Kobe Bryant and All-Time Ranking of Players

[permaximum]

Do you guys have any opinion about how to value votes from 1st place to 5th place? And what to do about pre-1980 years? Simply i'm trying to make a better and improved MVP award share that's not additive.

Edit: Well, there's an another problem. I can't seem to find all 2nd 3rd 4th 5th place votes. The best I could find is here: http://www.apbr.org/nbamvps.html and even that's missing some.

[permaximum]

MVP awards shares have it's flaws obviously. It needs improvement especially on how to value different place votes and what to do with those players that played significantly more years than others. Still, at it's current flawed state, it passes my eye test and looks better than any all-in-one metric's list.

1. Michael Jordan* 8.138
2. LeBron James 6.524
3. Kareem Abdul-Jabbar* 6.203
4. Larry Bird* 5.693
5. Magic Johnson* 5.129
6. Bill Russell* 4.827
7. Shaquille O'Neal 4.380
8. Karl Malone* 4.296
9. Tim Duncan 4.278
10. Wilt Chamberlain* 4.269
11. Kobe Bryant 4.206
12. David Robinson* 3.123
13. Kevin Durant 3.005
14. Moses Malone* 2.873
15. Kevin Garnett 2.753
16. Bob Pettit* 2.628
17. Hakeem Olajuwon* 2.611
18. Oscar Robertson* 2.479
19. Charles Barkley* 2.438
20. Steve Nash 2.429
21. Jerry West* 2.090
22. Dirk Nowitzki 1.810
23. Elgin Baylor* 1.659
24. Allen Iverson 1.567
25. Chris Paul 1.554
26. Bob McAdoo* 1.494
27. Patrick Ewing* 1.424
28. Julius Erving* 1.407
29. Dave Cowens* 1.338
30. Dwight Howard 1.249
31. Willis Reed* 1.073
32. Derrick Rose 0.981
33. Stephen Curry 0.977
34. Alonzo Mourning* 0.968
35. Jason Kidd 0.933
36. George Gervin* 0.911
37. Bob Cousy* 0.882
38. Tracy McGrady 0.855
39. Dominique Wilkins* 0.849
40. Gary Payton* 0.823
41. James Harden 0.815
42. Dwyane Wade 0.793
43. Clyde Drexler* 0.778
44. Scottie Pippen* 0.716
45. Sidney Moncrief 0.695
46. Dolph Schayes* 0.690
47. Wes Unseld* 0.639
48. Bernard King* 0.625
49. Rick Barry* 0.592
50. Chris Webber 0.588

[Mike G]

I'm highly opposed to the idea of using any kind of all-in-one metric to make a greatest of all time list ...
This brings me to the individual awards. Particularly, regular-season MVP votes.
... since we're interested in greatness this problem can be solved by a simple logic that if a guy's never considered a top 5 player at any time by any voter even if he was the 6th guy in all his career, he doesn't pass the requirement to be on the list. ..

http://bkref.com/tiny/rUDgg
One guy finished 3rd in MVP voting, the more recent player no higher than 8th.
Is top 5 in an 8- or 18-team league the same elite level as top-5 in a 30-team league?

Might several summary metrics have more validity than any one metric?

[schtevie]

I'm highly opposed to the idea of using any kind of all-in-one metric to make a greatest of all time list while they were proven not good enough over and over again in each retrodiction test accounting for roster turnover. I believe making a list by valuing opinions of people that watches, coaches, plays basketball to reach a consensus would be a strongly more reliable route. We can't objectively decide which one's opinion is closer to the truth but with a sample that's big enough, the list should be closer to the reality as long as there are some basic requirements to be included in that sample.

Hmmm. I find your relative faith in the wisdom of a crowd of self-proclaimed experts to be...um...unwise. Whatever you may think of the shortcomings of plus-minus statistics, it is worth reflecting on those of those who vote on post-season awards.

It seems that every couple of years I find myself trotting out this story, at once amusing and nutritious. In the off-season before KG's first season in Boston, much media opinion was expressed as to the effect that his (and Ray Allen, etc.) joining the team would have. And how wise was that electorate? Go back and look. Not very. I think the consensus view was that though obviously a one-sided trade, it was but a classic "trade youth for aging veterans" move, with no collective sense that one of the best NBA teams of all-time had been constructed. Representing the 5th Estate's competence, we had "the Dean of NBA Sportswriters", our local Bob Ryan, saying (and I wish I had the article to link to) that the Celtics had basically become a 0.500 team.

Now, was Bob Ryan an idiot? Well, this I don't know, but let's presume that the presumptive king of the clan was not. If so, how could he (and so many others) have made such egregious analytic errors?

Well, the sympathetic answer is that there was then (and remains) a systematic bias against recognizing defensive contributions, owing largely (but not solely) to the difficulty in perceiving same. One part of this relates to the lack of relative box score statistics to guide intuition, and the other is the difficulty in perceiving plus-minus changes as they occur on the court (what, of course, afflicts the offensive end as well). And both these points are relevant to the discussion of the apparent shortcomings of DBPM to DxRAPM.

But perhaps the post-season award electorate has cleaned up its act and is now a font of collective wisdom. Pull another one.

