Back in June, I wrote a post detailing a
new type of statistic. I didn't have a name for it then, but I think a good name would be
Differential Production (DP), divided into
Differential Offensive Production (DOP) and
Differential Defensive Production (DDP). I explained the general idea in the original post, but the idea in a nutshell is this: a team's offensive rating and defensive ratings are defined in terms of points per 100 possessions. For example, the 76ers scored 87 points in 91 possessions vs. the Heat, for an offensive rating of (87/91)*100 = 95.6. DP determines the
individual contributions to a team's offensive and defensive ratings, based on the assumption that an "average" team will have offensive/defensive ratings of 100 (this isn't strictly true, but it makes the statistic easier to understand). So, the 76ers scored 87 points in 91 possessions for a net of -4 on offense. Who contributed to that -4 total? And the same question holds, perhaps more interestingly, on defense, where the Sixers were a net of -5 (97 points in 92 possessions). Sites like 82games.com show individual offensive and defensive ratings, but that only reflects the team's points per possession while the player was on the floor, not the individual player's contribution to it.
As noted in the original post, there are several ways to assign differential production. Remember that everything is relative to 1 point per possession, so a made 2-point basket is +1, for example. Here are the possible ways to accumulate DP:
On offense:- A made basket is either +1 (for a 2-pointer) or +2 (for a 3-pointer) and is divided between the shooter and the passer depending on how important the pass was. I divide into quarters at most, for relative simplicity.
- 2 made free throws is a +1. 1 of 2 free throws is recorded as +0.5 under made shots and -0.5 under missed shots (net of 0).
- A turnover is -1. In some circumstances, some blame might go to the other player (e.g., someone fumbling a good pass -- even though the official turnover always goes to the passer).
- A missed field goal attempt resulting in a defensive rebound is -1. A missed field goal attempt resulting in an offensive rebound is -0.5. This is intuitively not right, but is necessary for symmetry with the defensive stats.
- An offensive rebound is +0.5. Partial credit can be assigned if multiple players were involved (e.g., a tip to another player).
On defense:- A made basket is either -1 or -2 (for 2 or 3-pointers), and blame can be divided depending on who is responsible for the made basket.
- A contested shot resulting in a miss is +0.5 for the defender (I label this a "stop"). An uncontested shot resulting in a miss is +0.5 for "team" (essentially, no one caused the miss).
- A defensive rebound is +0.5. An offensive rebound allowed is -0.5. Partial credit can be assigned if multiple players were involved (e.g., a tip to another player).
- A forced turnover is +1.
- A shooting foul is automatically a -1 or -2 for the defender (or divided among multiple defenders). If the shooter misses the first free throw, the "team" is credited with +1 for stop. If the shooter misses the last free throw, the "team" is credited with +0.5 for a stop, and then rebounding is assigned as above.
Without further ado, here is the table for the 76ers-Miami game.
tables
|
OFFENSE |
|
Made
Shots |
Missed
Shots |
Off
Reb |
Asst |
TO |
Off
Tot |
| Iguodala |
3.75 |
-6 |
0.75 |
1.75 |
-3 |
-2.75 |
| Brand |
5.5 |
-5 |
2.25 |
0 |
-1 |
1.75 |
| Hawes |
0 |
-2.5 |
0.5 |
0.5 |
0 |
-1.5 |
| Holiday |
3.5 |
-6.5 |
0.75 |
1 |
-5 |
-6.25 |
| Kapono |
0.5 |
-0.5 |
0.5 |
0 |
-1 |
-0.5 |
| Turner |
7.75 |
-3 |
0.5 |
0.75 |
-2 |
4 |
| Young |
7 |
-2 |
0.75 |
0 |
-1 |
4.75 |
| Williams |
8.5 |
-8 |
0 |
2 |
-3 |
-0.5 |
| Speights |
0.5 |
-2.5 |
0 |
0 |
0 |
