As you know, I am currently reading The Book of Basketball. If you’re an NBA fan, its a fun book, filled with basketball information and provocative arguments, along with a buttload of cultural references. (note: What’s with Simmons obsession with the movie Boogie Nights? There are — and I’m not exaggerating — at least a dozen references to the movie or its central character in the book.)
One of the early chapters in the book raises the greatest argument of all among NBA fans: who was better, Boston Celtics C Bill Russell or Phi/SF/LA C Wilt Chamberlain? In the book, Simmons argues that Bill Russell was better. In a prior post I criticized the nonsensical way Simmons put together his argument, but remained agnostic regarding his conclusion. Today I am prepared to say that while the evidence is close, I agree that Russell was the better “win force” than Chamberlain. (I take no position on who had more “basketball skill”. That is an impossible question to answer).
Applying Historic Marginal Win Score (MWS48)
I reached my conclusion after painstakingly constructing Career Win Resumes for each player using a version of the Marginal Win Score metric (explained in a separate Page on this blog) I call “Historic Marginal Win Score”.
I will write another Page explaining Historic Marginal Win Score, but for now I’ll just say that it is the same as MWS48 except it relies on inductive reasoning and historical precedent to fill in the statistical gaps that one encounters in every NBA season prior to 1977-78. For the main bit of missing information, Opposition Win Score, it works kind of like this. If I know I have a “2” (Team Win Score) and I know the final answer is around “5” (Pythagorean Wins), I can conclude that the missing number is probably around “3”. That’s way more simplistic, but its the gist. (The process also relies on “defensive position placement” that’s usually arduous but in this case is actually easy because I know both of these players spent all of their minutes at center. I’ll explain the process of placement when I do the Page).
Here are the Career Win Resumes I came up with for both players:
MWS48: Russell was the larger “win force”
As you can see from the two resumes, Bill Russell — according to Marginal Win Score — was the slightly larger “win force” (if you will). Meaning, throughout his career, and on a per minute basis, Russell outproduced his contemporary opponent centers by a bit more than Wilt Chamberlain in the categories that correlate with wins.
In a “typical” season for Bill Russell (for all the following numbers and terms, please refer to the “How to Read Win Charts” page in the blog column), the big man posted a Marginal Win Score per 48 of +6.10, he produced 20.2 wins for his team and (-7.2) losses, he was responsible for 13.7 wins above .500% (meaning if you added him to a 41-41 team, he would typically make that team a 55-27 team… absolutely Ruthian impact), and his Win Contribution Index would be +1.010. In Chamberlain’s “typical” season, he posted a Marginal Win Score of +5.08, he produced 20.6 wins and (-5.4) losses, he was responsible for +13.0 wins above .500%, and his Win Contribution Index would be +0.932.
So while Chamberlain produced slightly more wins in a typical season, he needed nearly 400 more minutes per season to do so. Thus in my opinion — while the decision was a close one — Russell was nevertheless the more valuable player. Russell was more efficient with his marginal production, and Chamberlain’s extra minutes, while valuable, could not overcome that fact. (Please also note that part of the reason for Chamberlain’s win advantage was that in Russell’s first three seasons the NBA played only a 75 game schedule).
Russell also gets the advantage because he was more consistent. Up until his last two seasons, he produced MWS48s of +5.00 or better every single season, with most seasons being +6.00. To get a feel for how awesome that kind of production is, check out the Win Chart from last season’s Cleveland Cavaliers.
As you can see, last season’s NBA MVP, LeBron James, had an MWS48 of +6.00, remarkable production. By my estimates, Bill Russell averaged better than that for his career. Repeat, he outproduced his opponents, on average, at a rate slightly better than LeBron James did in his spectacular 2008-09 MVP season. Absorb that. (Also, don’t get the idea in your head that Russell produced such awesome numbers only because he had some astronomical physical or athletic advantage over the 1960s competition. Not so. Go on Youtube and search “NBA 1965″ and watch some of the Celtic games that pop up. Russell doesn’t even really stand out in physical terms. And he played most of his career against high caliber centers the likes of Chamberlain, Zelmo Beatty, Wayne Embry, Nate Thurmond, Willis Reed, and Jerry Lucas. In other words, he was not picking low hanging fruit. He couldn’t shoot well, but he produced points, assists, and Rodman-like rebound numbers by playing with phenomenal passion and intelligence).