What comes most recently to mind is a recent ESPN Tim Legler "end-of-career-compare-the-relative-merits-of" interview, discussing Kobe Bryant vs. Tim Duncan. Classic, ridiculous false-equivalence media, where the resolution of the supposed near-tie at least came up with the answer of the (much) better player, but what was evident is that there was no rational basis for assessing relative worth. Indeed, post-season awards won were the unanchored, self-referential basis for the apparent confusion.

Wisdom of the crowd, indeed.

[permaximum]

http://bkref.com/tiny/rUDgg
Is top 5 in an 8- or 18-team league the same elite level as top-5 in a 30-team league?

No. Basketball and it's rules change constantly too. Even ball physics change constantly too. Does any metric account for those?

Might several summary metrics have more validity than any one metric?

No. Blends don't improve the outcome meaningfully.

Edit: I understood it wrong. Their validity depends on the person who evaluates them. So, it would be subjective.

I know that this approach is not ideal but all-in-one metrics haven't come to the point that they isolate AND capture a single player's effect on the outcome of games reliably. Let alone in a game in 1977 to make an all-time list. At the moment, I would trade metrics' unbiased approach for basketball people's bias. Only I would try to reduce the bias by taking opinions from as many basketball people as I can. That's why I thought about MVP voting.

Do you agree that, MVP award shares list looks better than for example BPM list or WS list?

[permaximum]

I'm highly opposed to the idea of using any kind of all-in-one metric to make a greatest of all time list while they were proven not good enough over and over again in each retrodiction test accounting for roster turnover. I believe making a list by valuing opinions of people that watches, coaches, plays basketball to reach a consensus would be a strongly more reliable route. We can't objectively decide which one's opinion is closer to the truth but with a sample that's big enough, the list should be closer to the reality as long as there are some basic requirements to be included in that sample.

Hmmm. I find your relative faith in the wisdom of a crowd of self-proclaimed experts to be...um...unwise. Whatever you may think of the shortcomings of plus-minus statistics, it is worth reflecting on those of those who vote on post-season awards.

It seems that every couple of years I find myself trotting out this story, at once amusing and nutritious. In the off-season before KG's first season in Boston, much media opinion was expressed as to the effect that his (and Ray Allen, etc.) joining the team would have. And how wise was that electorate? Go back and look. Not very. I think the consensus view was that though obviously a one-sided trade, it was but a classic "trade youth for aging veterans" move, with no collective sense that one of the best NBA teams of all-time had been constructed. Representing the 5th Estate's competence, we had "the Dean of NBA Sportswriters", our local Bob Ryan, saying (and I wish I had the article to link to) that the Celtics had basically become a 0.500 team.

Now, was Bob Ryan an idiot? Well, this I don't know, but let's presume that the presumptive king of the clan was not. If so, how could he (and so many others) have made such egregious analytic errors?

Well, the sympathetic answer is that there was then (and remains) a systematic bias against recognizing defensive contributions, owing largely (but not solely) to the difficulty in perceiving same. One part of this relates to the lack of relative box score statistics to guide intuition, and the other is the difficulty in perceiving plus-minus changes as they occur on the court (what, of course, afflicts the offensive end as well). And both these points are relevant to the discussion of the apparent shortcomings of DBPM to DxRAPM.

But perhaps the post-season award electorate has cleaned up its act and is now a font of collective wisdom. Pull another one.

What comes most recently to mind is a recent ESPN Tim Legler "end-of-career-compare-the-relative-merits-of" interview, discussing Kobe Bryant vs. Tim Duncan. Classic, ridiculous false-equivalence media, where the resolution of the supposed near-tie at least came up with the answer of the (much) better player, but what was evident is that there was no rational basis for assessing relative worth. Indeed, post-season awards won were the unanchored, self-referential basis for the apparent confusion.

Wisdom of the crowd, indeed.

You point out otliers. With 130 voters I believe we can find a sweet spot.

BTW, What makes your metric or BPM or RPM or PER etc. better than a random person's opinion who only watches games? Metrics can be objective but do they capture enough? I remember NBA efficiency rating and per/48 stats were the things back then. Then it was PER. With the advancement of capturing and problem solving methods we get newer and supposedly better metrics and there will be new and "better" metrics wich will improve the accuracy of predicting the outcome of the next game by 0.3%. What then? Will we forget about BPM, RPM and WS to use those to evaluate player performances?

I have to see the results. Show me an out-of-sample retrodiction test which proves a metric's signficantly lower RMSE compared to MPG and non-empirical PER at 100% roster turnover rate (all minutes from new players). Do I have to take your word?

[schtevie]

You point out otliers. With 130 voters I believe we can find a sweet spot.

Let us not forget that when it comes to the All-Time Ranking of Players, the entire conversation is about outliers. Unless one is interested in the accuracy of rank #486 vs. #503 (in which case I recommend spending more time with friends and family and/or therapy) we are talking about the extreme tail of the NBA performance distribution, aggregated (somehow) over time. And if outlier defensive performance is systematically mismeasured (for paint-protectors, in particular) then this is a big problem, as defense is half the game.

BTW, What makes your metric or BPM or RPM or PER etc. better than a random person's opinion who only watches games? Metrics can be objective but do they capture enough? I remember NBA efficiency rating and per/48 stats were the things back then. Then it was PER. With the advancement of capturing and problem solving methods we get newer and supposedly better metrics and there will be new and "better" metrics wich will improve the accuracy of predicting the outcome of the next game by 0.3%. What then? Will we forget about BPM, RPM and WS to use those to evaluate player performances?