-2 |
| Nocioni |
4 |
-4.5 |
0.5 |
0.5 |
-1 |
-0.5 |
| Songaila |
0 |
0 |
0 |
0 |
0 |
0 |
| Team |
0 |
0 |
0.5 |
0 |
-1 |
-0.5 |
| Totals |
41 |
-40.5 |
7 |
6.5 |
-18 |
-4 |
|
|
|
|
|
|
(87/91) |
tables
|
DEFENSE |
TOT |
|
Scored
On |
Stops |
Def
Reb |
ORA |
TO
Forced |
Def
Tot |
| AI |
-8.75 |
3 |
3.25 |
-0.5 |
3 |
0 |
-2.75 |
| EB |
-10.5 |
3.25 |
2.5 |
-1 |
3.25 |
-2.5 |
-0.75 |
| SH |
-1.5 |
0 |
1.5 |
0 |
0 |
0 |
-1.5 |
| JH |
-6 |
1.5 |
0 |
0 |
1.75 |
-2.75 |
-9 |
| JK |
-1.5 |
1 |
0 |
0 |
1 |
0.5 |
0 |
| ET |
-6 |
0.75 |
3 |
-0.5 |
0 |
-2.75 |
1.25 |
| TY |
-7.5 |
1 |
1 |
-1 |
2 |
-4.5 |
0.25 |
| LW |
-8.25 |
0.5 |
0.75 |
0 |
2 |
-5 |
-5.5 |
| MS |
-1 |
0.5 |
0 |
0 |
0 |
-0.5 |
-2.5 |
| AN |
-4 |
2 |
2.5 |
-1 |
1 |
0.5 |
0 |
| DS |
-3 |
0 |
0 |
0 |
0 |
-3 |
-3 |
| TM |
0 |
11 |
2 |
0 |
2 |
15 |
14.5 |
| Tot |
-58 |
24.5 |
16.5 |
-4 |
16 |
-5 |
-9 |
|
|
|
|
|
|
(97/92) |
|
Before I discuss some of the analysis, here are some ways to interpret the results:
- Net Points = DP + (Possession Differential). For the 76ers, it was (-10)=(-9)+(-1).
- Turnovers should equal the turnover column in a boxscore, but sometimes I assign blame (on offense) or credit (on defense) to an individual for a 24-second violation. In boxscores, 24-sec. violations are team turnovers and don't show up.
- To get the total points for a team, the formula is [(Made Shots)+(Assists)-(Extra Points)]*2 + (Extra Points). "Extra Points" are the additional point for a 3-pointer or an and-one (or a technical free throw).
- Multiplying the rebound columns by 2 will come close to the boxscore total, but remember that I sometimes assign partial credit for rebounds, and the boxscore doesn't count team rebounds.
Now some thoughts on the results:
- The only positive DPs for the game were Turner (+1.25) and Thad (+0.25).
- Thad's +4.75 was highest on offense (Turner was second at +4), but he gave almost all of it back on defense (-4.5).
- Worst players on defense were Lou (-5), Thad (-4.5), and Songaila (an impressive -3 in about 2 minutes).
- Worst player of the game by far was Jrue (-9), but he was actually a +5 for the 4th quarter, 3rd best quarter of the game behind Iguodala's +5.75 in the 1st and Turner's +5.25 in the 4th). That shows how poorly Jrue played in the 1st 3 quarters (-5.75 in the 3rd and -5.25 in the 1st only trailed Lou's -6 in the 3rd).
- Jrue and Lou together were -11.75 in the 3rd. If you want to know why Miami blew the game open in that quarter, look no further.
- Iguodala had a strong first quarter (+5.75) but was poor thereafter (-8.5). In fairness, he was always guarding either LeBron or Wade, so his defensive numbers were probably lower than usual (still decent at 0).
- Nocioni acquitted himself fairly well, actually scored one of two positive DDPs (surprise: Kapono was the other). Positive ODPs went to Brand (+1.75) and the aforementioned Turner and Young.
I've given the full spreadsheet with breakdowns by quarter and by half to Brian, who can distribute it to anyone who wants to see it. As always, comments and questions welcome.
Editor's Note: If you'd like to download the spreadsheet, which is broken down by quarter and half,
click here.
Unbelievable job, Statman. I really like the 1-to-1 relationship w/ PPP.
Were you able to compile the data just from watching the game live, did you rewatch it, or liberal use of the rewind on the DVR. I'd love to see what your graph paper looks like.
One question, on a breakaway dunk, uncontested, do you assigned the scored points to the player responsible for the turnover, if it is a turnover? On a long outlet pass, does the player who failed to get back on defense get the points?
For the results, interesting to see Nocioni as a positive on the defensive end. Also, was Thad responsible for most of the James Jones threes?
