I've been somewhat out of commission posting-wise because of the holidays and work, but I've been thinking of a simpler version of my "Differential Production" (DP) statistic. I'm sure the idea isn't new: calculate a reverse boxscore for a game. Next to a player's name will be the statistics of
all the players he was guarding. The totals of the reverse boxscore should equal the team totals of the opposing team.
A few notes about calculating the reverse boxscore:
- Unlike in the DP calculation, I don't divide blame for opponent field goals (or free throws) made. I try to determine who is the most culpable for a given play. On switches and fastbreaks, this is somewhat subjective.
- Rebounds and assists are assigned to whichever player was most recently guarded by or guarding the player who attained the rebound or assist.
- Steals, blocks, turnovers, and fouls drawn are always assigned directly.
In the reverse boxscore, turnovers and fouls (drawn) are "good," while all other categories are "bad." The goal is to achieve a low opponent FG percentage and low totals in other categories. The Cleveland game on 12/7 was an interesting case study, because while the Sixers excelled on offense, they weren't great on defense.
Here is the reverse boxscore for the Sixers in the 12/7 game vs. the Cavs.
|
MP |
FGM-A |
3PM-A |
FTM-A |
OR-TR |
AST |
PF |
ST |
TO |
BL |
PTS |
| AI |
36 |
3-6 |
0-1 |
2-2 |
1-2 |
0 |
6 |
1 |
2 |
1 |
8 |
| EB |
30 |
6-11 |
0-1 |
3-4 |
0-5 |
0 |
2 |
0 |
2 |
0 |
15 |
| SH |
32 |
3-13 |
1-2 |
3-4 |
3-7 |
4 |
3 |
0 |
1 |
0 |
10 |
| JH |
40 |
8-15 |
3-7 |
0-0 |
0-2 |
8 |
5 |
1 |
1 |
1 |
19 |
| JM |
30 |
3-6 |
0-0 |
0-0 |
0-3 |
5 |
0 |
2 |
0 |
1 |
6 |
| TY |
28 |
8-14 |
3-5 |
0-1 |
1-6 |
1 |
2 |
0 |
2 |
0 |
19 |
| LW |
21 |
5-7 |
0-0 |
1-2 |
0-1 |
3 |
4 |
0 |
2 |
0 |
11 |
| ET |
15 |
3-5 |
0-1 |
2-2 |
0-2 |
1 |
0 |
1 |
1 |
0 |
8 |
| AN |
8 |
0-1 |
0-0 |
1-2 |
0-1 |
0 |
0 |
0 |
1 |
0 |
1 |
| Tot |
240 |
39-78 |
7-17 |
12-17 |
5-29 |
22 |
22 |
5 |
12 |
3 |
97 |
A few notes about the reverse boxscore:
- Spencer Hawes was even better than his stats indicated. I couldn't believe how many shots he challenged off switches. He had his usual problems covering the pick-and-roll, but he more than made up for it with his rebounding and help defense.
- Thad Young was high scorer in the game, but tying for 2nd-high scorer (with Lou) was Bizarro Thad Young, along with Bizarro Jrue Holiday.
- Iguodala impressively "fouled out" his opponents.
With these stats and the conventional boxscore stats, we can compute PER and opponent PER for the game. I wrote a program to do this, making a few assumptions along the way (if you want to know what they were, ask me offline -- it's probably too boring for most people). But here are the results:
|
PER |
OPP
PER |
| AI |
18.4 |
4.6 |
| EB |
18.3 |
12.6 |
| SH |
21.4 |
6.0 |
| JH |
15.9 |
17.1 |
| JM |
4.6 |
17.3 |
| TY |
50.2 |
15.9 |
| LW |
37.7 |
15.5 |
| ET |
-1.1 |
23.1 |
| AN |
21.5 |
-6.9 |
| Tot |
20.9 |
12.5 |
Some comments about the PER results:
- Thad and Lou had ridiculously high PERs, leading the team to an impressive total PER of 20.5 (PER is normalized to 15.0).