First, I don't have a particular metric, but a general approach of which I have always approved (plus-minus) and which over its particular years, provides a very useful guide about relative player value (which is different from the intertemporal question at hand, about which I have significant misgivings, as already stated). So, if there are refinements to this approach, I am happy to accept the improved results (of which, for example, I presume RPM is an improvement on Jeremias' earlier work, and I look forward to these estimates being extended back to 2001, if possible.)

I have to see the results. Show me an out-of-sample retrodiction test which proves a metric's signficantly lower RMSE compared to MPG and non-empirical PER at 100% roster turnover rate (all minutes from new players). Do I have to take your word?

I am confident that there is little I could say that would dislodge you from your views (these are the interwebs, no?) so no, I am not asking you take my word. I would, however, ask you to consider if the test you propose is the relevant one for the issue at hand. And also, to reflect on your belief that media - who are ever so clearly basing their votes on a very biased model of NBA value - can really be expected to produce superior results to xRAPM data (what only exist since 2001) or BPM (even given the aforementioned limitations).

[Mike G]

..
Do you agree that, MVP award shares list looks better than for example BPM list or WS list?

Don't know. But as I suggested, it's straightforward enough to take a blend of 2 or more metrics. Each time you add one to the mix -- if it's better than a random number -- you tend to get a better list.

But my own metrics predate all those you mentioned (except mpg), and I find some careers in the MVP-shares Top 50 who do not make my top 80.
Others (from my top 40) would seem to have had greater careers.

Code: Select all

. Out       In
Harden    Stockton
Curry     Pierce
Rose      Havlicek
Moncrief  Hayes
King      Gilmore
Reed      Parish
Unseld    Lanier

Of course, regular season MVP doesn't take into account the postseasons these players may have -- other than subjectively influencing some voters toward players who deserve an MVP for their career efforts.

If some standardization can be achieved in the MVP Share from era to era, it may indeed be useful as another ingredient in a career metric. But still, great players whose careers happened to coincide with Russell's or Jordan's or LeBron's -- or who played center in an age of centers, say -- will be undervalued.

Looking at others that may be too low or too high, by MVP shares:

Code: Select all

. High         Low
20  Nash      Wilt     10
2   LeBron    Duncan    9
4   Bird      Pippen   44
26  McAdoo    Olajuwon 17
6   Russell   Erving   28
25  Paul      Wade     42
34  Mourning  Drexler  43
14  Moses     Garnett  15
47  Unseld    Barry    49
5   Magic     Schayes  46
24  Iverson   Shaq      7
36  Gervin
29  Cowens
39  Wilkins

It is nice to see Bird and Magic neck-and-neck. Until Magic caught up with Bird's (3) MVP's, with 2 that Jordan probably should have had, it was a heavy mismatch in this "stat".

[Mike G]

Here's a stab at making MVP shares equivalent across eras.
In olden times, MVP votes were gotten 1 from each voter -- no 2nd place or 3rd, etc votes. In this scenario, the b-r.com "mvp share" is actually the fraction of 1.00 total mvp share.

In later decades, the MVP share is the fraction of highest possible. A 2nd place total might be as high as .813 (Shaq'05) -- he got enough 1st and 2nd place votes to total .813 of what he'd get with all the 1sts.

So an alternative way of doing it is:
MVP = 1.00
2nd = 1/2 = .50
3rd = 1/3 = .33
etc

This inverse method eliminates most of the disparity in total shares awarded each season; but it doesn't distinguish between Shaq getting 100% of the 1st place votes in 2000, and Kareem winning with 28% in 1976.

Both measures are useful, in a season or over a career. And the average of the 2 might be best.
I looked up the top 11 (from each player page), to include Kobe -- since this is/was the Kobe thread.

Code: Select all

mvp        %        mvp    inv        mvp    avg
Jordan   8.14      Kareem  8.77      Jordan  7.82
LeBron   6.53      Jordan  7.50      Kareem  7.48
Kareem   6.20      Russell 7.31      LeBron  6.46
Bird     5.69      LeBron  6.39      Russell 6.07
Magic    5.13      Wilt    6.18      Bird    5.78
Russell  4.83      Bird    5.87      Magic   5.38
Shaq     4.38      Magic   5.63      Wilt    5.22
Malone   4.30      Malone  4.88      Malone  4.59
Wilt     4.27      Shaq    4.15      Shaq    4.26
Kobe     4.21      Kobe    4.04      Kobe    4.13
.       53.67             60.72             57.19

There's still the issue of the relative value of an MVP award, or a 4th best vote, in 1960 vs 2010. That would be Part 2.

