared out of 32 (expected that Johnson and Czyz would have minimal) and one of these was a gimmick lineup with Scheyer at PG:

Paulus-Pocius-McClure-Thomas-Zoubek 2-3 (-1)
Paulus-Scheyer-Williams-Singler-Zoubek 2-3 (-1)
Paulus-Pocius-Williams-McClure-Zoubek (x2) 5-7 (-2)
Paulus-Pocius-Williams-Czyz-Zoubek 0-2 (-2)
Scheyer-Williams-Henderson-Singler-Zoubek 0-2 (-2)
Smith-Scheyer-McClure-Singler-Zoubek (x4) 7-10 (-3)

Only once do the names Smith, Henderson, Thomas or Czyz appear with Zoubek in a combo with negative plus/minus. The other names all appear 3-4 times, whcih would be expectd more for guys who play a lot of minutes like Singler and Scheyer and perhaps indicate McClure should not be paired with Z much.

Paulus and Zoubek combo is slow at two critical positions and having inexperienced Williams or non sterling Pocius as co-defender just makes it worse, but still it only happened in 6 of 32 worst combos.

Min Expect Actual Plus/Minus Player
006 00.5 01 -0.5 Johnson
026 02.3 05 -2.7 Czyz
120 10.7 06 4.7 Zoubek
208 18.5 10 8.5 Smith
207 18.4 12 6.4 Henderson
071 06.3 12 -5.7 Plumlee
265 23.6 13 10.6 Scheyer
171 15.2 13 2.2 Thomas
261 23.2 14 9.2 Singler
129 11.5 15 -3.5 McClure
065 05.8 16 -10.2 Pocius
149 13.2 20 -6.8 Paulus
122 10.8 23 -12.2 Williams
1800 160 160 0.0 Total Duke

Other observations, not unexpectedly, Paulus does worse paired with Scheyer or Pocius than Smith or Williams:

10 Paulus-Pocius
7 Paulus-Scheyer
2 Paulus-Smith
1 Paulus-Williams

Pocius-Williams is a bad combination on defensive end:

12 Pocius-Williams
1 Pocius-Czyz
1 Pocius-Henderson
1 Pocius-McClure
1 Scheyer-Pocius

11 Pocius-Williams
7 Scheyer-Williams
4 Williams-Henderson

Hard to track Elliott since he has played 1-2-3, so I excluded certain combos.

If I get some time, I will do a similar analysis on the top several combos that resulted in favorable cumulative plus/minus. Too much work to aggregate them by total minutes in each combo, so just doing occurrences.

I don't really understand your methodology. Could you summarize it? Also, how are you able to to draw conclusions about "defensive" contributions? Plus/minus numbers don't differentiate between whether someone is hurting/helping on offense or defense.

ACCBBallFan
12-11-2008, 07:11 PM
I tired to analyze all the combinations that resulted in a positive cumulative +2 or better (27 combinations @ 5 players = 135).

Based on actual minutes played, I developed how many occurrences across the 1800 minutes would be expected. Contrary to what happened on the analysis of worst combinations, surprisingly, the players with the best plus/minus varaiations are the reserves. On the other hand, perhaps it adds credence to the conventional wisdom that defense is more important than offense to coach K when deciding who plays:

Sorted by minutes played which is interesting in and of itself:

Min Expect Actual Plus/Minus Player
265 19.9 17 -2.9 Scheyer
261 19.6 18 -1.6 Singler
208 15.6 15 -0.6 Smith
207 15.5 11 -4.5 Henderson
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
129 09.7 12 2.3 McClure
122 09.2 12 2.9 Williams
120 09.0 08 -1.0 Zoubek
071 05.3 05 -0.3 Plumlee
065 04.9 08 3.1 Pocius
026 02.0 01 -1.0 Czyz
006 00.5 00 -0.5 Johnson
1800 135 135 0.0 Total Duke


