This is our old blog. It hasn't been active since 2011. Please see the link above for our current blog or click the logo above to see all of the great data and content on this site.

Shocker of the Day: Losing Your Leading Scorer Hurts Your Offense

Posted by Neil Paine on May 18, 2011

In gathering links for StatHead yesterday, I came across this post at the Wages of Wins, wherein Prof. Berri mentions that the losses of Allen Iverson, Carmelo Anthony, & Rudy Gay did not hurt their respective teams. He then writes:

"In each of these examples, the loss of a scorer led people to forecast doom.  In each case, the team losing the scorer managed to survive and even improve.

Readers of The Wages of Wins and Stumbling on Wins understand this basic story. Scoring is overvalued by many NBA observers.   Top scorers do not always have the impact on wins that people imagine.  But no matter how often this story repeats, each time a scorer is lost we still see the same arguments offered by adherents to the conventional wisdom (for example, this week the Grizzlies insisted they would never dream of letting Gay depart)."

That's anecdotal evidence, though. What if we looked at every instance of a team losing its leading scorer? Would the typical team in that situation be impervious, or are those just a few cherry-picked exceptions to a larger rule?

Well, luckily, at BBR we have boxscores for every regular-season game since 1985-86. So I gathered our data, considering a team's "leading scorer" to be the player who led the team in PPG among players who played more than half of the team's games. I then looked at each team's offensive rating in every game, noting whether the designated "leading scorer" played in that game or not.

I also accounted for the strength of the opposing defense in each game by measuring how many pts/possession the opponent allowed in every game of the season except the one at hand. The end result will measure how well each offense performed relative to what we would expect from a league-average team facing the same opponent -- split by whether the team's "leading scorer" played or not.

For example, here is the data from 2011:

w/ Leading Scorer w/o Leading Scorer
Team Leading Scorer G Pts Poss ORtg avRtg vsAvg G Pts Poss ORtg avRtg vsAvg Diff
ATL Joe Johnson 72 6854 6382.1 107.4 108.1 -0.7 10 936 895.7 104.5 107.0 -2.5 1.8
BOS Paul Pierce 80 7707 7171.0 107.5 108.2 -0.7 2 206 199.6 103.2 111.0 -7.8 7.1
CHA Stephen Jackson 67 6284 6011.3 104.5 108.0 -3.5 15 1366 1323.8 103.2 107.4 -4.2 0.7
CHI Derrick Rose 81 7990 7307.3 109.3 108.3 1.1 1 97 97.0 100.0 108.0 -8.0 9.1
CLE Antawn Jamison 56 5332 5240.8 101.7 107.4 -5.7 26 2495 2377.4 104.9 108.4 -3.5 -2.2
DAL Dirk Nowitzki 73 7396 6657.7 111.1 108.4 2.7 9 824 795.7 103.6 107.4 -3.9 6.5
DEN Carmelo Anthony 50 5408 4773.4 113.3 108.2 5.1 32 3403 3032.7 112.2 108.2 4.0 1.1
DET Rodney Stuckey 70 6814 6271.6 108.6 107.8 0.9 12 1137 1058.0 107.5 107.0 0.5 0.4
GSW Monta Ellis 80 8256 7583.4 108.9 108.1 0.8 2 221 194.6 113.6 107.8 5.8 -5.0
HOU Kevin Martin 80 8465 7572.8 111.8 108.1 3.7 2 220 181.2 121.4 104.6 16.8 -13.1
IND Danny Granger 79 7930 7477.5 106.1 108.1 -2.1 3 253 279.6 90.5 105.1 -14.6 12.5
LAC Blake Griffin 82 8089 7612.1 106.3 108.2 -1.9 0 0 0.0 0.0 0.0 0.0 0.0
LAL Kobe Bryant 82 8321 7452.5 111.7 108.4 3.3 0 0 0.0 0.0 0.0 0.0 0.0
MEM Zach Randolph 75 7500 6886.8 108.9 108.3 0.6 7 695 668.8 103.9 108.0 -4.1 4.7
MIA LeBron James 79 8074 7159.7 112.8 108.1 4.6 3 295 269.7 109.4 107.4 1.9 2.7
MIL Brandon Jennings 63 5788 5661.2 102.2 108.5 -6.2 19 1746 1693.8 103.1 107.2 -4.1 -2.2
MIN Kevin Love 73 7380 7037.0 104.9 108.1 -3.2 9 908 867.3 104.7 108.2 -3.5 0.3
NJN Brook Lopez 82 7722 7427.8 104.0 107.9 -4.0 0 0 0.0 0.0 0.0 0.0 0.0
NOH David West 70 6657 6211.8 107.2 108.2 -1.0 12 1127 1048.9 107.4 108.6 -1.1 0.1
NYK Amare Stoudemire 78 8313 7433.5 111.8 107.7 4.1 4 421 388.1 108.5 107.4 1.1 3.1
OKC Kevin Durant 78 8203 7286.7 112.6 108.3 4.3 4 393 386.0 101.8 107.7 -5.9 10.2
ORL Dwight Howard 78 7762 7132.2 108.8 108.2 0.6 4 373 345.1 108.1 106.1 2.0 -1.4
PHI Elton Brand 81 8029 7458.3 107.7 108.0 -0.4 1 90 90.1 99.9 113.8 -13.9 13.5
PHO Steve Nash 75 7931 7154.9 110.8 108.1 2.8 7 680 655.8 103.7 106.7 -3.0 5.8
POR LaMarcus Aldridge 81 7810 7109.2 109.9 108.1 1.7 1 86 93.1 92.4 111.7 -19.3 21.0
SAC Tyreke Evans 57 5547 5366.9 103.4 108.4 -5.0 25 2604 2459.3 105.9 107.7 -1.8 -3.2
SAS Tony Parker 78 8115 7201.6 112.7 108.2 4.5 4 387 363.6 106.4 107.8 -1.4 5.9
TOR Andrea Bargnani 66 6627 6159.2 107.6 108.1 -0.5 16 1497 1444.0 103.7 106.1 -2.4 1.9
UTA Deron Williams 53 5296 4822.9 109.8 107.9 1.9 29 2857 2658.7 107.5 108.1 -0.7 2.5
WAS Nick Young 64 6207 6029.0 103.0 107.7 -4.7 18 1770 1705.9 103.8 108.1 -4.3 -0.4
League Average 2183 217807 201052.0 108.3 108.1 0.2 277 27087 25573.7 105.9 107.7 -1.8 0.9

