The Iverson Effect

[Mike G]

Why do certain players with bad shooting percent and/or lots of turnovers get so many minutes and take so many shots? Sometimes they are All-Stars. And sometimes with highly successful teams.

Here's an abbreviated summary of the 2003-04 Sixers, by minutes played.

Code: Select all

Sixers '04  G   MPG   PER    TS%    TO%   Usg%   WS/48
Snow        82  36   13.5   .494   18.0   16.5   .090
K Thomas    74  37   15.9   .527   15.3   19.6   .118
Dalembert   82  27   16.1   .566   13.0   14.1   .151
McKie       75  28   13.9   .550   14.1   16.2   .108
Iverson     47  43   19.3   .478   13.6   35.3   .066
Salmons     77  21   10.6   .447   14.4   15.6   .064
G Robinson  42  32   15.0   .518   13.6   27.4   .056
Korver      74  12   10.1   .489   10.8   20.1   .054
Coleman     34  25   11.6   .471   16.5   19.2   .047

I picked the 2003-04 season because it was Iverson's career low in TS%, eFG% and WS/48; an 11-year low in PER; career high TO% ...
Yet he still played 42.5 mpg for a team that was 33-49. He missed 35 games. The Sixers won 41% when Iverson played and 39% when he didn't.

Using this page -- http://www.basketball-reference.com/tea ... 4/lineups/ -- which lists the team's top 20 lineups, top 20 four-man arrangements, 3-man, and 2-man groupings, one can see how different lineups produced with and without Iverson.
For example, these (3) 4-man groups for all minutes, and with Iverson:

Code: Select all

.                                       all minutes together                with Iverson
Sixers '04                        Min  eFG%   Pts   ORb   DRb    TO      Min  eFG%   Pts   ORb  DRb   TO
Dalembert, Robinson, Snow, Thomas 476 -.018  -2.1    .1   1.7   1.5      232  .019   6.3   2.5  6.3    .3
Coleman, Robinson, Snow, Thomas   311  .005  -1.4   -.8   1.3   3.9      169  .032   3.2  -2.3   .6   1.3
Dalembert, McKie, Snow, Thomas    609  .000  -4.3  -2.3   1.7   1.3      117  .041  -4.2  -9.2  3.3   1.1

These 3 arrangements cover just 518 of Iverson's 2040 minutes. But as heavily-used lineups, they wouldn't include many garbage minutes.
In each case, they shoot much better with Iverson. They also turn it over a lot less.
Note that the numbers represent Team/Opponent difference. Negative turnover differential is a good thing.

It's straightforward to subtract the With Iverson minutes from the total minutes, to find the same numbers strictly without Iverson; and then to get the differences:

Code: Select all

.           without Iverson                         difference with Iverson
Sixers  Min    eFG%    Pts   ORb   DRb   TO       eFG%    Pts   ORb   DRb    TO
DRST    244   -.053  -10.1  -2.2  -2.7   2.6      .072   16.4   4.7   9.0   -2.3
CRST    142   -.027   -6.9   1.0   2.1   7.0      .059   10.1  -3.3  -1.5   -5.7
DMST    492   -.010   -4.3  - .7   1.3   1.3      .051     .1  -8.5   2.0   - .2
.               minutes weighted averages :       .063   10.7   -.9   4.0   -3.0

Losing Iverson from these lineups cost the Sixers some 11 points (per 100 possessions)
This was largely due to losing .063 in TS%, 4 DReb, and 3 TO
The loss of roughly 1 OReb is likely due to missing fewer shots.

We can get more minutes, at the expense of diluting the Iverson Effect, by checking the difference he makes to groups of 3

Code: Select all

.                                    all minutes together                with Iverson
.Sixers '04               Min   eFG%    Pts   ORb  DRb   TO      Min   eFG%   Pts   ORb   DRb   TO
Robinson, Snow, Thomas    752  -.046   -8.6    .0   .4   3.3     476  -.018  -2.1    .1   1.7   1.5
Dalembert, Snow, Thomas  1118   .001   -1.8  -1.9  3.4   1.7     452   .006   1.5  -1.0   5.3   1.5
Dalembert, Snow, Robinson 579  -.013   -4.3  -2.2  2.9   3.5     366   .011   -.6  -2.1   4.6   3.1
Coleman, Snow, Thomas     546  -.010   -2.6   1.9   .1   4.5     333   .000  -1.3   1.6    .4   2.6
Dalembert, McKie, Snow    921  -.007   -3.0  -2.8  -.8   -.4     269   .039   2.8  -9.1    .7  -2.3
Dalembert, McKie, Thomas  782   .007   - .1  - .4  2.0    .5     214   .035   8.0   -.3   4.9   -.8