Sorted by best to worst, it negates some conventional wisdom that plus/minus gets worse as Duke uses its bench. The reserves appear on the top 27 combos more than expected and the starters less:

Min Expect Actual Plus/Minus Player
065 04.9 08 3.1 Pocius
122 09.2 12 2.9 Williams
129 09.7 12 2.3 McClure
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
071 05.3 05 -0.3 Plumlee
006 00.5 00 -0.5 Johnson
208 15.6 15 -0.6 Smith
026 02.0 01 -1.0 Czyz
120 09.0 08 -1.0 Zoubek
261 19.6 18 -1.6 Singler
265 19.9 17 -2.9 Scheyer
207 15.5 11 -4.5 Henderson
1800 135 135 0.0 Total Duke


Perhaps the best way to interpret these somewhat surprising results is to merge the two. This does explain why some players play a lot and others not so much, and also increases my suspicion that Gerald in addition to Greg is playing with more of an injury than has been disclosed:

Best +/- Worst +/- Overall Player
-0.6 +08.5 +7.9 Smith
-2.9 +10.6 +7.7 Scheyer
-1.6 +09.2 +7.6 Singler
+2.2 +02.2 +4.4 Thomas
-1.0 +04.7 +3.7 Zoubek
-4.5 +06.4 +1.9 Henderson
-0.5 -00.5 -1.0 Johnson
+2.3 -03.5 -1.2 McClure
-1.0 -02.7 -3.7 Czyz
+1.8 -06.8 -5.0 Paulus
-0.3 -05.7 -6.0 Plumlee
+3.1 -10.2 -7.1 Pocius
+2.9 -12.2 -9.4 Williams
+0.0 -0.0 +0.0 Total Duke

It has to be remembered that these are variations relative to expectation in occurences based on minutes played rather than true plus/minus anlayses.

Indoor66
12-11-2008, 08:09 PM
I tired to analyze all the combinations that resulted in a positive cumulative +2 or better (27 combinations @ 5 players = 135).

Based on actual minutes played, I developed how many occurrences across the 1800 minutes would be expected. Contrary to what happened on the analysis of worst combinations, surprisingly, the players with the best plus/minus varaiations are the reserves. On the other hand, perhaps it adds credence to the conventional wisdom that defense is more important than offense to coach K when deciding who plays:

Sorted by minutes played which is interesting in and of itself:

Min Expect Actual Plus/Minus Player
265 19.9 17 -2.9 Scheyer
261 19.6 18 -1.6 Singler
208 15.6 15 -0.6 Smith
207 15.5 11 -4.5 Henderson
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
129 09.7 12 2.3 McClure
122 09.2 12 2.9 Williams
120 09.0 08 -1.0 Zoubek
071 05.3 05 -0.3 Plumlee
065 04.9 08 3.1 Pocius
026 02.0 01 -1.0 Czyz
006 00.5 00 -0.5 Johnson
1800 135 135 0.0 Total Duke


Sorted by best to worst, it negates some conventional wisdom that plus/minus gets worse as Duke uses its bench. The reserves appear on the top 27 combos more than expected and the starters less:

Min Expect Actual Plus/Minus Player
065 04.9 08 3.1 Pocius
122 09.2 12 2.9 Williams
129 09.7 12 2.3 McClure
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
071 05.3 05 -0.3 Plumlee
006 00.5 00 -0.5 Johnson
208 15.6 15 -0.6 Smith
026 02.0 01 -1.0 Czyz
120 09.0 08 -1.0 Zoubek
261 19.6 18 -1.6 Singler
265 19.9 17 -2.9 Scheyer
207 15.5 11 -4.5 Henderson
1800 135 135 0.0 Total Duke


Perhaps the best way to interpret these somewhat surprising results is to merge the two. This does explain why some players play a lot and others not so much, and also increases my suspicion that Gerald in addition to Greg is playing with more of an injury than has been disclosed:

Best +/- Worst +/- Overall Player
-0.6 +08.5 +7.9 Smith
-2.9 +10.6 +7.7 Scheyer
-1.6 +09.2 +7.6 Singler
+2.2 +02.2 +4.4 Thomas
-1.0 +04.7 +3.7 Zoubek
-4.5 +06.4 +1.9 Henderson
-0.5 -00.5 -1.0 Johnson
+2.3 -03.5 -1.2 McClure
-1.0 -02.7 -3.7 Czyz
+1.8 -06.8 -5.0 Paulus
-0.3 -05.7 -6.0 Plumlee
+3.1 -10.2 -7.1 Pocius
+2.9 -12.2 -9.4 Williams
+0.0 -0.0 +0.0 Total Duke

It has to be remembered that these are variations relative to expectation in occurences based on minutes played rather than true plus/minus anlayses.

Wayyyyyyyyyyyyyyyyyyyyyyyyy too many numbers for me to absorb, fathom or enjoy, :eek:

Jumbo
12-11-2008, 10:10 PM
I tired to analyze all the combinations that resulted in a positive cumulative +2 or better (27 combinations @ 5 players = 135).

Based on actual minutes played, I developed how many occurrences across the 1800 minutes would be expected. Contrary to what happened on the analysis of worst combinations, surprisingly, the players with the best plus/minus varaiations are the reserves. On the other hand, perhaps it adds credence to the conventional wisdom that defense is more important than offense to coach K when deciding who plays:

Sorted by minutes played which is interesting in and of itself:

Min Expect Actual Plus/Minus Player
265 19.9 17 -2.9 Scheyer
261 19.6 18 -1.6 Singler
208 15.6 15 -0.6 Smith
207 15.5 11 -4.5 Henderson
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
129 09.7 12 2.3 McClure
122 09.2 12 2.9 Williams
120 09.0 08 -1.0 Zoubek
071 05.3 05 -0.3 Plumlee
065 04.9 08 3.1 Pocius
026 02.0 01 -1.0 Czyz
006 00.5 00 -0.5 Johnson
1800 135 135 0.0 Total Duke


Sorted by best to worst, it negates some conventional wisdom that plus/minus gets worse as Duke uses its bench. The reserves appear on the top 27 combos more than expected and the starters less:

Min Expect Actual Plus/Minus Player
065 04.9 08 3.1 Pocius
122 09.2 12 2.9 Williams
129 09.7 12 2.3 McClure
171 12.8 15 2.2 Thomas
149 11.2 13 1.8 Paulus
071 05.3 05 -0.3 Plumlee
006 00.5 00 -0.5 Johnson
208 15.6 15 -0.6 Smith
026 02.0 01 -1.0 Czyz
120 09.0 08 -1.0 Zoubek
261 19.6 18 -1.6 Singler
265 19.9 17 -2.9 Scheyer
207 15.5 11 -4.5 Henderson
1800 135 135 0.0 Total Duke


Perhaps the best way to interpret these somewhat surprising results is to merge the two. This does explain why some players play a lot and others not so much, and also increases my suspicion that Gerald in addition to Greg is playing with more of an injury than has been disclosed:

Best +/- Worst +/- Overall Player
-0.6 +08.5 +7.9 Smith
-2.9 +10.6 +7.7 Scheyer
-1.6 +09.2 +7.6 Singler
+2.2 +02.2 +4.4 Thomas
-1.0 +04.7 +3.7 Zoubek
-4.5 +06.4 +1.9 Henderson
-0.5 -00.5 -1.0 Johnson
+2.3 -03.5 -1.2 McClure
-1.0 -02.7 -3.7 Czyz
+1.8 -06.8 -5.0 Paulus
-0.3 -05.7 -6.0 Plumlee
+3.1 -10.2 -7.1 Pocius
+2.9 -12.2 -9.4 Williams
+0.0 -0.0 +0.0 Total Duke

It has to be remembered that these are variations relative to expectation in occurences based on minutes played rather than true plus/minus anlayses.