As you can see, the average NBA team (weighted by possessions without the scorer) was 0.9 points/100 poss. worse on offense in games where their designated "leading scorer" did not play. As it turns out, this was actually a good year for teams who lost scorers; since 1985-86, the average NBA team is 2.0 points per 100 poss. better when their leading scorer plays vs. when he doesn't. (For the full dataset, click here.)

Which teams were impacted the most? With a minimum of 10 games missed, here were the teams whose ORtg suffered the most without their leading scorer:

w/ Leading Scorer w/o Leading Scorer
Year Team Leading Scorer G Pts Poss ORtg avRtg vsAvg G Pts Poss ORtg avRtg vsAvg Diff
2010 CHA Stephen Jackson 72 6981 6535.5 106.8 108.5 -1.6 10 832 883.1 94.2 108.2 -14.0 12.4
1997 PHO Kevin Johnson 70 7314 6546.6 111.7 107.7 4.0 12 1117 1126.7 99.1 106.9 -7.7 11.7
1997 ATL Steve Smith 72 6935 6255.7 110.9 107.5 3.3 10 839 851.7 98.5 106.9 -8.3 11.7
1999 NJN Stephon Marbury 31 2938 2796.6 105.1 102.9 2.2 19 1631 1760.0 92.7 101.6 -9.0 11.1
2008 CHI Ben Gordon 72 7073 6693.0 105.7 108.0 -2.3 10 908 935.2 97.1 109.6 -12.5 10.2
1997 ORL Anfernee Hardaway 59 5751 5247.0 109.6 107.6 2.0 23 1968 2005.9 98.1 106.2 -8.1 10.1
1996 MIA Alonzo Mourning 70 6890 6407.4 107.5 108.5 -1.0 12 1019 1047.5 97.3 108.3 -11.0 10.0
1993 ATL Dominique Wilkins 71 7721 6960.7 110.9 108.2 2.7 11 1093 1060.3 103.1 110.0 -6.9 9.6
1994 DET Joe Dumars 69 6742 6459.3 104.4 106.3 -1.9 13 1207 1243.4 97.1 108.6 -11.5 9.6
2007 NOK David West 52 5168 4727.3 109.3 107.5 1.8 30 2665 2689.6 99.1 106.7 -7.6 9.5
2000 PHI Allen Iverson 70 6746 6523.2 103.4 105.0 -1.6 12 1025 1082.4 94.7 105.0 -10.3 8.7
2007 TOR Chris Bosh 69 6917 6351.8 108.9 107.1 1.8 13 1240 1217.6 101.8 107.8 -5.9 7.8
1998 HOU Clyde Drexler 70 7014 6397.6 109.6 106.3 3.4 12 1085 1081.0 100.4 104.6 -4.2 7.6
1999 ATL Steve Smith 36 3188 3091.1 103.1 102.3 0.8 14 1127 1160.1 97.1 103.8 -6.6 7.5
2005 NJN Vince Carter 57 5339 5123.4 104.2 106.7 -2.5 25 2157 2210.0 97.6 107.5 -9.9 7.4
1993 WSB Harvey Grant 72 7396 7040.1 105.1 108.3 -3.3 10 957 976.8 98.0 108.5 -10.5 7.3
2004 LAL Kobe Bryant 65 6528 6075.5 107.4 103.7 3.7 17 1524 1508.7 101.0 104.6 -3.6 7.3
1999 CHH Eddie Jones 30 2872 2711.2 105.9 102.0 4.0 20 1772 1795.8 98.7 101.7 -3.1 7.0
2007 GSW BaRon Davis 63 6760 6190.1 109.2 107.3 2.0 19 1977 1933.6 102.2 107.2 -5.0 6.9
1991 SEA Eddie Johnson 66 7066 6304.6 112.1 108.4 3.7 16 1678 1608.5 104.3 107.5 -3.2 6.9