Great, great job. This is the type of stuff the Sixers' front office should be doing.
Good questions. Here are some answers:
Were you able to compile the data just from watching the game live, did you rewatch it, or liberal use of the rewind on the DVR. I'd love to see what your graph paper looks like.
I actually watched the game live with a friend and then did the stat breakdown later via DVR. I needed to rewind several times to assign proper credit/blame on defense (if there's switching involved, it's not always apparent at first glance). I took my notes by hand on a pad of paper because I wanted the ability to write two columns easily (though a laptop would have been better for subsequent search). You don't want to see the original notes ;-)
One question, on a breakaway dunk, uncontested, do you assigned the scored points to the player responsible for the turnover, if it is a turnover? On a long outlet pass, does the player who failed to get back on defense get the points?
If the turnover results in a breakaway, then yes, the player who caused the turnover gets the -1. Jrue had a couple of those in the 3rd quarter (and it was his man, Wade, who scored, so not too difficult). More of a gray area is when there is a defender back. On another Jrue turnover, Iguodala was forced to guard a 2-on-1 with LeBron and Arroyo and just missed a block on Arroyo's lay-in. So I assigned Jrue a -0.75 and Iguodala a -0.25 for that play. Failing to get back would result in most of the blame, depending on how realistic a shot the nearest defender has in stopping the shot.
For the results, interesting to see Nocioni as a positive on the defensive end. Also, was Thad responsible for most of the James Jones threes?
Here's how I scored the Jones threes:
* Songaila -2
* Nocioni -1.5, Jrue -0.5 (Nocioni's man, Jrue nearest)
* Thad -2
* Lou -2
* Thad -1.5, Iguodala -0.5 (Iguodala caught on pick, Thad overhelps)
* Thad -1, Brand -1 (Snow criticized Thad heavily for this, but the breakdown began when Brand lost his man, forcing Turner to rotate to Brand's man next to the basket, forcing Thad to guard two players. Thad chose to guard Stackhouse, leaving Jones open.)
So, while Thad was primarily responsible, it wasn't as egregious as the famous Matt Bonner game a couple years ago.
Excellent work man. In this case the stats seem to really tell the story.
This is awesome, great work. Are you going to be doing this throughout the year at all?
Thanks (to you and everyone else) for all the compliments.
I'll probably do this a few times throughout the year, on "interesting" games. (Who would care if the Sixers get blown out?) The initial note-taking and subsequent entering into the spreadsheet takes about 30 minutes per quarter. Maybe if I can get three volunteers to take a quarter each (and teach them how to assign the points), we can do it more often ...
Wow, great job Statman!!!. It's unbelievable how all of it mirrors exactly what we saw in the game (Turner and Young played well, Brand and Nocioni were solid, Holiday waas horrible through 3 quarters).
Keep it up!!!
Statman, this is very neat! It gives the whole +/- stat a whole new level of context.
Interesting breakdown Statman.
OT, this is a link to bold and timely news out of Sixerland:
http://basketball.realgm.com/src_wiretap_archives/69824/20101029/76ers_brass_changes_stance_on_season/
""Their strategy seems to have changed from "we can win with this group" to "we want to see what we have and make changes." They, meaning president Rod Thorn to general manager Ed Stefanski and on down through the front-office guys to Collins, all realize that this is not a team that they are going to try and tweak to make the playoffs. It appears they believe that a player added here or there really isn't going to make a difference," Cooney wrote."
...
Official statements (of the obvious) like this are important. Especially early on in the tenure of a new coach and President. In fact this could lead to an Iguodala trade and another "Thorn special" lottery appearance.
interesing. how do you account for your own bias? on defense especially on close calls, you need a second opinion. also on breakaways with one defender back who fouls shouldn't get any points assigned. your cause and effect analysis "Thad -1.5, Iguodala -0.5 (Iguodala caught on pick, Thad overhelps)
* Thad -1, Brand -1 (Snow criticized Thad heavily for this, but the breakdown began when Brand lost his man, forcing Turner to rotate to Brand's man next to the basket, forcing Thad to guard two players. Thad chose to guard Stackhouse, leaving Jones open.)" is subject yo your interpretation of who is supposed to be doing what. Are you sure you know the sixers defensive rotations that well? I didn't see the game so I really can't comment on how accurate your numbers are. Also what is an uncontested shot? if a player is playing off his man as he is supposed to do in certain situations shouldn't he get credit for doing the right thing? how can base you crediting a shooter for missing a shot on whther the team gets the rebound or not? a missed shot is a missed shot.
mchezno,
Like personal fouls in the NBA, some things are subjective. What he is capturing is subjective to a large extent.