[Mike G]

Most straightforward, perhaps, is to just take each season's 1st place MVP votes, divide by that year's total 1st place votes (# of voters), and sum them for careers. Starting in 1956:

Code: Select all

%1st    mvp       %1st   mvp
5.27   Jordan     .96   Kobe
4.10   Kareem     .93   Rose
3.79   LeBron     .93   Olajuwon
3.43   Bird       .83   Cowens
3.30   Russell    .82   Iverson
3.25   Wilt       .77   Curry
2.17   Moses      .76   Nowitzki
2.06   Magic      .65   Erving
1.81   Shaq       .62   Gervin
1.76   Pettit     .62   Cousy
1.46   Garnett    .61   Reed
1.46   Oscar      .54   Baylor
1.36   Duncan     .52   Walton
1.33   K Malone   .52   Unseld
1.31   Nash       .42   Schayes
1.18   Durant     .39   Arizin
1.15   West       .38   Hayes
1.06   Barkley    .37   Kidd
1.04   Robinson   .31   Mourning
1.01   McAdoo     .29   Paul

West is by far the highest accumulator who never won it, almost double Gervin and Cousy. Unseld is the lowest ranker who did win one -- he got 99% of his total in his rookie year.
This list could be combined with all-NBA selections, to arrive at a more comprehensive list. Some years, Russell was MVP, and Wilt was 1st team C. MVP is more of a "team award".
Then factor in size and strength of league; ABA; All-Star appearances; and playoffs!

[Mike G]

Rainy day fun. Here's a "real mvp" list based on the avg of 3 "wins" appropriations.
WS = WS/48 * MP/48
perW = (PER-5.75)/92.5 *MP/48
bpmW = (BPM+5.14)/51.4 *MP/48

Most years, the MVP winner has been the xyz leader. When the MVP is further down, his name is capitalized.

Code: Select all

xyzW    2015           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.6   James Harden   HOU   2981   26.7  .265   8.4   14.1   16.5   16.4
14.9   STEPHEN CURRY  GSW   2613   28.0  .288   9.9   13.1   15.7   15.9
14.6   Chris Paul     LAC   2857   26.0  .270   7.5   13.0   16.1   14.6
                           
xyz     2014           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.9   Kevin Durant   OKC   3122   29.8  .295   8.8   16.9   19.2   17.6
16.0   LeBron James   MIA   2902   29.3  .264   8.9   15.4   16.0   16.5
14.3   Kevin Love     MIN   2797   26.9  .245   8.4   13.3   14.3   15.3
                           
xyz     2013           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.5   LeBron James   MIA   2877   31.6  .322  11.6   16.8   19.3   19.5
17.0   Kevin Durant   OKC   3119   28.3  .291   7.7   15.8   18.9   16.2
12.4   James Harden   HOU   2985   23.0  .206   5.5   11.6   12.8   12.9
                           
xyz     2012           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
14.2   LeBron James   MIA   2326   30.7  .298  11.0   13.1   14.4   15.2
11.5   Chris Paul     LAC   2181   27.0  .278   7.9   10.4   12.6   11.5
11.5   Kevin Durant   OKC   2546   26.2  .230   5.2   11.7   12.2   10.7
                           
xyz     2011           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.8   LeBron James   MIA   3063   27.3  .244   8.6   14.9   15.6   17.1
13.2   Dwight Howard  ORL   2935   26.1  .235   4.8   13.5   14.4   11.8
13.2   Pau Gasol      LAL   3037   23.3  .232   5.3   12.0   14.7   12.9
13.0   Chris Paul     NOH   2880   23.7  .232   6.4   11.6   13.9   13.5
12.9   DERRICK ROSE   CHI   3026   23.5  .208   5.9   12.1   13.1   13.5
                           
xyz     2010           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.9   LeBron James   CLE   2966   31.1  .299  12.5   16.9   18.5   21.2
14.8   Kevin Durant   OKC   3239   26.2  .238   5.1   14.9   16.1   13.4
14.5   Dwyane Wade    MIA   2792   28.0  .224   9.4   14.0   13.0   16.5
                           
xyz     2009           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
20.2   LeBron James   CLE   3054   31.7  .318  13.0   17.8   20.2   22.5
18.2   Chris Paul     NOH   3002   30.0  .292  11.2   16.4   18.3   19.9
17.1   Dwyane Wade    MIA   3048   30.4  .232  10.7   16.9   14.7   19.6
                           
xyz     2008           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.1   LeBron James   CLE   3027   29.1  .242  11.2   15.9   15.3   20.0
16.8   Chris Paul     NOH   3006   28.3  .284   9.2   15.3   17.8   17.5
13.6   KOBE BRYANT    LAL   3192   24.2  .208   5.4   13.3   13.8   13.6
                           
xyz     2007           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
14.6   Dirk Nowitzki  DAL   2820   27.6  .278   6.7   13.9   16.3   13.5
14.5   LeBron James   CLE   3190   24.5  .206   7.4   13.5   13.7   16.2
13.3   Kobe Bryant    LAL   3140   26.1  .199   4.7   14.4   13.0   12.5
                           
xyz     2006           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.6   LeBron James   CLE   3361   28.1  .232   9.3   16.9   16.2   19.7
15.6   Dirk Nowitzki  DAL   3089   28.1  .275   5.7   15.5   17.7   13.6
15.4   Kobe Bryant    LAL   3277   28.0  .224   5.8   16.4   15.3   14.5
14.9   Kevin Garnett  MIN   2957   26.8  .242   7.9   14.0   14.9   15.6
14.5   Elton Brand    LAC   3099   26.5  .229   6.3   14.5   14.8   14.4
14.5   Dwyane Wade    MIA   2892   27.6  .239   7.5   14.2   14.4   14.8
14.0   Gilbert Arenas WAS   3384   23.8  .193   5.5   13.8   13.6   14.6
13.8   Shawn Marion   PHO   3263   23.6  .214   5.3   13.1   14.5   13.8
12.9 Chauncey Billups DET   2925   23.4  .254   4.7   11.6   15.5   11.7
12.3   Pau Gasol      MEM   3135   22.7  .184   5.1   12.0   12.0   13.0
11.9   Paul Pierce    BOS   3084   23.6  .178   4.3   12.4   11.4   11.8
11.1   STEVE NASH     PHO   2796   23.3  .212   3.7   11.1   12.3   10.0
                           