Meanwhile, these teams thrived without their leading scorers:

w/ Leading Scorer w/o Leading Scorer
Year Team Leading Scorer G Pts Poss ORtg xORtg vsAvg G Pts Poss ORtg xORtg vsAvg Diff
2004 ATL Shareef Abdur-Rahim 53 4712 4765.9 98.9 103.3 -4.4 29 2899 2696.9 107.5 104.0 3.5 -7.8
2009 PHO Amare Stoudemire 53 5606 5007.9 111.9 109.3 2.7 29 3368 2850.0 118.2 108.6 9.6 -6.9
2002 CHI Brad Miller 48 4155 4290.4 96.8 105.4 -8.6 34 3180 3110.6 102.2 104.7 -2.4 -6.1
1989 DET Adrian Dantley 42 4415 4059.6 108.8 108.8 0.0 40 4325 3811.4 113.5 107.9 5.5 -5.6
1994 LAC Danny Manning 42 4239 4193.0 101.1 106.9 -5.8 40 4208 3967.3 106.1 106.6 -0.5 -5.3
2000 LAC Maurice Taylor 62 5632 5804.6 97.0 104.2 -7.2 20 1914 1859.8 102.9 105.0 -2.1 -5.1
2005 ATL Antoine Walker 53 4807 4833.3 99.5 106.5 -7.0 29 2798 2676.6 104.5 106.9 -2.4 -4.7
2001 WAS Juwan Howard 54 4947 4958.4 99.8 103.4 -3.7 28 2698 2586.1 104.3 103.6 0.7 -4.4
2006 ORL Steve Francis 46 4291 4077.9 105.2 107.2 -2.0 36 3493 3196.8 109.3 107.0 2.3 -4.3
2003 SEA Gary Payton 52 4746 4614.2 102.9 104.1 -1.3 30 2809 2614.8 107.4 104.6 2.8 -4.1
1999 NYK Patrick Ewing 38 3259 3309.9 98.5 102.5 -4.0 12 1061 1034.2 102.6 102.6 0.0 -4.0
2007 CHA Gerald Wallace 72 6980 6726.9 103.8 107.2 -3.4 10 965 894.5 107.9 107.6 0.3 -3.7
1998 GSW Joe Smith 49 4300 4524.5 95.0 106.5 -11.4 33 2937 2988.2 98.3 106.0 -7.7 -3.7
1996 BOS Dino Radja 53 5521 5211.8 105.9 108.5 -2.6 29 2974 2730.0 108.9 108.1 0.9 -3.5
1989 CHH Kelly Tripucka 71 7416 7165.0 103.5 108.3 -4.8 11 1150 1078.3 106.7 108.1 -1.4 -3.4
2002 CLE Lamond Murray 71 6758 6431.9 105.1 105.1 0.0 11 1054 980.3 107.5 104.2 3.4 -3.4
2011 SAC Tyreke Evans 57 5547 5366.9 103.4 108.4 -5.0 25 2604 2459.3 105.9 107.7 -1.8 -3.2
2005 SAC Chris Webber 46 4699 4285.6 109.6 106.6 3.0 36 3806 3371.7 112.9 107.0 5.9 -2.9
1993 MIL Frank Brickowski 66 6739 6348.0 106.2 108.5 -2.3 16 1653 1515.9 109.0 108.5 0.5 -2.8
2008 HOU Yao Ming 55 5304 4975.6 106.6 108.4 -1.8 27 2627 2405.8 109.2 108.4 0.8 -2.6

