Posted by Neil Paine on December 14, 2009
One of the long-held NBA aphorisms used to explain the gulf in championships between Bill Russell and Wilt Chamberlain, arguably the two greatest centers of all time (OK, Kareem & Shaq might dispute this notion, but that's another argument for another day), is the idea that Russell played on Celtics teams stacked to the rafters with All-Star and/or Hall of Fame talent, while Chamberlain suited up with lesser teammates who held him back. Certainly that's the impression you'll get from our Quality of Teammates post, where we found that Russell played with the 10th-most talented set of teammates (weighted by career minutes played) of any player in NBA history, while Chamberlain's teammates were average at best, and hardly spectacular like Russell's.
In fact, according to career Win Shares per minute, Russell's teammates were worth 8.10 WS/3000 MP over his career, while Chamberlain's were worth 6.06. So if both men played 3,300 minutes per season, with a schedule of 80 games and 48.3 MPG (the NBA's all-time average), that gives roughly 16,000 minutes to each center's teammates in total for each year:
((8.10 - 6.06) / 3000) * 16000 = 10.88
In other words, Russell's teammates alone were worth approximately 11 more wins than Chamberlain's per regular season... And in the playoffs since 1957, teams with 10-12 more regular-season wins than their opponent won 71 of 85 series (83.5%). So should it have been any surprise that Russell and the C's were coming out ahead of Chamberlain's Warriors & Sixers?
Bill Simmons disagrees. In The Book of Basketball, Simmons tries to make a case that Chamberlain's teammates were in fact closer to Russell's than anyone thinks, going through the rosters on a year-by-year basis to see who had the stronger supporting cast. He finishes by saying:
"Russell played with four members of the NBA's Top 50 at 50 (Havlicek, Cousy, Sharman, and Sam Jones); Wilt played with six members (Baylor, West, Greer, Cunningham, Arizin, and Thurmond). And Russell's teammates from 1957 to 1969 were selected to twenty-six All-Star games, while Wilt's teammates from 1960 to 1973 were selected to twenty-four. Let's never mention the supporting-cast card again with Russell and Chamberlain. Thank you."
Er, sorry Bill, but I'm going to have to take a closer look... Let's go through and see how many minutes Russell and Chamberlain played alongside fellow members of Simmons' own Pyramid, his top 96 players of all time. We'll calculate "minutes together" like this: In 1960, Wilt played 3,338 minutes, or 92.1% of the Warriors' total court time. Tom Gola played 2,870 MP, or 79.2% of the Warriors' total. The probability of Wilt and Gola being on the court at the same time is given as p(Wilt_Gola) = 0.921 * 0.792 = 0.729, and therefore the expected # of minutes Wilt and Gola played together would be 72.9% of Philadelphia's 3,624 total MP, or 2,644 minutes. We can do this with every teammate the two great centers ever had, and see how much court time they logged alongside members of Simmons' Pyramid over the course of their careers.
What does this tell us? Well, not much at first glance, except that Russell's teammates were a bit better. But let's turn those rankings into points, giving 96 for 1st place, 95 for 2nd, etc. until 0 for unranked players. Then take a weighted average by minutes played:
Now you can see Russell's "score" is more than twice that of Wilt, but I think we're also overselling the unranked players by this method -- in essence, we're tying every unranked player as the 97th best of all-time. However, 3,902 players have suited up in NBA history, meaning the typical unranked player's ranking should be more like 2000th overall. So, finally, what if we took another weighted average ranking but assigned #2000 to all unranked players?
Obviously this is just a fun exercise, and far from scientific, but you can still see that Chamberlain's teammates were in fact significantly less talented than Russell's, by both our Quality of Teammates metric and even by Bill Simmons' own ranking method. So I don't think it's quite fair to say, "let's never mention the supporting-cast card again with Russell and Chamberlain," because it's still pretty obvious that Wilt's supporting cast was inferior to Russell's by a good margin.