.          without Iverson                             improvement with AI
Sixers  Min    eFG%    Pts    ORb   DRb   TO       eFG%    Pts   ORb   DRb    TO
RST     276   -.094  -19.8    -.2  -1.8   6.4      .076   17.7    .3   3.5   -4.9
DST     666   -.002   -4.0   -2.5   2.1   1.8      .008    5.5   1.5   3.2    -.3
DSR     213   -.054  -10.7   -2.4    .0   4.2      .065   10.1    .3   4.6   -1.1
CST     213   -.026   -4.6    2.4   -.4   7.5      .026    3.3   -.8    .8   -4.9
DMS     652   -.026   -5.4    -.2  -1.4    .4      .065    8.2  -8.9   2.1   -2.7
DMT     568   -.004   -3.2    -.4    .9   1.0      .039   11.2    .1   4.0   -1.8
.                                  averages        .047    9.6   -.8   3.1   -2.7

Here we see the most consistent savings are still in eFG% and TO. The total here is 2110 minutes, and there is substantial overlap.

Just to cover all the bases, 2 man combos with and without Iverson:

Code: Select all

.                                all minutes together                with Iverson
.Sixers04          Min   eFG%    Pts   ORb   DRb   TO      Min   eFG%   Pts   ORb   DRb   TO
Snow, Thomas      2136  -.015   -3.3    .1   2.0   2.2    1006  -.014  -1.9    .5   1.9   1.1
Dalembert, Snow   1656   .002   -1.6  -2.5   1.5    .7     751   .025   2.0  -3.7   3.4    .6
Robinson, Snow    1110  -.021   -6.7  -1.6   -.1   2.9     669  -.007  -5.7  -2.6    .8   2.5
Dalembert, Thomas 1385   .001    -.4  -1.0   2.8    .9     609   .012   3.7   -.3   4.9    .3
Robinson, Thomas   894  -.038   -7.7  -1.0    .6   3.3     582  -.009  -1.3   -.9   2.3   2.0
McKie, Snow       1520  -.016   -2.2   -.1  -1.6   -.4     529  -.012    .1  -2.1  -2.9  -3.1
McKie, Thomas     1403  -.012   -2.7   1.8   1.1   1.4     506  -.025   -.7   2.0    .1  -1.2


.          without Iverson                             improvement with AI
Sixers  Min    eFG%   Pts    ORb   DRb   TO       eFG%   Pts   ORb   DRb    TO
ST     1130   -.016   -4.5   -.3   2.1   3.2      .002   2.6    .8   -.2   -2.1
DS      905   -.017   -4.6  -1.5   -.1    .8      .042   6.6  -2.2   3.5    -.2
RS      441   -.042   -8.2   -.1  -1.5   3.5      .035   2.5  -2.5   2.3   -1.0
DT      776   -.008   -3.6  -1.5   1.2   1.4      .020   7.3   1.2   3.7   -1.1
RT      312   -.092  -19.6  -1.2  -2.6   5.7      .083  18.3    .3   4.9   -3.7
MS      991   -.018   -3.4   1.0   -.9   1.0      .006   3.5  -3.1  -2.0   -4.1
MT      897   -.005   -3.8   1.7   1.7   2.9     -.020   3.1    .3  -1.6   -4.1
.                               averages          .024   6.0   -.7   1.5   -2.1

Here at last we find a pair of players (McKie-Thomas) associated with lineups that shot worse with Iverson among them. But the point spread still improves by 3.1, TO by 4.1 per 100 possessions.

[steveshea]

Why do certain players with bad shooting percent and/or lots of turnovers get so many minutes and take so many shots?

I think sometimes it's because they get so many minutes and (at least feel as though) they are required to take so many shots that the players shoot a bad percentage and turn the ball over a lot.

I think these numbers are to a great extent a reflection of the team around Iverson.

The numbers do suggest that even when a player is not particularly efficient, he can still positively impact teammates' production. Defenses expected Iverson to shoot. He was Philadelphia's best offensive weapon. He was probably receiving a disproportionate amount of the defense's energy and attention. So, when he was on the court, the teammates had better looks. At least that's my subjective opinion. Of course, this does not imply that Philadelphia wouldn't have benefited from Iverson passing a bit more.