Historically, though, those are the exceptions, not the rule. The complete data since 1986 says that when forced to survive without their leading scorer, teams often struggle to put the ball in the basket.

51 Responses to “Shocker of the Day: Losing Your Leading Scorer Hurts Your Offense”

  1. DSMok1 Says:

    Brilliant Research, Neil. Case Closed.

  2. brgulker Says:


    I appreciate the work, and the data set makes for some interseting stuff.

    I'm not sure it's an apples to apples comparison with Dr. Berri's point, however. DB's argument was premised on replacing a leading scorer with an "average" player (defined as .100 WP48).

    Here, though, you've simply subtracted the team's leading scorer.

    I don't have the time to go through all of this data to see who replaced the leading scorer in all of these games. It would be my sense, however, that these players weren't replaced by "average" players more often than otherwise.


  3. Neil Paine Says:

    I think there were two separate points made in that post -- that a team can survive the loss of its leading scorer (in general), and that replacing a team's leading scorer with an average player by WP48 will usually make the team better.

    I can't speak to the latter, but the former seems like a more practical concern for real-life teams. Average players aren't often in great supply on a team's bench, so they make do with what they have, and as we see from this data it can be very detrimental to the offense.

  4. yariv Says:

    It seems reasonable that missing most central players will hurt the team offensively. The leading scorer probably plays many minutes per game, of course. How about comparing the change in ORtg (vsAvg, as you did here) to fraction of team points in games the player participated in? (Of course, only for players playing many minutes per game, and missing enough games in the season to get meaningful data).

  5. ElGee Says:

    Great post Neil. Just as a case study -- particularly in response to the WoW claims about him -- look at Iverson's team offense w and w/out:

    06 4.8 worse
    05 2.2
    04 -0.3
    02 4.5
    01 2.2
    00 8.7
    99 5.7

    This was not a guy hurting his team on offense, despite such allegations of chucking and poor shooting efficiency. This is yet another piece of evidence his offense was lifting those teams.


  6. Dre Says:


    Really liked this! I like that in essence we're viewing several parts of the same problem. Dave's point is that scorers tend to be viewed as the "best" player on their team. Joe Johnson and Rudy Gay for instance got mad pay days and Lopez was called "untouchable" in trade talks for Melo (thank goodness!) As Ben pointed out if you found average players (in the case of the Hawks Childress, who has traditional been above average, was available for 1/3 the cost) to replace them you'd probably do better.

    Your point says "I have a team that uses player X to score, what happens if I lose them for part of the season?" As you and Ben agree, odds are a team doesn't have a productive player just rotting on the bench waiting to play (Rockets with Mutombo and Magic with Gortat though do come to mind) and they tend to replace their top scorer with something less desirable. So again this is a different angle and brings a fascinating question. What is the effect given my current roster of losing a given player? I'd argue Bynum is a better player than Kobe. However, if Bynum can't play the Lakers have two top big men in Odom and Gasol who can eat the minutes. If Kobe can't play, do we really think Artest or Brown is a good replacement?

    Finally the "games with and games without" is a tricky tricky scenario. I did the same thing you did for just Kevin Love ( and the fact is the only thing that changed in those games wasn't that Kevin Love went down. Darko also didn't play. The Wolves actually did have some decent players on the bench. In short this is not a single variable experiment. When a top scorer goes down other variables change and without adjusting for them we can't definitely say "Oh of course they matter a ton!" (sorry dsMok)

    Anyway loved the line of thought and would love to see more follow up articles on Wages of Wins Network pieces!


  7. brgulker Says:

    I think there were two separate points made in that post -- that a team can survive the loss of its leading scorer (in general), and that replacing a team's leading scorer with an average player by WP48 will usually make the team better.