I'm still confused -- sorry.

ACCBBallFan
12-12-2008, 11:00 AM
Just another way of looking at your plus/minus combinations.

To net it out, looking at the top 27 combinations of +2 or better and all 32 with negative plus/minus, number of times a players name appeared realtive to minutes played, players who performed better than expected in order are:

Overall Player
+7.9 Smith (208)
+7.7 Scheyer (265)
+7.6 Singler(261)
+4.4 Thomas (171)
+3.7 Zoubek (120)
+1.9 Henderson (207)
-1.2 McClure (129)
-5.0 Paulus (149)
-6.0 Plumlee (71)
-7.1 Pocius (65)
-9.4 Williams (122)
+0.0 Total Duke (1800)

So the minutes in parenthesis correspond pretty well with overall performance and if anything Elliott is perhaps being given more opportunity than his performance warrants. Not saying this is wrong, just that people complaining he is not playing enough may be overstating their case.

Henderson in addiiton to Paulus IMO is more injured than has been publicly admitted.

shadowfax336
12-12-2008, 11:28 AM
I sorta get it...
are you trying to see who leads to the extreme cases of productivity per minute vs the extreme cases of unproductivity?

You can't say thats "offense" vs "defense" though, because having large negative differentials per minute might just mean that they can't score at all. Which has occassionally been the case with Email and often been the case with McClure (although there are signs of life there)

Devilsfan
12-13-2008, 08:27 AM
It's seems funny but the top nine are the ones I would have on almost any ACC team. Maybe I'd throw in Elliot based on his potential.

ACCBBallFan
12-13-2008, 10:09 AM
I sorta get it...
are you trying to see who leads to the extreme cases of productivity per minute vs the extreme cases of unproductivity?

You can't say thats "offense" vs "defense" though, because having large negative differentials per minute might just mean that they can't score at all. Which has occassionally been the case with Email and often been the case with McClure (although there are signs of life there)

Correct on all counts - More of a who outscored the Opponent or got outscored regardless of whether that was due to offense or defense, and most likely a combination of both.

jv001
12-13-2008, 12:06 PM
Man I'm really confused.

ACCBBallFan
12-15-2008, 05:27 PM
Just another way of looking at your plus/minus combinations.

To net it out, looking at the top 27 combinations of +2 or better and all 32 with negative plus/minus, number of times a players name appeared realtive to minutes played, players who performed better than expected in order are:

Overall Player
+7.9 Smith (208)
+7.7 Scheyer (265)
+7.6 Singler(261)
+4.4 Thomas (171)
+3.7 Zoubek (120)
+1.9 Henderson (207)
-1.2 McClure (129)
-5.0 Paulus (149)
-6.0 Plumlee (71)
-7.1 Pocius (65)
-9.4 Williams (122)
+0.0 Total Duke (1800)

So the minutes in parenthesis correspond pretty well with overall performance and if anything Elliott is perhaps being given more opportunity than his performance warrants. Not saying this is wrong, just that people complaining he is not playing enough may be overstating their case.

Henderson in addiiton to Paulus IMO is more injured than has been publicly admitted.After a lot of metrics, hard to form any conclusions.

Nolan Smith's relative success is more a factor of Greg playing injured, than Nolan playing so well.

OTOH, it validates the minutes played by whom are appropriate but on the other it supports arguments that the starters play too many minutes when game is in hand in vein that reserves would have fared better vs. the weaker competition too.

So while I thought beforehand it might be useful, in retrospect, I will stick to enjoying Jumbo's analyses of each combination.

Jumbo
12-18-2008, 12:35 AM
Now updated through the UNC-Asheville game ...

ACCBBallFan
12-18-2008, 12:43 PM
Jumbo, while I agree individual game +/- is more skewed than full season, could you also post the individual +/- from UNC -Asheville.