I'd like to compare Iverson's production that season to his production in a better offensive system, where he had another prolific scorer to draw the defense's attention. Although it wasn't a perfect union, Iverson did get to play with a 23 year-old Carmelo in 07-08. That year Iverson had a .482 2P% and a .345 3P%. I believe the 3P% was the best of his career, and the 2P% was 2nd best. It is perhaps not a coincidence that Carmelo shot .492 from the field that year and .509 on 2-pointers, which were both bests of his career.

[Mike G]

I think these numbers are to a great extent a reflection of the team around Iverson.

Indeed. Iverson's replacements that year were a rookie Willie Green, Greg Buckner, and other filler.
In a better year for Iverson and for his supporting cast, we might see different results.

At the time, though, many metrics -- WS/48 would be prominent, if it were around -- declared that he was among the worse players in the league. He was so bad individually, he should hardly play at all; and when he did, he should shoot a lot less.

With hindsight and pbp, we can see that he was indispensable. In a year with less support, his numbers suffered.

I was hoping to find similar results for Russell Westbrook (since he's a current player). But lineups without him come with very few minutes. The Iverson example is just a demonstration of how individual "efficiencies" can be disconnected from team effects.

[steveshea]

I think you make a great point Mike, and I'd like to see more numbers like the ones you present for Philadelphia.

Certain role players can produce efficiently with limited offensive responsibility. A team of all role players is not likely to be successful. The theory is that the players like Iverson (that certain metrics might malign) improve their teammates' efficiency. Take Iverson (or another star) out of the equation and the role players will not see the same looks on offense, will occasionally have to force the tough shot, and will a drop in efficiency. It would be nice to have a thorough study to support the theory.

[steveshea]

I guess I should add that Philadelphia was 33-49 in 03-04. Even though the team was better with Iverson, the team as a whole was not very good. Iverson's inefficiency was part of the problem. (Iverson's inefficiency is in part a product of the quality of his team and teammates). In 02-03, the team was 48-34. Iverson had a .414FG% and played all 82 games.

I don't want to understate efficiency either. While Iverson's efficiency is not necessarily a direct reflection of his value or ability as a player, I suspect it would be hard for any team to win with so much of the offense going through a player performing as inefficiently as the 03-04 Iverson.

[colts18]

If you look at RAPM, Iverson's impact on offense was positive every year until 2008.

Year Year Rank Name Offense per 100 Defense per 100 Off+Def per 200
2008 23 Allen Iverson 2.3 0.8 3.1
2007 69 Allen Iverson 1.7 0 1.7
2006 42 Allen Iverson 3.1 -1 2.1
2006 NPI 48 Allen Iverson 2.1 -0.4 1.6
2005 106 Allen Iverson 0.9 0 0.9
2005 NPI 203 Allen Iverson 0 -0.5 -0.5
2004 81 Allen Iverson 1.2 -0.3 0.9
2004 NPI 181 Allen Iverson 0.3 -0.6 -0.3
2003 105 Allen Iverson 1.2 -0.7 0.5
2003 NPI 163 Allen Iverson 0.4 -0.6 -0.2
2002 86 Allen Iverson 0.9 -0.3 0.5
2002 NPI 21 Allen Iverson 1 1.8 2.8
2001 NPI 38 Allen Iverson 1.9 0.3 2.1
02-11 73 Allen Iverson 1.9 -0.4 1.5

[mtamada]

Those stats are nice evidence which illustrates what most of us probably believed subjectively about Iverson: rather poor efficiency stats but his team needed him, he literally was a player who made his teammates better. Would the 76ers have been better off still with a more efficient and equally high usage player? Almost certainly but such players are very hard to get, and Iverson was the player that they had.

Over the years I grudgingly had to give Iverson his due; he made the 76ers go and they had some pretty good teams with him. A clear all-star and probable all-pro.

The harder question is: was he MVP level? And Hall of Fame level? The voters have already said yes to the MVP question and my guess is they'll say yes to the Hall of Fame question too. But personally that's where I still have not drunk the Iverson kool-aid. If you choose to build a team around Iverson, there are limits to how good that team can be. I think that 2001 Finals team was more lucky than good, and no threat to win the championship.

Carmelo Anthony has some of the same characteristics, although my guess is that if the Knicks ever did manage to create a truly good roster, there's a chance that Anthony could be convinced to reduce his usage level and play more of a team game, and be part of a championship team. Adrian Dantley almost did that. Whereas Iverson could play only one way: with the ball in his hands. No sharing of the ball, or even coming off the bench, for him. It's a lot harder to build a championship team when a key player has that sort of attitude. And that's why if I were a voter, I would not vote for him being in the Hall of Fame. But I have to admit that he was not just a gunning turnover machine. He was the franchise player for some pretty good teams.