xyz     2005           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.9   Kevin Garnett  MIN   3121   28.2  .248   9.7   15.8   16.1   18.8
16.0   LeBron James   CLE   3388   25.7  .203   8.3   15.2   14.3   18.5
14.2   Dirk Nowitzki  DAL   3020   26.1  .248   5.7   13.8   15.6   13.3
...  et cetera -- no sign of NASH
                           
xyz     2004           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.4   Kevin Garnett  MIN   3231   29.4  .272   9.9   17.2   18.3   19.7
12.7 Andrei Kirilenko UTA   2895   22.6  .192   8.2   11.0   11.6   15.7
12.7   Tim Duncan     SAS   2527   27.1  .249   7.3   12.2   13.1   12.7
                           
xyz     2003           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.7   Tracy McGrady  ORL   2954   30.3  .262   9.7   16.3   16.1   17.8
16.5   Kevin Garnett  MIN   3321   26.4  .225   8.7   15.4   15.6   18.6
15.9   TIM DUNCAN     SAS   3181   26.9  .248   7.4   15.2   16.4   16.2
                           
xyz     2002           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.0   Tim Duncan     SAS   3329   27.0  .257   7.6   15.9   17.8   17.2
13.7   Kevin Garnett  MIN   3175   23.8  .194   6.7   12.9   12.8   15.2
13.2   Elton Brand    LAC   3020   23.6  .216   6.2   12.1   13.6   13.9
                           
xyz     2001           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.3 Shaquille O'Neal LAL   2924   30.2  .245   7.3   16.1   14.9   14.7
13.5   Vince Carter   TOR   2979   25.0  .208   7.0   12.9   12.9   14.7
13.2   Tim Duncan     SAS   3174   23.8  .200   5.4   12.9   13.2   13.6
13.2   Ray Allen      MIL   3129   22.9  .211   5.6   12.1   13.8   13.6
13.0   Dirk Nowitzki  DAL   3125   22.8  .224   4.7   12.0   14.6   12.5
13.0   Tracy McGrady  ORL   3087   24.9  .189   5.6   13.3   12.2   13.4
13.0   Kevin Garnett  MIN   3202   23.9  .176   5.7   13.1   11.7   14.1
12.9   Steve Francis  HOU   3194   21.6  .184   6.6   11.4   12.2   15.2
12.8   Karl Malone    UTA   2895   24.7  .217   5.9   12.4   13.1   13.0
12.0   ALLEN IVERSON  PHI   2979   24.0  .190   4.8   12.2   11.8   12.0
                           
xyz     2000           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.5 Shaquille O'Neal LAL   3163   30.6  .283   9.7   17.7   18.6   19.0
14.8   Gary Payton    SEA   3425   23.6  .195   6.8   13.8   13.9   16.6
14.6   Karl Malone    UTA   2947   27.1  .249   6.9   14.2   15.3   14.4
                           
xyz    1999            Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
8.9   Karl Malone     Uta   1832   25.6  .252   6.8    8.2    9.6    8.9
8.8  Shaquille O'Neal LAL   1705   30.6  .255   6.3    9.5    9.1    7.9
8.1   Tim Duncan      SAS   1963   23.2  .213   4.9    7.7    8.7    8.0
                           
xyz     1998           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.5   Karl Malone    UTA   3030   27.9  .259   7.0   15.1   16.3   14.9
14.1   MICHAEL JORDAN CHI   3181   25.2  .238   4.6   13.9   15.8   12.6
13.0   David Robinson SAS   2457   27.8  .269   7.8   12.2   13.8   12.9
                           
xyz     1997           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.3   Karl Malone    UTA   2998   28.9  .268   8.5   15.6   16.7   16.6
16.2   Michael Jordan CHI   3106   27.8  .283   6.7   15.4   18.3   14.9
15.1   Grant Hill     DET   3147   25.5  .223   8.0   14.0   14.6   16.8
                           
xyz     1996           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.0   Michael Jordan CHI   3090   29.4  .317   8.6   16.5   20.4   17.2
17.3   David Robinson SAS   3019   29.4  .290   9.2   16.1   18.2   17.5
15.0   Karl Malone    UTA   3113   26.0  .233   7.3   14.2   15.1   15.7
                           
xyz     1995           Tm    MP    PER   WS/48   PM   perW    WS    bpmW
16.8   David Robinson SAS   3074   29.1  .273   8.4   16.2   17.5   16.9
13.8 Shaquille O'Neal ORL   2923   28.6  .230   5.4   15.0   14.0   12.5
13.8   Karl Malone    UTA   3126   25.1  .212   5.8   13.6   13.8   13.9
                           