The reserves played a lot more and these stats might be interesting to peruse, to see what progress they are making which may get masked in the overall.

-jk
12-18-2008, 01:36 PM
Jumbo, while I agree individual game +/- is more skewed than full season, could you also post the individual +/- from UNC -Asheville.

The reserves played a lot more and these stats might be interesting to peruse, to see what progress they are making which may get masked in the overall.

He does a separate thread for them. http://www.dukebasketballreport.com/forums/showthread.php?t=13279

-jk

ACCBBallFan
12-19-2008, 02:44 PM
He does a separate thread for them. http://www.dukebasketballreport.com/forums/showthread.php?t=13279

-jkThanks, Jk.

ice-9
12-19-2008, 08:48 PM
Zoubek's numbers are eye popping. He's really come into form this season...go Z!

Jumbo
12-21-2008, 08:22 PM
Now updated through the Xavier game ...

Jumbo
01-01-2009, 11:58 AM
Now updated through the Loyola game ...

CameronCrazy'11
01-01-2009, 11:42 PM
Zoubek's numbers are eye popping. He's really come into form this season...go Z!

A lot of that is because he starts, but doesn't play that many minutes total. That means that most of his minutes are with the rest of the starters also in the game, which is when Duke typically does the most damage. Still though, he's been an excellent player in his own right.

Jumbo
01-09-2009, 01:45 AM
Now updated through the Davidson game ...

Jumbo
01-11-2009, 10:05 PM
Now updated through the Florida State game ...

Jumbo
01-15-2009, 09:28 AM
Now updated through the Georgia Tech game ...

Hancock 4 Duke
01-15-2009, 02:37 PM
What is a Cumulative plus minus? Like what would this mean: 1060-712

-jk
01-15-2009, 03:33 PM
Start at the beginning and read a bit of the thread. It gets pretty thoroughly explained.

-jk

Jumbo
01-20-2009, 12:14 AM
Now updated through the Georgetown game ...

Jumbo
01-20-2009, 11:29 PM
Now updated through the NC State game ...

Jumbo
01-24-2009, 09:51 PM
Now updated through the Maryland game ...

Jumbo
01-30-2009, 12:15 AM
Now updated through the Wake Forest game ...

Jumbo
02-01-2009, 07:37 PM
Now updated through the Virginia game ...

Jumbo
02-01-2009, 11:28 PM
Here's an interesting stat. This season, Coach K has already used 153 different player combinations. All of last season, he only used 136.

Jumbo
02-06-2009, 12:01 AM
Now updated through the Clemson game ...

Jumbo
02-07-2009, 05:45 PM
Now updated through the Miami game ...

Truth
02-10-2009, 11:07 AM
Sorry if I missed this... what is the difference between the cumulative +/- and the "Net +/-" posted at the end of Jumbo's initial post?

Hancock 4 Duke
02-16-2009, 06:15 PM
Now updated through the Miami game ...

Individuals
Kyle Singler 1,411-1,023 (+388)
Jon Scheyer 1,434-1,049 (+385)
Nolan Smith 1,066-763 (+303)
Gerald Henderson 1,268-972 (+296)
Brian Zoubek 665-428 (+237)
David McClure 705-572 (+133)
Greg Paulus 773-688 (+85)
Lance Thomas 717-640 (+77)
Elliot Williams 407-368 (+39)
Martynas Pocius 220-196 (+24)
Miles Plumlee 239-232 (+7)
Steve Johnson 71-71 (0)

[/code]

I don't mean to be a "stickler", but you didn't mention Olek Czyz on this list.

Jumbo
02-18-2009, 12:04 AM
Now updated through the UNC game ...

Jumbo
02-18-2009, 12:04 AM
Now updated through the Boston College game ...

bird
02-18-2009, 07:34 AM
The NY Time Mag Shane Article mentions a method to adjust +/- stats to account for the "playing with other good players" effect, but said that Houston considers the adjustments secret. Any ideas?