Chamberlain had the best single season
If you notice, while Russell has the better career average MWS48, in 1966-67 Wilt Chamberlain turned in the best single season when he somehow recorded an MWS48 of +8.11 and produced 28.7 wins for a 76er team that many rank among the greatest teams of all time. I have not calculated the Win Credits or MWS48s for more than a handful of seasons, but I would venture to say that Chamberlain’s 1966-67 season was hands down the greatest single season of all time.
A close runner-up, however, was Bill Russell’s 1964-65 season when he recorded an MWS48 of +7.75 and produced 26.0 wins for that outstanding Celtic championship team. Another great season, obviously. If you want to see the Win Credits I calculated for the entire ’65 Celtics, click here.
Chamberlain’s apparent inconsistency
Chamberlain’s Win Resume shows that he did not become the mega dominator that we remember him to be until he left his initial team, the Philadelphia/San Francisco Warriors. Why that is, I have no idea. The Warrior teams that featured Chamberlain were very sometimes poor defensive teams, especially considering they had a 7’1” force in the middle. Chamberlain must take some of the blame for that, and MWS48 gives it to him.
Once Chamberlain left the Warriors, it was mostly all uphill. He had three of the most dominant seasons in NBA history when he wore the “Phila” jersey of the 76ers, and he continued at a high level when he moved his act to Los Angeles.
In fact, the curious thing about both Chamberlain and Russell’s careers are they both retired while performing at a level that could be deemed “elite”. Unlike Kareem (and Shaq for that matter), who sort of hung on until there was nothing left of his productive capacity, Chamberlain and Russell seem to have either believed that one shouldn’t play past a certain age, or that they did not want to play anymore if they could not play at least close to the stratospheric levels they played at in their primes. (Click here for an interesting post on the aging of NBA stars)
Basketball-Reference’s Win Shares disagree
This summer the Basketball-Reference.com blog did a similar “fill in the blanks” calculation of wins produced by players prior to 1977-78 which they call “Historical Win Shares”.
Their results strongly disagree with mine. According to their calculations, Wilt Chamberlain was the far superior win producer. In fact, I would go so far as to say that Historical Win Shares considers Bill Russell to be just a “very good” player, not a great one.
A comparison of the numbers bears this out. While MWS48 estimates Russell produced 263 wins, Historical Win Shares estimates that Russell only produced 163 wins — obviously, a huge difference.
If you make the logical assumption that a player is responsible for 1/5th of a game every 48 minutes of action, and if you make the further assumption that a player is either producing wins or he is producing losses, then you can easily translate the results into wins and losses produced.
Bill Russell’s Career Wins Produced
Win Shares: 163.5 wins and 6.1 losses; .964% winning percentage
MWS48: 263.1 wins and (-93.5) losses; 1.551% winning percentage
Less than Tim Duncan?
Let’s put Russell’s numbers in perspective by comparing Russell’s winning percentages under the two systems to a similar contemporary player. Compare his winning percentage to Tim Duncan’s career winning percentage under Win Shares:
Tim Duncan’s career wins produced
Win Shares: 156.0 wins and (-14.3) losses; 1.100% winning percentage
MWS48 (roughestimate): 149.7 wins and (-8.0) losses; 1.056%
Those numbers show Win Shares regards Tim Duncan as a greater win force in his era than Russell was in the 1960s. While MWS48 basically agrees with Win Shares on Duncan’s win impact, you can see that MWS48 believes that Russell’s career production and win impact far exceeded Duncan’s.
Which win credit system is right?
That’s an easy question to answer. Neither. There is no “right” in this case. There are only ideas and estimates based on those ideas. The rest is conversation.
We will never settle the debate.
That said, I am more than happy to argue for a calculation that concludes that Bill Russell was a better than “very good” win producer.