[mbosset]

I think this is just further evidence to Dean Oliver's original idea that an increase in usage rate leads to a decrease in efficiency. I don't have any statistics of those old 76ers teams. However, with Iverson out I am confident his replacement did not have a usage rate close to Iverson's. This would result in the other four players having to increase there usage, thus decreasing their efficiency. This means that taking Iverson away from his team would be largely detrimental. However, I do agree that a team built this way is relatively limited as a good portion of your teams usage is inefficient, limiting your offensive upside.

[Mike G]

Carmelo Anthony has some of the same characteristics, although my guess is that if the Knicks ever did manage to create a truly good roster, there's a chance that Anthony could be convinced to reduce his usage level and play more of a team game,...

Excellent. A player who seems to have a middling or worse TS% and/or ORtg than teammates, while hogging the ball, and on a good team. Er, last year they were good -- 3rd best in the league on offense and 7th overall (SRS 3.73)

And, by luck, Carmelo missed 15 games. This puts his minutes in the window of on/off effectiveness regarding lineups. Here are those Knicks:

Code: Select all

.NYK'13    G  mpg    PER    TS%    TO%   Usg%  ORtg   WS/48 
Smith      80  34   17.6   .522    8.8   26.5   107   .120
Anthony    67  37   24.8   .560    9.3   35.6   112   .184
Felton     68  34   15.2   .505   14.2   22.1   105   .087
Chandler   66  33   18.9   .671   14.5   13.0   133   .207
Kidd       76  27   13.5   .511   15.0   11.6   114   .128

Novak      81  20   11.3   .587    2.6   13.0   125   .116
Prigioni   78  16   13.0   .575   27.1   11.7   116   .123
Shumpert   45  22   11.7   .516   11.1   15.6   108   .094
Copeland   56  15   16.8   .583   10.7   25.3   110   .117
Brewer     46  16   10.2   .432    6.8   13.6   100   .057
Stoudemire 29  24   22.1   .637   13.4   25.7   118   .191

The team had ORtg = 111.1 and TS% = .550

Using abbreviations of names above, player combos with and without Carmelo, showing team/opponent differential.
The summary differences are minutes weighted (using lesser of with/without minutes)

Code: Select all

lineups    - -  With Carmelo Anthony - - - -      - -  Without Carmelo Anthony - - - 

NYK'13    Min   eFG%    Pts  ORb%  TRb%   TO      Min   eFG%   Pts   ORb%  TRb%   TO
C F K Sm  269   .109   27.3  10.4  11.8  -2.3      98  -.112  -10.5   3.7  -3.9  -8.7
B C F K   210  -.054     .8    .1  -8.0  -7.9      55  -.040    7.1  25.6  22.8  -2.1
C F N Sm  102  -.068  -11.9  -6.3  -8.7    .5      94  -.053   -1.1   6.0  -1.2  -5.3
C K N Sm   53  -.001   20.4    .2  -4.3 -12.1     120  -.102  -11.9   4.2  -2.1  -3.7

. diff    300   .083   13.5  -7.0  -3.2   1.4
 
NYK'13    Min   eFG%    Pts  ORb%  TRb%   TO      Min   eFG%   Pts   ORb%  TRb%   TO
C F Sm    681   .029   11.1   3.1   2.1  -3.0     177  -.039   -1.0   2.6  -1.3  -4.9
C F K     675   .020   12.1   5.2   3.2  -4.5     163  -.052   -2.8  10.3   4.2  -5.0
C K Sm    457   .042   17.7   7.7   6.6  -4.5     215  -.077   -5.4   7.7   3.5  -4.2
F K Sm    441   .077   14.1   5.4   6.3  -1.5     221  -.043    1.5   1.2  -2.4  -6.3
N P Sm    168   .011    9.1   1.4   1.4  -5.0     428   .061   11.2  -4.2  -5.4  -5.4
C N Sm    202   .000   10.2   -.2  -2.1  -4.3     243  -.031     .7   6.0   1.7  -2.1
B C F     344  -.018    5.7  -2.8  -7.6  -6.9      79  -.018    1.4  10.6   6.9  17.7
F N Sm    197  -.001    1.6  -3.7  -5.4  -3.0     191  -.034    4.4   4.2  -1.3  -5.8
C F Sh    307  -.061   -9.7  -1.1  -4.6   -.3      70  -.029   -5.4   1.1   2.4   2.4