    What he said was that 19 teams would get worse, and 10 of those teams would get much worse. I'm not sure how that supports your reading at all. In fact, it says the opposite for these teams -- replacing their leading scorer would be very, very hard to do.

    By DB's analysis, though, 11 teams would get much better if they replaced their leading scorers (because they're inefficient, really bad at other components of the game, etc.) with an average "win producer." The number 11 is a minority of teams, so again, I don't see how that supports your reading.

  8. ElGee Says:

    Btw -- would be awesome if you could do the same run with DRtg (either using leading scorer, or leading DWS guy or something).

  9. Neil Paine Says:

    #6 - Sorry about that, I misread the initial post, thinking the 19 represented teams that would be better without their scorer.

  10. Neil Paine Says:

    So... what, then? Sounds like we're actually in agreement that the majority of teams would be hurt with the loss of their best scorer. But that in turn seems inconsistent with the paragraphs I quoted at the beginning of this post, which implied that scorers are overrated and easily replaceable.

  11. Jason Says:

    If I am not mistaken, the WoW post was dealing with wins, and this post is concerned with offensive rating. It, indeed, isn't shocking that a team's offense could suffer without their leading scorer (especially if replaced by a below-average player). But it might be plausible that a team's wins are unaffected or improved after the loss of such a player.

  12. Jason J Says:

    I'm with #4. It would be interesting to see how different the results would be taking away any of the big minute players from a team.

  13. brgulker Says:


    Regarding comment #10, that's what I was trying to say. DB's post did conclude that many teams would be significantly hurt by losing their top scorers. He wasn't arguing against that point in his piece, but that seems to be how you read it.

  14. David Says:

    I do not see your post and Berri's posts in disagreement really. Berri's central claim is that "Top scorers do not always have the impact on wins that people imagine" and that, more generally, scoring is overvalued in player assessment. What you've shown here is that teams have difficulty in scoring once they lose their leading scorer. Maybe I'm being dense but how are those 2 statements in opposition? Maybe look at W-L when the best scorer is missing would be more apples-to-apples?

  15. David Says:

    Let me add to the above as I've re-read both posts. Berri replaces the lead scorer with an average player and sees most teams (using 2010-2011) improving. The conclusion here is that a high(est) PPG player does not perforce translate into an enhanced probability of winning a game. With Berri's metric this is simply that a high PPG player has a sub-average metric. Picture, say, someone with negative net turnovers, shooting low 40s but scoring 25 pts. You look at this differently (conditionally) and get 2 pts less per game. Scoring 2 pts less per game does not equate to winning less though. What if the replacement player is better (in effect what Berri found with his metric)? A defensive ace (so Drtg more than compensates)? What was the "mean player" used to replace the missing scorer in your tables above? I think that's the a good way to make the studies more comparable. Good reads all around.

  16. Jason J Says:

    Just as an aside on this whole topic, Prof Berri's WP, WP48, and PAWS metrics will always tend to downplay the value of scoring wings who don't provide large rebounds because of the nature of the math, which is very pro-rebound and anti-usage. The return on points per shot and assists is low compared to the subtraction for misses and turnovers.

    I'm not saying whether I agree or not, but anyone familiar with his methods could have predicted that Berri would conclude that losing a Melo or an AI would be a good thing, right or wrong.

    On the other hand, if Prof Hollinger ran a similar test using his methods, PER and the like, he'd likely have a very different read on things because PER favors high usage and is less impressed by rebounding generally. Which again, might be right or wrong.

  17. yariv Says:

    #16, this is a good point, and the reason Neil's method is important. Of course, you can look into Win% or ORtg-DRtg, but it's interesting in any case. There are much less assumptions hiding, you work with reality more than with an abstract model. Then again, you're replacing the missing player with a player from your bench, not an average player, so you should adjust for that (something like average difference over MP/G seems reasonable point to compare to).

  18. ElGee Says:

    I think what Jason (#16) is saying is that if I used a metric that were anti-assists -- whether that accurately reflected reality or not -- I could obviously run a simulation "showing" that the removal of all top point guards helps teams **by my metric.** That doesn't mean in real life teams would be better off if we removed all the top PG's.