The best thing he could do is make a list of standardized rules and follow them strictly. As time evolves, the rules evolve.
There is a degree of subjectivity to this statistic, no question. I do believe, however, that my (subjective) attempt to assign credit/blame for every play captures more than many individual defensive statistics for the NBA, which either (a) do not account for man-to-man defense at all [measuring only stats like steals, blocks, and defensive rebounds] or (b) make generic assumptions about who is guarding whom based on position.
But you're right, how much you trust this statistic does depend on how much you trust the statkeeper.
Players are not given "credit" if their team rebounds their own missed shot. It's more like they are doubly penalized if the other team rebounds the missed shot. The idea is as follows. On defense, a stop is a +1, and I assign a +0.5 for the defender and +0.5 for the rebounder. For symmetry, on offense I assign a -0.5 for the missed shot and another -0.5 for the rebound. However, it is in most cases unfair to assign a -0.5 to the offensive counterpart to the defensive rebounder, so I assign both -0.5's to the shooter. If you always assign -1 for a missed shot, then an offensive rebound would be worth +1 and offensive/defensive rebounds would be weighted differently.
Sorry for formatting and possible errors. Here are the player rankings per minute according to Win score, Games score(simplified Hollinger PER) and Statman's DP.
I wanted to see the rankings so I made this. Just figured I'd share.
Starters WS Game Score Statman
Andre Iguodala 5 6 6
Elton Brand 3 4 5
Jrue Holiday 10 7 9
Spencer Hawes 7 10 7
Jason Kapono 8 8 3
Reserves
Louis Williams 6 5 8
Evan Turner 1 2 1
Thaddeus Young 2 1 2
Andres Nocioni 4 3 3
Darius Songaila 9 9 10
Marreese Speights 11 11 11
And the actual totals per 48 minutes just in case someone wanted them... (please correct me if I missed anything)
And WS is not position adjusted as it typically is because I am lazy.
Starters WS Game Score Statman
Andre Iguodala 7.400428266 10.85396146 -3.391862955
Elton Brand 15.59055118 17.00787402 -1.062992126
Jrue Holiday -6.765857478 1.353171496 -20.29757244
Spencer Hawes -3.364485981 -5.719626168 -5.046728972
Jason Kapono -3.716129032 1.11483871 0
Reserves
Louis Williams -0.711111111 11.09333333 -7.822222222
Evan Turner 17.2173913 22.85217391 1.956521739
Thaddeus Young 16.73337637 26.02969658 0.46481601
Andres Nocioni 14.47236181 19.77889447 0
Darius Songaila -4.832214765 -3.865771812 -28.99328859
Marreese Speights -46.02739726 -24.98630137 -32.87671233
Interesting, thanks. Is Win Shares both offensive and Defensive Win Shares? As far as I can tell, Game Score doesn't account for individual defense, while Defensive Win Shares assigns everyone on the court equal credit/blame for a missed/made shot.
There is an interesting point from the writeup on defensive efficiency (which goes into Defensive Win Shares), that forcing a missed shot is harder than getting a defensive rebound. So theoretically, I could tweak my statistic to assign, say, +0.667 for the defender of a missed shot and +0.333 for the defensive rebounder. But again, that would wind up weighting offensive rebounds more than defensive rebounds (because an offensive rebound would be +0.667 to balance the -0.667 for a missed shot).
Win score, not win shares. Win score is the simple box score version of wins produced, so basically wins produced without the team defense adjustment and more rounding. (probably more to it)
As for defense, win score takes into account PFs, DReb, steals. That is it w/respect to defense, which is all the box score has really.
Win shares is a bit too complicated to do in 10-15 minutes in excel.
As for the missed shot thing, if you read up on PER and game score I think Hollinger does the exact thing you say in that the DRebounder only gets .33 points and an oreb gets .667 or something like that.
I think that simplicity has value, so watch the tweaks...
Awesome job, Statman. It really does paint a pretty accurate picture of what we all thought we saw out there. Iggy's defense dropped in the 3rd when Wade was scoring on him.