. diff   1486   .052    9.0  -1.9    .7   -.4                     


NYK'13    Min   eFG%    Pts  ORb%  TRb%   TO      Min   eFG%   Pts   ORb%  TRb%   TO
C F      1245   .003    6.9   2.7    .7  -3.2     282  -.019    -.1   5.9   2.9  -2.1
F Sm     1057   .026    7.6   1.4    .4  -3.3     369  -.024     .6   -.9  -3.5  -4.1
K Sm      773   .027    9.2   2.9   2.4  -3.2     579  -.015    1.7    .6  -1.6  -4.4
C Sm      969   .021   11.7   4.1   2.8  -3.8     382  -.029    1.4   7.3   3.9  -2.4
C K       960   .006    8.9   4.3   2.1  -4.4     314  -.043   -2.5  10.4   7.0  -2.4
N Sm      499   .008    7.0    .0  -1.0  -4.0     767   .020    6.3  -2.3  -4.3  -5.0
F K       866   .022    8.4   3.9   2.5  -3.4     309  -.020    4.6   3.1   -.2  -6.1
F Sh      532  -.035   -2.7   1.7  -1.7  -2.6     155  -.044   -8.5  -8.9  -9.2  -3.5
Sm St     388  -.028   -4.8   3.0   2.6    .3     211   .071   19.6  12.1  11.1  -3.7

. diff   3100   .023    4.4   -.2   1.0    .6                     

The Carmelo Effect lessens with the diminishing specificity of 3- and 2-man teammate combos. But with rising minutes validation, there's still a clear benefit to team eFG% and point differential.
Novak looks like the guy who got most non-Carmelo minutes; only in lineups with N do we see more minutes without Melo.
Novak is kind of the anti-Melo: Never turns it over, shoots a high%. But in the playoffs, Anthony went 40 mpg and Novak about 4.

The eFG% differentials are profound, as was the case with Iverson.
Perhaps, too, opponents shoot worse because they often have a defensive specialist on the floor to cover these scoring aces.

[xkonk]

For both Iverson and Carmelo there's been passing mention of their back-ups, but not much done with that information. In both cases, at least by WS/48, the back-ups are notably worse than the starters: last year Novak was a .116 to Carmelo's .184, and as bad as Iverson was on the 03-04 team (.066) his replacements were all worse (Salmons .064, Buckner .024, Green 0). How much of these 'effects' is due to Carmelo and Iverson improving their teammates and how much is due to the guys coming off the bench just being worse and lowering the line-up numbers?

[Mike G]

Good question. One reason I included eFG%, OReb% (same as DReb% differential), and TO, in addition to point differential, was to get a feel for how the pt-diff is achieved.

If a guy replacing Iverson is a lousy shooter, he's not going to try for 35% Usage. The other, middling users are going to step up their usage. And they don't do nearly as well, given that Iverson was generally a poor shooter, in terms of his own TS% or eFG%.

If Iverson takes 1/3 of the shots, at .030 worse than his teammates; yet the team improves by .040; then his teammates' shooting must be some .060 better with him than they are without him? That's phenomenal.

A replacement who can't (or won't) shoot will certainly be a drag on the offense. They might bring something like better defense than Iverson. Yet AI is also good for TO differential.

[Crow]

This season there are 13 starting PGs with usage at or above 25%. 7 of these teams have winning records, 6 losing.

Of the 6 guys with TS% above 52%, the team records are 5-1.

Of those PGs with TS% below 52%, the team records are 1-6. Only one of these guys had a positive raw +/- on the court (Teague at +0.2 per 100 possessions).

[Crow]

From above post Iverson had 2 seasons from 2001-8 that were near +3 on overall RAPM. +3 level is star level or maybe bottom of superstar level. The teams that win titles usually have someone higher or multiple players equal or higher, I would think (haven't researched to confirm yet). In getting his +3 twice and not getting there 6 times, did Iverson's style prevent other +2s or just enough else to win a title? Only high usage PGs to win a title in last 30 years were Magic, Isiah and Tony Parker (IIRC). Most teams won without a high usage PG, often with a high usage playmaking wing.

[permaximum]

Iverson's normal-season career TS% equals NBA average of his time and is greater than guards' average despite he had taken more shots than almost everyone else (Kobe, Carmelo and Wade I think takes more shots per min). So I can't talk about the Iverson effect when I see a post that starts with his so-called bad shooting. I don't remember anyone called T-Mac's shooting poor although his TS% is the same. Don't even get me talk about turnovers when it comes to high-usage players.

Iverson is definetely one of the most underrated players of all time.

[AcrossTheCourt]

One key point here is comparing Iverson to someone like Carmelo is that high usage translates to higher teammate efficiency better when you pass. Those double teams you draw don't help the open man when he never gets the ball.