    There are clear reasons why the math used behind WP concludes rebounders are more important than scorers. In this case, "leading scorer" is going to encompass a range of players from Kobe Bryant, to Allen Iverson, to Allan Houston, to the default leading scorer on the 2011 Cavs. Berri's claim is from a metric that thinks Reggie Evans is way more valuable than Allan Houston. He's replacing many of the non-Kobe leading scorer types with players who are better by WP. So, obviously the teams will be better in those forecasts. After all, he's saying "replace someone that WP thinks is better than Amare Stoudemire, and WP thinks New York will be better."

    Here, Neil uses actual data to show that's not what happens on average. And while scoring is overrated (agreed), it's not nearly as overrated as the ridiculous conclusions from the WP regression would suggest. Iverson -- one of the people singled out in the article -- actually helped Philadelphia's offense when he played there. In 2007, when Iverson left, the 76ers ORtg went from average to 26th in the league. I guess that counts as "not hurting the team." *shrug*

  19. Panic Says:

    Having just read Berri's piece, I'm not sure what point he's trying to make. He waxes poetic at first about how overrated leading scorers are. Then he uses his own (dubious) methods to show that in most cases they are, in fact, quite valuable. He devotes about three times more words to the players who "make their teams worse" and never comes out and says the average loss of wins across all thirty teams (or even reports it - you have to calculate it yourself if you think this number - which should be the whole point of the piece - is worth knowing).

    All in all, it's pretty classic Berri, for whom intellectual honesty rarely seems to be a priority.

  20. AHL Says:

    2009 Stoudemire can suuuuuck it.

  21. AYC Says:

    Wait, are you suggesting Dennis Rodman isn't the second most productive player of the last 30 years??

  22. Jacob Says:

    Cool stuff. Still, I'm with #11 - I think it's more interesting to see the impact on W% (or ORtg/DRtg) than on ORtg only. Losing Carmelo will cost you points, but what is the net gain if his sucky defense is removed?

    Better yet, we could tweet the resulting table to Bill Simmons, so he can finally validate his Ewing Theory.

  23. Neil Paine Says:

    If we look at W% instead of ORtg, though, wouldn't we also simply be measuring the scarcity of great two-way players, instead of isolating the effect of a high-usage player on an offense?

    I suppose it comes down to whether this is strictly an offensive question -- are high-usage players really necessary to an offense? -- or a question of whether scorers tend to be so bad in other areas of the game that it offsets their offense. I can see pros and cons to both interpretations.

  24. yariv Says:

    There might be some correlation between scoring and defence, both because of the basic skills required (physical skills) and due to cultural effects (scoring is overrated among players as well, I would guess). If such a correlation exists, the question about how will a team manage when losing the main scorer might not be the same as how the offence will manage. I would say that the general question is more important, although of course both are interesting.

    Neil, what do you think about testing the effect of losing high-minute player in general? It seems to me that this should be the baseline to what you did in this article. Maybe 0.02 points/possession is actually not much when losing such a player?

  25. DSMok1 Says:

    Yeah, dre, I rather regretted that terminology after posting that comment.

    One must admit the With-or-Without-You research is pretty airtight in terms of valuing a player for that specific team. It goes straight to the core of a player's valuation, and has a large enough sample size to be meaningful. In fact, handled properly, I think WOWY like this could be used to validate individual player models/replacement level settings.

    It honestly didn't come to a very different answer than Dave did--most teams are hurt somewhat by losing their best scorer, but not all of them. In fact, shouldn't Dave take it, to some extent, as validation of what his model said? This WOWY analysis has the advantage of not being biased by handling of box scores, and gives a very meaningful result in terms of the spread of "best scorers" actual impact on their team. Note, of course, the difference between replacing with a league-average player, and replacing with....whoever sat on the bench.

  26. Neil Paine Says:

    #24 - That is a good idea. How would you break it out -- maybe two groups, players who led the team in MPG but not PPG, and players who led the team in PPG but not MPG? There's probably going to be significant overlap between PPG and MPG leaders, though.

  27. Dre Says:

    Wins Produced and Win Shares, which are built on the same principles essentially work in aggregate (I know how all of the games played out so I can get a score). You actually have opened a really fun case worth more investigation (consider it added to my queue) If I have a team of X players I can use any metric (and no Dsmok the plus-minus family ones are not an airtight set) to get a feel. A really important question is which players I can't afford to lose or put another way I need to make sure I have a back up for. For example, on Los Angeles I argue Kobe is actually a player they can't afford to lose given their weakness at Guard. On Dallas, Arturo's actually said with and without Dirk was a huge difference. This is actually quite a different question than if scoring is overrated, which I hope we can agree to some degree is.