xyz     1994           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
19.8   David Robinson SAS   3241   30.7  .296  10.9   18.2   20.0   21.1
16.3 Shaquille O'Neal ORL   3224   28.5  .252   6.6   16.5   16.9   15.3
15.0  HAKEEM OLAJUWON HOU   3277   25.3  .210   7.0   14.4   14.3   16.1
                           
xyz     1993           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.3   Michael Jordan CHI   3067   29.7  .270   9.5   16.5   17.3   18.2
16.4  Hakeem Olajuwon HOU   3242   27.3  .234   8.4   15.7   15.8   17.8
15.2   Karl Malone    UTA   3099   26.2  .238   7.6   14.3   15.4   16.0
14.3  CHARLES BARKLEY PHO   2859   25.9  .242   8.3   13.0   14.4   15.6
                           
xyz     1992           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.8   Michael Jordan CHI   3102   27.7  .274   8.6   15.3   17.7   17.3
14.1   David Robinson SAS   2564   27.5  .260  10.0   12.6   13.9   15.7
13.8   Karl Malone    UTA   3054   25.4  .237   5.3   13.5   15.1   12.9
                           
xyz     1991           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
19.2   Michael Jordan CHI   3034   31.6  .321  10.8   17.7   20.3   19.6
16.4   David Robinson SAS   3095   27.4  .264   8.4   15.1   17.0   17.0
14.8   Karl Malone    UTA   3302   24.8  .225   5.8   14.2   15.5   14.6
14.7   Magic Johnson  LAL   2933   25.1  .251   8.3   12.8   15.3   16.0
                           
xyz     1990           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
19.2   Michael Jordan CHI   3197   31.2  .285  10.6   18.3   19.0   20.4
16.9  Charles Barkley PHI   3085   27.1  .269   9.8   14.8   17.3   18.7
15.9   MAGIC JOHNSON  LAL   2937   26.6  .270   9.5   13.8   16.5   17.4
                           
xyz     1989           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
20.6   Michael Jordan CHI   3255   31.1  .292  12.6   18.6   19.8   23.4
16.5  Charles Barkley PHI   3088   27.0  .250   9.8   14.8   16.1   18.7
15.6   MAGIC JOHNSON  LAL   2886   26.9  .267   9.3   13.7   16.1   16.9
                           
xyz     1988           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
21.3   Michael Jordan CHI   3311   31.7  .308  12.2   19.4   21.2   23.3
16.8  Charles Barkley PHI   3170   27.6  .253   9.0   15.6   16.7   18.2
15.5   Larry Bird     BOS   2965   27.8  .243   8.8   14.7   15.0   16.8
                           
xyz     1987           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.6   Michael Jordan CHI   3281   29.8  .247   8.6   17.8   16.9   18.3
15.6   Larry Bird     BOS   3005   26.4  .243   9.2   14.0   15.2   17.5
15.1   MAGIC JOHNSON  LAL   2904   27.0  .263   8.1   13.9   15.9   15.6
                           
xyz     1986           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.9   Larry Bird     BOS   3113   25.6  .244   9.1   13.9   15.8   18.0
12.5  Charles Barkley PHI   2952   22.4  .176   7.9   11.1   10.8   15.6
11.9   Adrian Dantley UTA   2744   24.6  .223   4.9   11.6   12.7   11.2
                           
xyz     1985           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.0   Larry Bird     BOS   3161   26.5  .238   8.5   14.8   15.7   17.5
15.0   Michael Jordan CHI   3144   25.8  .213   8.2   14.2   14.0   17.0
11.9   Magic Johnson  LAL   2781   23.2  .220   5.4   10.9   12.7   11.9
11.8   Isiah Thomas   DET   3089   22.2  .173   5.2   11.4   11.1   12.9
                           
xyz     1984           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
14.0   Larry Bird     BOS   3028   24.2  .215   7.8   12.6   13.6   15.9
12.9   Adrian Dantley UTA   2984   24.6  .235   4.4   12.7   14.6   11.5
11.6  Sidney Moncrief MIL   3075   19.8  .198   4.8    9.7   12.7   12.4
                           
xyz     1983           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
13.9   Larry Bird     BOS   2982   24.1  .225   7.6   12.3   14.0   15.4
12.9   Magic Johnson  LAL   2907   23.0  .207   7.4   11.3   12.5   14.8
12.6   MOSES MALONE   PHI   2922   25.1  .248   3.4   12.7   15.1   10.1
                           
xyz     1982           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
14.8   Moses Malone   HOU   3398   26.8  .218   4.3   16.1   15.4   13.0
13.6   Magic Johnson  LAL   2991   22.9  .207   8.3   11.6   12.9   16.3
13.5   Julius Erving  PHI   2789   25.9  .229   7.8   12.7   13.3   14.6
                           
xyz     1981           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
13.9   Julius Erving  PHI   2874   25.1  .231   8.0   12.5   13.8   15.3
13.8   Adrian Dantley UTA   3417   24.3  .191   4.6   14.3   13.6   13.5
13.4  Kareem A-Jabbar LAL   2976   25.5  .230   5.3   13.2   14.3   12.6
                           
xyz     1980           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
14.6  Kareem A-Jabbar LAL   3143   25.3  .227   6.7   13.8   14.9   15.1
13.0   Julius Erving  PHI   2812   25.4  .213   7.2   12.4   12.5   14.1
11.7   Moses Malone   HOU   3140   24.1  .183   2.8   13.0   12.0   10.1
                           