Jumbo
02-21-2009, 07:12 AM
Now updated through the St. John's game ...

Jumbo
02-23-2009, 06:09 PM
Now updated through the second Wake Forest game ...

pfrduke
02-23-2009, 06:37 PM
The NY Time Mag Shane Article mentions a method to adjust +/- stats to account for the "playing with other good players" effect, but said that Houston considers the adjustments secret. Any ideas?

Here's one way to do it (http://www.countthebasket.com/blog/2008/06/01/calculating-adjusted-plus-minus/). It's, um, cumbersome (to say the least). Also, I'm not 100% sure, but it may require a more self-enclosed league like the NBA than the 341 team world of D-I hoops. Possibly could set it up to run ACC-only data. But you also need the play-by-play data, which not all D-I schools provide.

There's huge, expansive room for improvement of statistical analysis of college basketball for anyone with the time, energy, and funding to do it. Lamentably, I only have the energy....

Jumbo
02-26-2009, 11:46 PM
Now updated through the second Maryland game ...

tele
02-28-2009, 09:02 AM
I'd like to see more of the lineups with Thomas and Zoubek or even Thomas and Plumlee at the same time. I guess +9 and +6 aren't very high numbers but seems like it would be good to show this some before the post season.

Jumbo
03-01-2009, 09:10 PM
Now updated through the second Virginia Tech game ...

Jumbo
03-04-2009, 07:31 PM
Now updated through the second Florida State game ...

Jumbo
03-10-2009, 11:30 PM
Now updated through the second North Carolina game ...

Jumbo
03-14-2009, 02:46 PM
Now updated through the Boston College ACC Tournament game ...

Jumbo
03-14-2009, 09:13 PM
Now updated through the Maryland ACC Tournament game ...

Jumbo
03-15-2009, 11:42 PM
Now updated through the Florida State ACC Tournament game ...

bjornolf
03-20-2009, 05:37 PM
Any way you can send this to Mike Gminski? I noticed during one of the Duke games he was commentating, he said that Dave McClure's stats don't show all he does for this team. He then said "I'd love to see the plus-minus when that guy's in the lineup."

Just a thought.

Jumbo
03-20-2009, 05:53 PM
Any way you can send this to Mike Gminski? I noticed during one of the Duke games he was commentating, he said that Dave McClure's stats don't show all he does for this team. He then said "I'd love to see the plus-minus when that guy's in the lineup."

Just a thought.

How, exactly, would you like me to do that? ;) I supposed you could get in touch with Duke, CBS or FSN, send them an e-mail, and link this thread.

Jumbo
03-20-2009, 08:46 PM
Now updated through the Binghamton game ...

Jumbo
03-22-2009, 01:23 PM
Now updated through the Texas game ...

pfrduke
03-30-2009, 10:30 AM
Paulus-Scheyer-Henderson-Singler-Thomas (x34) 125-133 (-8)

Interesting to note that last season, this lineup was our most effective (and most used) non-Nelson lineup:

Paulus-Scheyer-Henderson-Singler-Thomas (x32) 101-60 (+51).

In one season, that group of five played a little better on offense but were 73 points worse on defense. If I get the chance, I'll look a little deeper to see when those respective lineups were used to see if there's a readily apparent explanation for this (e.g., in 2007-08, that lineup played a lot against Eastern Kentucky and Princeton, and not much in the ACC, and it was vice versa in 2008-2009), but going from +51 to -8 in a single season is a shocking drop off.

Jumbo
03-30-2009, 09:03 PM
Now updated with final 2008-09 totals ...

-jk
03-30-2009, 09:49 PM
And thanks for the effort!

-jk

Jumbo
03-31-2009, 09:10 PM
And thanks for the effort!

-jk

Thanks!

cspan37421
04-02-2009, 11:40 AM
I can't decide if the most underappreciated player on this past year's team was Greg Zoubek or Brian Koubek. :D