When Bill Russell joined the Boston Celtics in 1956-57 they were a decidedly average team (1955-56 Pythagorean: 38-34). Immediately upon his arrival the team won its first championship. During his career the team went on to win 10 more championships in 12 seasons. Immediately following his retirement the same roster minus him slumped under .500%, and the Celtics did not win another championship for six seasons.
How much of that was due to Russell? MWS48 would argue that quite a bit of it was due to Russell, but that’s for you to decide.
The “With and Without You” test
Another unscientific test I like to use to judge how accurate MWS48 is at describing a player’s win impact is to look at the player’s team the season before and the season after his arrival.
In 1955-56, the Boston Celtics recorded 37.6 pythagorean wins and they were last in the NBA in opponent points per game. In Bill Russell’s first season, with virtually the same roster plus Russell, the team improved its Pyth wins total to 48.6, plus 9 wins. Win Shares credits Russell with producing 6.2 wins that season, MWS48 credits Russell with 11.4 wins. About equally off, with one shooting too high, the other two low.
In 1968-69, the Boston Celtics recorded 55.2 pythagorean wins, with Win Shares giving Russell 10.2 and MWS48 giving Russell 17.1. The very next season, with basically the same roster sans Russell, the Celtics recorded only 36.4 ptyh wins — (-18.8). MWS48 seems to capture the impact better.
Chamberlain’s career is less helpful because Win Shares and MWS48 agree most of the time. In Chamberlain’s first season the Warriors improved by 14.4 pyth wins. Win Shares credits him with 17.0 wins, MWS credits him wit 16.5 wins. Both are in the ballpark.
In Chamberlain’s first partial season with the 76ers the team improved by +10.1 pyth wins. Win Shares gives Chamberlain 7.6 wins with the Sixers, MWS48 gives him 9.0 wins.
In Chamberlain’s first full season with Philly the team recorded 21.6 more pyth, and Win Shares seems right on the mark, crediting Wilt with 21.4 wins that season while MWS48 gives him 24.6 wins.
What happened to the Warriors in his absence? In his final full season with the team, Win Shares credits Chamberlain with producing 25.0 wins, while MWS48 credits him with 22.3 wins. In their first full season without the Dipper, the Warriors recorded only 33.7 pythagorean wins, compared to 53.1 in his last full season with the team (-19.4 wins).
Finally, in Chamberlain’s last season with the Lakers, the team recorded 18.4 more pyth wins than they would record in their first season without him. In this case Win Shares seems dead on the money, giving Wilt 18.2 wins in his last season with LA while MWS48 is not far off, giving him 20.1 wins.
The results are not decisive, but it buoys me that MWS48 is, in every instance, right in the ballpark. (I’ve tested elsewhere in history — for instance, the Blazers last season with Bill Walton and first without — and gotten similarly encouraging results. Not decisive… encouraging.)
The long and short of it is that Historic Marginal Win Score finds that Bill Simmons argument that he made in his Book of Basketball, namely that Bill Russell was the more valuable player than Wilt Chamberlain, was accurate. MWS48 thinks it was close, but that the nod goes to Russell.
But not everyone agrees. Other metrics, namely Basketball-Reference.com’s Win Shares, believe Wilt Chamberlain was the far more valuable player. Win Shares believes that although the Celtics won 11 world championships during Russell’s tenure, and although their roster turned over at least three times in that span, Bill Russell was not even as valuable to the Boston franchise as Tim Duncan has been to the San Antonio franchise. (You like the Straw Man I built?)
Which “win credit” system comes closer to the truth? That’s for you, the jury, to decide.
However, if you ask me whether I am comfortable resting the validity of the MWS48 system on the argument that Boston’s Bill Russell was the greatest win producer in NBA history, my answer to you would be a resounding “yes”.
Footnote: If you are interested in seeing Historic Marginal Win Score applied to an entire season from the Chamberlain-Russell Era, click here.
PS — if you ever wondered how you could construct a team that included two of the absolute greatest players of all time and yet never even sniff an NBA title, go to the above link and click on the Cincinnati Royals.