  28. yariv Says:

    I think I would assume that the effect on the team should be proportional to MPG, and compare PPG leader to all other "main" players (MPG>25 sounds reasonable requirement). Compare WOWY-diff/MPG for both groups. Comparing only MPG-leader is very little data, probably, because of the overlap between PPG and MPG leaders.

  29. yariv Says:

    Dre, I think that WOWY measuring games (success in games with and without the player), as opposed to plus/minus on minutes, is airtight when it can be used. The problem is that for most players we don't have enough data to make this argument. I was really impressed by Benjamin Morris' work with it in "The Case for Dennis Rodman".

  30. Neil Paine Says:

    #28 - Huh, this actually isn't as bad as I thought. Since 1986, there have been 727 team-seasons. Of those, 473 (65%) had the same person as their MPG and PPG leader (minimum half of team games played). That leaves a 254-team sample where the MPG leader and PPG leader were two different players.

  31. Michael E Sullivan Says:

    I would suggest that nobody every suggests that losing your leading scorer in general is no big deal. I think many people suggest that losing a leading scorer who isn't particularly efficient and doesn't play solid defense is no big loss.

    High scorers who are relatively efficient are extremely valuable, and their loss will invariably hurt unless replaced by a similar caliber player (possible in trade, but unlikely when due to injury or other problem).

    OTOH, High scorers who are inefficient and weak defensively, may not actually be more valuable than the next guy on the bench. Rudy Gay, for instance, was a 15 pt. scorer in his rookie year, but his WS/48 were atrocious at .011, probably below replacement level. Every year except this year his WS/48 were under .1 suggesting he was a relatively average player. (think about it, 5 guys on the floor if they average .1, you have a .500 win chance). And I don't think he has a ton of intangibles that WS isn't capturing. He's a guy who takes a lot of shots, but is no more efficient than average. That's not a guy who creates value. Even this year, his first above average at .123, he was seventh on his team in WS/48, and only a bit ahead of guys 8 and 9. So Memphis had a nice stable of above average players this year, and most of Rudy's minutes ended up being replaced by guys who were about as good as he was or better.

    The guy who can take a lot of shots and be *more* efficient than average, that guy creates value and helps win games. Hello, Dirk's game last night. Also, welcome to his .214 lifetime WS/48 average. Lose Dirk, and you lose a lot.

    Basically looking at all leading scorers lumps all your huge stars that score like Dirk, Dwade, Lebron, Jordan etc. in with the guys who are a bit suspect because they are much less efficient. I don't think anybody doubts that when Dirk or Kobe goes out, it hurts you offensively. The question is whether a guy like Anthony makes as big a difference as his proponents think.

  32. yariv Says:

    #30, How many games did those missed (lower of the sums for 1-the 254 PPG leaders, 2-the 254 MPG leaders)? It should be enough to draw some conclusions. I would suggest looking at the distance in each measure as well. The PPG-leaders I would expect to be very close in MPG to the actual leaders, but not the other direction. If this is true, then we can expect that if MPG is the main factor they will have close WOWY-values, whereas if PPG is a significant they might not even be close. What I'm saying is that you should present the weighted average of MPG and PPG of both groups, and the weighted WOWY-numbers should be compared to both, if the PPG-leaders have only slightly higher WOWY-value, it might still imply MPG is a stronger factor than PPG (depending on numbers, as suggested above).

  33. Jason J Says:

    Good thoughts buzzing around here Yariv & Neil. I'm interested to see how this turns out if you come up with something.

  34. false~cognate Says:

    My one issue is that you are comparing stats for the team while said leading scorer was STILL on the team, ignoring that teams can be assumed to being built with a plan. If you have a scorer, maybe you don't move to replace him until you trade him or he leaves via free agency. I'd like to see the same analysis done comparing teams after permanent departures.

  35. Nick Says:

    Wouldn't the average numbers reflect the situation better if you removed the top scorers who play every game from the mix? Having guys who are irrelevant to the point being made, but still create statistically significant data (since they are a guaranteed 0.0, below the average) seems like a bad idea.

  36. yariv Says:

    Nick, The averaging is weighted by number of games (missed), otherwise it would be way too noisy. At least, I assume this is the case, also it gives the same number...