- Iguodala and Hawes led the way on defense to a team opponent PER of 12.5.
- Looking closely at the PER calculation, one can see why SFs typically have low opponent PERs, because SFs get relatively few rebounds and assists. PFs and Cs see higher opponent PERs because of rebounds (and blocks), while PGs see higher opponent PERs because of assists.
- I was surprised by Meeks' low PER number, but then I saw that he didn't score overwhelmingly much or efficiently (16 points on 13 shots in 30 minutes), and his 3 turnovers outnumbered all his other ancillary positive stats combined (2 rebounds+assists+steals+blocks).
That's all I have for now. Thoughts and comments welcome, as usual. Depending on how the Boston game goes, I might have DP statistics for that game.
Wow. So Thad's man scored 19 on 14 shots and he still beat him in PER by 34.3. Unbelievable. Rough game for Turner and I never would've guessed Nocioni's differential was so good.
Great stuff, again, Statman. Thanks for putting this together.
If you want another "I can't believe that stat", look at 82games.com's PER differential for the year, and look at Young's. We're starting to get to the point where we're at a decent enough sample size, too.
Those stats don't even include the Cavs game. I'll be interested to see if Thad can keep it up vs. the Celtics (historically, he's struggled against them, 38% FG for his career).
By the way, I missed it the first time through, but on re-watching the game, I couldn't believe the 5-minute debate between Eric Snow and Zumoff about LeBron. LeBron's teammates made him better? The same teammates that are challenging for the worst record in the league now? I think Snow is decent as a commentator (better than Salmi and Pinckney, for example), but he definitely had an old axe to grind. And it was weird to see the play-by-play guy veer into opinion, but Snow's argument was so preposterous that it couldn't be helped ...
I could see the argument that the Cavs had players that were more complimentary for Lebron's skill set and in that way made him "better"
Yup. Thad's playing his way into that $55-$60M extension.
Quick question about methodology: I don't think the Sixers played any zone in that game, but how do you assign the credit/blame when they do? If a guard comes streaking in from the perimeter for an offensive rebound, is it his man, or the closest man?
Eddie Jordan
For a zone, I would assign credit/blame to the nearest defender. If the zone is played properly, there should be somebody in the vicinity on every shot ;-). This was the general rule that I followed on most shots, with the exception of the "blow by" play (AKA the Lou Williams special -- bad defense causes somebody else to foul) and the "turnover leading to uncontested fastbreak" play (which never happened in this game).
For rebounds/assists, I assigned credit/blame to whoever was guarding or guarded by that player most recently.
Cool. So what's your feeling, is this like a shorthand version of DP, but not quite as illuminating? I think this is probably easier to digest, people are used to looking at box scores.
The classical boxscore stats are much easier to digest, and the reverse boxscore is faster and easier to tabulate than DP. By nature, though, the boxscore doesn't capture partial credit/blame and it doesn't account for the redundancy between made field goals and assists (though the PER calculation itself attempts to do this). But I would say the reverse boxscore captures about 80% of what is going on in individual defense, which is one of the primary reasons I came up with DP ... so maybe I will wind up doing the reverse boxscore more often (since it is faster to tabulate).
Awesome Statman, it gives a good look at Jodie's relative effect on a game may be through what he does to help others. If he's not shooting ultra-efficiently from three, his numbers won't be that high. This is simply because he doesn't really do anything else right now.
If Iguodala's defense doesn't impress people, they have something wrong with them. He routinely excels in these stats, which aren't some crazy formula either. You simply watch the game and assign blame if a player's man scores. There's no real bias in this or DP, and his defense often is excellent. Of course, Cleveland doesn't have a ton of great wing guys, so he can really excel. That stat doesn't weight him having to guard a superstar.
I am very discouraged with Turner, but I can't really say anything more than that.