xyz     1979           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
15.0  Kareem A-Jabbar LAL   3157   25.5  .219   7.8   14.0   14.4   16.6
13.5   MOSES MALONE   HOU   3390   23.7  .200   4.0   13.7   14.1   12.6
12.6   Artis Gilmore  CHI   3265   22.3  .186   4.7   12.2   12.7   13.0
                           
xyz     1978           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
12.4  Kareem A-Jabbar LAL   2265   29.2  .257   9.0   12.0   12.1   13.0
12.0   Artis Gilmore  CHI   3067   23.5  .180   4.6   12.3   11.5   12.1
11.8   David Thompson DEN   3025   23.2  .202   3.6   11.9   12.7   10.7
...
8.7   BILL WALTON     Por   1929   24.8  .209   7.0    8.3    8.4    9.5
                           
xyz     1977           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.4  Kareem A-Jabbar LAL   3016   27.8  .283  10.8   15.0   17.8   19.5
11.5   Elvin Hayes    WSB   3364   19.8  .173   3.4   10.6   12.1   11.6
11.4   Artis Gilmore  CHI   2877   21.6  .203   5.0   10.3   12.2   11.8
                           
xyz     1976           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
18.1  Kareem A-Jabbar LAL   3379   27.2  .242  10.2   16.3   17.0   21.0
13.1   Bob McAdoo     BUF   3328   23.3  .178   5.0   13.2   12.3   13.7
10.9   Dave Cowens    BOS   3101   18.9  .165   5.0    9.2   10.7   12.7
                           
xyz     1975           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
16.0   Bob McAdoo     BUF   3539   25.8  .242   4.7   16.0   17.8   14.1
13.5   Bob Lanier     DET   2987   24.4  .199   7.6   12.5   12.4   15.4
13.4  Kareem A-Jabbar MIL   2747   26.4  .225   8.0   12.8   12.9   14.6
                           
xyz     1974           Tm    MP    PER   WS/48  BPM   perW    WS    bpmW
17.7  Kareem A-Jabbar MIL   3548   24.4  .250   8.5   14.9   18.5   19.6
14.6   Bob Lanier     DET   3047   23.9  .227   8.5   12.5   14.4   16.8
13.9   Bob McAdoo     BUF   3185   24.7  .231   4.8   13.6   15.3   12.8

[Mike G]

From 1974 to '79, Kareem's BPM run 10-30% higher than his equivalent PER. His BPM are 6 to 23% above what his Win Shares would have you suppose.
Lanier is also higher in BPM than the other 2, by 5 to 30%. Cowens is higher in BPM by 15 to 40%.
Hayes, Gervin, and Rudy T are notably lower in bpm-wins than in WS.

Is this valuable information, if you are speculating on BPM before 1974, and basing it on Win Shares, PER, or both?
You run into similar extrapolations when you guess at turnovers, pre-1974. Some have more and some less, relative to other stats, after 1974.

Below is a streamlined chart from the table I posted earlier.
The red line represents an interpretation of zero bias, a factor of 1.00 from year to year of league competitiveness, for players aged 24 to 28 yrs.
Competitive Index is the avg of yr2/yr1 minutes and rebound rates, for the 24-28 yr olds in the league.
This study concludes in 2007. Up to that point, you can see similarities and differences with DSMok1's BPM-based chart.
I used the midpoint of NBA history, 1977, as the standard 1.00 index strength.
It may well be that the bias (or lack thereof) is not the same before/after 1977.
The question is: Do 24-27 yr olds tend to be just as good when they are 25-28 years old? Slightly better, or slightly worse? That may have changed somewhere in the timeline.

The most correct curve might be one tail to the left of 1977, and a different color moving to the right.

[schtevie]

I ran a full era adjustment for everything, which helps the 1960s and 1980s and hurts the 1970s and particularly the ABA. Here's the adjustment chart:



It's very interesting. This looks at all players near their peak that played in both seasons, and sees whether they perform better in the first or second season, adjusted for aging. Statistically, there is a fair amount of error possible, not in the overall shape of the adjustment curve, but in the overall slope--in other words, it is quite possible the left end of the line should be moved up or down a point or two do to some slight systematic bias. So take it with a grain of salt.

Hopefully a productive conversation can be had about this very interesting graph (perhaps it even merits being split away from this thread because the issue of all-time ranking of players is but derivative).

My initial reaction to seeing this story told (including the title!) is that there seems to be big problems with it (or at least questions) and further consideration reinforces that impression.

What is or ought to be our common, a priori view about "era adjustments"? Well, first, the term itself should be considered something of an imprecision. What should determine changes in the average quality of league play across time? In expectation, there is but one general factor: talent dilution/concentration. And here the sole, general determinant is the tension between league expansion and demography, writ large. Then overlaying this is a stochastic element, where incoming and outgoing cohorts in any given year have certain random, quality variation, e.g. not every year does a LeBron enter the league.

So, in essence, there ought to be no real "eras" to speak of; instead, there is the overlay of the historical schedule of league expansion/contraction (within and without the NBA) and relevant demographic trends (growth rates in player population subsets) what in turn are influenced by economic and social factors.