  37. Sean Says:

    Losing a leading scorer just sounds like a big hole to fill. I would think it depends on who is left to fill the void, what skill set the missing player has (beyond scoring) and to what degree that missing player's 'game' is infused with his teamates' respective 'games'.

    If a player does much more than lead the team in scoring and is a facilitator for the other players' games------that team will have more problems than if the player who leaves was existing more on an island away from teamates, even if he is the leading scorer. Less avenues to scoring are damaged if a player that goes missing wasn't so involved with everyone else's scoring to begin with.

    Also, if a team loses a leading scorer, they might play at a different pace without him--------and the 'new' offensive numbers might indicate a problem where there really is none. A team could play at a different pace offensively and be relatively just as effective.

  38. yariv Says:

    #37 the numbers are for offensive rating, so it's already adjusted to pace. Now, the team might have better defence and so overcome the weakened offence, but the weakened offence is still a problem (on average).

  39. Mikewillblitz Says:

    This article just points out what I have been saying about the Bulls... oh look, another article on why the bulls won't win it:

  40. Sean Says:

    Thanks, Yariv, for clearing the pace issue up for me.

  41. marparker Says:

    Apparently mentioning DB will get you 40 comments around here.

    I'm wondering why you didn't use efficiency differential to make your point. What does offensive efficiency tell you by itself?

  42. Neil Paine Says:

    41 now (well, actually, 42).

    "Top scorer" seems to be primarily an offensive designation, no?

  43. marparker Says:


    I thought it was understood that maximizing efficiency differential was the goal.

  44. Guy Says:

    Any chance you could replicate your games played/not played analysis, but instead of looking at each team's top scorer look at the player with the highest Wins Produced? That is, repeat Dave Berri's earlier analysis of how much teams would lose without their highest WP player, but using your methodology. It would be interesting to see the actual change in efficiency differential, compared to the projected change based on WP.

  45. Neil Paine Says:

    #43 - I think there are 2 semi-related but ultimately distinct questions here:

    * Are scorers overrated offensively? (i.e., is high-volume shooting easy to replace on offense?)

    * Are scorers overrated as all-around players? (i.e., are they so bad at defense, passing, etc. that the impact of their scoring is offset?)

    The first question is what I was driving at, and it's strictly an offensive one. However, plus/minus regressions have shown that volume shooters do tend to be worse defensively than low-usage players, all else being equal. This could be because players who don't create shots must be doing something else to justify being on the floor, but it could also imply that scorers tend to be bad defenders, in which case #2 might be true even if #1 isn't.

    #44 - Now that's a great idea. Might have to use win score instead of WP. Replicate the (pending) study of MPG vs PPG by looking at winscore/G vs PPG.

  46. marparker Says:


    Now, I see where you are coming from more clearly.

  47. Owen Says:

    "I think there were two separate points made in that post -- that a team can survive the loss of its leading scorer (in general), and that replacing a team's leading scorer with an average player by WP48 will usually make the team better."

    That's a pretty bizarre reading of the post, especially "in general" and "usually."

    I think what he is saying is that some percentage of the time, maybe 25%-30% of the time, if you replaced a team's leading socrer with an average player the team would improve. That's because scoring is overrated.

    Honestly, this isn't even controversial is it?

    What happens to the Bulls if Kyle Lowry replaces Derrick Rose? Would love to see what happens there.....

  48. Neil Paine Says:

    I already corrected myself on the second remark -- I misread the initial DBerri post as saying that 19 of 30 teams could be improved by replacing their leading scorer with an average WP48 player. But the actual conclusion of his post makes the opening quotation seem even more like an oversell.

  49. Jay Says:

    Why doesn't this site run Wins Produced numbers for players? If PER can get on there, I see no reason why WP shouldn't.

  50. Matt, Colombia Says:


    How does 25-30% of teams improving because they lost their leading scorer show that scoring is overrated? It may show that those particular teams overrate it, but the majority of the league still doesn't.

    Also, he's only saying that his method shows that without actually giving any proof of it through looking at whether those teams actually improved without those example players he uses. Philly and Denver both added players. Memphis didn't really win that much more without Gay and really could have used him against the Thunder for shot creation.

    That 25% of the league has a high volume low-efficiency scorer leading their team in ppg, just goes to show valuable a skill efficient scoring really is. If those teams had someone who could create their own shot and score efficiently, they would be doing it, or so one assumes.