As to the former, there is the relative, economic attractiveness of supplying labor as a player in the NBA vs. all other avocations (athletic and non). And here this has risen markedly over time. Then as to the latter, the most important factors have been the unwinding the malign effects of past discrimination against African-Americans then later in time against "Euros".

Given this general analytic structure (which we all agree upon...right?) what are the implications about what we would expect an "era adjustment" plot to look like, generally?

Well, replying to what Daniel says is a shortcoming of his presentation (taking the present day as a baseline) there is the issue of the "vertical adjustment" for the initial year (1952 in his plot?) And if one admits any reasonable lack of opportunity for African-Americans beginning in what was the end of the Jim Crow era, that one-off adjustment will be way down (a direction reinforced by all other relevant factors, by the way). And then there would be catch-up over time according to some rule, until the present racial/non-US native composition was achieved. (And a similar approach would be taken for the influx of non-native US players.) But I will leave the discussion of such particulars for now, but we can discuss details if there is any desire to.

Then overlaying such a relatively smooth demographic plot we would expect to see discontinuities for realized league expansions and contractions (what we can infer from Daniels' chart as well). As for the expected size of the jumps, these can probably be reasonably estimated for their average expected effect based upon the contemporaneous distribution of plus-minus data. (Again, a discussion of particulars can be undertaken should there be interest.)

And that should be about it, what suggests a very different time path of "era adjustment" than revealed by Daniels' plot.

Let me discuss one "era" in question that particularly raises my suspicions: the twenty year, near monotonic and linear decline from the peak of 1983-84 to trough in 2003-04. Why this raises special concern is that I suspect that all demographic/expansion issues probably not just cancel themselves out, but imply what should have been a net talent concentration.

(By this I mean that if you total the minutes or Win Shares taken by non-native US players in 2004 and account for the general US population growth over time, league expansion was running behind the relevant population growth - and I assume that by 1984 the racial representation issues were essentially off the table...but perhaps not. And, of course, in terms of economic opportunity for athletes, the NBA was increasingly the best place to be.)

But instead we learn from the exhibited plot that this was a period of dramatic, really dramatic, and relentless deterioration in what I am interpreting is the quality of the average NBA player. And the story for this is what exactly? It cannot be randomness can it, in the sense that it just so happened that for about twenty consecutive years inferior classes of athletes just happened to enter the NBA?

My suspicion, and I don't understand enough of the method used to really comment, is that what the plot is picking up is a period of exogenous improvements in defense that diffuse league-wide over time - at least that is a story consistent with my understanding of the historiography. Specifically, as the story goes, the end of the 80s coincides with Pistons' efforts to slow down Michael Jordan, followed by Pat Riley's influential interpretation and generalization of the same philosophy. We all remember the introduction of the "no layups" rule, right? And then there was the "mini-era" of the shortening of the 3-point line that ironically reinforced these same defensive tendencies.

And as goes this two decade "era" there are similar problems throughout. Daniel writes that errors aren't possible in the overall shape of the adjustment curve, and though I am loathe to question his judgment, I would like at least for there to be some story offered for these "unrandom" results. Might breaking down the plot in terms of offense and defense possibly clarify matters?

[Mike G]

The table I offered up correlates very closely to DSMok1's chart -- up to about 1993, when his index continues southward, while my own data suggest a reversal and recovery.

it is quite possible the left end of the line should be moved up or down a point or two do to some slight systematic bias.

A point or (especially) two is quite a bit and changes the shape enormously.
Working that bias into the 60-year history is the fun part.

[permaximum]

Under APBRmetrics it's written: "The discussion of the analysis of basketball through objective evidence, especially basketball statistics."

I have to say, I'm losing hope for the analytics. Still, I have never seen anyone share the evidence (I'm asking it for the 100th time) that these metrics improve significantly upon such a simple thing as MPG which has big flaws to use as a predictor. If you're interested in this stuff, just run an out-of-sample retrodiction test at full roster turnover rate and share the results here. Just show the objective evidence.

For example, one comes with RPM but never gives the complete steps about how he calculates it. While everyone's RAPM numbers for the same season differs (even more than any reasonable penalty for the regression allows), we have to take a much-complicated RPM (that's what we're "told") as an objective evidence. Or WS, BPM, PER. Best of those metrics improve our chance to predict the winner of an NBA game by a few percents upon HCA.

I'm especially bringing Iverson argument because it seems the greatest outlier. His coach Lary Brown says he is the greatest athlete, competitor, toughest, player, best player at size etc. He even said without him 2001 76ers wouldn't even make the playoffs let alone win a game against supposedly unbeateble Lakers of 2001 in the Finals. (which I still think is the best team of all time). Shaq, Wade, Lebron, Durant think he's the best player ever pound for pound, one of the 5 greatest ever etc. Back to the topic Kobe recently said, Jordan and Iverson were the greatest competitiors. But for example in 14-year RAPM he's 180th or something. In 2001's RAPM, Mutombo has a higher place.

These bring me to an another outlier. Nick Collison. He sets great screens right? So he's close to top in RAPM numbers always.

With these thoughts, I want to see the "objective evidence". Why should anyone trust any metric more than a Larry Brown or Kobe or Lebron or Durant or Phil Jackson or Popovich or anyone else? I can't lie, I'm losing the interest and I'm losing it fast.