Championship Usage Patterns II
Posted by Neil Paine on May 19, 2010
After yesterday's post about optimal championship usage patterns, I got a lot of good feedback about possible alternative versions of the same study that would better capture the effect I was going for. When setting up for the initial study, I struggled between sorting by minutes played and by raw modified shot attempts (MSA), each of which had unique advantages. But a nice compromise (suggested by reader Brian) would be to isolate the top 5 players on each team by minutes -- thereby approximating their most frequent 5-man unit -- and then sort by MSA%, the percentage of team MSA that each player took while on the floor:
| Modified shot att% by team rank | ||||||
|---|---|---|---|---|---|---|
| Year | Team | #1 MSA% | #2 MSA% | #3 MSA% | #4 MSA% | #5 MSA% |
| 1952 | Minneapolis Lakers | 31.6 | 22.9 | 18.1 | 16.3 | 15.7 |
| 1953 | Minneapolis Lakers | 28.4 | 23.1 | 21.1 | 15.6 | 15.6 |
| 1954 | Minneapolis Lakers | 29.3 | 19.3 | 18.2 | 17.1 | 16.2 |
| 1955 | Syracuse Nationals | 25.8 | 20.7 | 19.8 | 18.5 | 15.0 |
| 1956 | Philadelphia Warriors | 26.6 | 23.0 | 22.7 | 17.9 | 16.4 |
| 1957 | Boston Celtics | 25.8 | 24.2 | 22.5 | 17.2 | 16.6 |
| 1958 | St. Louis Hawks | 25.7 | 25.0 | 18.0 | 17.2 | 14.3 |
| 1959 | Boston Celtics | 24.7 | 24.0 | 23.2 | 23.1 | 13.6 |
| 1960 | Boston Celtics | 26.0 | 25.8 | 22.9 | 17.9 | 15.5 |
| 1961 | Boston Celtics | 26.8 | 24.2 | 20.7 | 20.0 | 17.3 |
| 1962 | Boston Celtics | 26.8 | 25.4 | 22.2 | 17.5 | 12.0 |
| 1963 | Boston Celtics | 27.0 | 26.0 | 23.1 | 16.8 | 12.9 |
| 1964 | Boston Celtics | 25.7 | 22.7 | 15.6 | 14.5 | 14.0 |
| 1965 | Boston Celtics | 25.3 | 24.2 | 17.3 | 14.5 | 13.8 |
| 1966 | Boston Celtics | 27.1 | 24.4 | 17.7 | 16.8 | 12.3 |
| 1967 | Philadelphia 76ers | 24.1 | 22.3 | 20.2 | 18.0 | 13.1 |
| 1968 | Boston Celtics | 24.5 | 24.1 | 20.5 | 19.6 | 15.8 |
| 1969 | Boston Celtics | 27.0 | 23.4 | 19.1 | 16.3 | 13.2 |
| 1970 | New York Knickerbockers | 24.0 | 20.6 | 19.2 | 18.3 | 17.6 |
| 1971 | Milwaukee Bucks | 25.5 | 24.7 | 22.8 | 16.6 | 14.9 |
| 1972 | Los Angeles Lakers | 30.5 | 25.3 | 18.7 | 14.8 | 12.5 |
| 1973 | New York Knickerbockers | 22.0 | 22.0 | 21.5 | 20.6 | 19.8 |
| 1974 | Boston Celtics | 26.0 | 23.1 | 20.8 | 14.1 | 12.9 |
| 1975 | Golden State Warriors | 30.8 | 23.8 | 21.6 | 18.9 | 9.7 |
| 1976 | Boston Celtics | 25.0 | 24.1 | 22.5 | 19.2 | 11.9 |
| 1977 | Portland Trail Blazers | 25.2 | 22.8 | 20.9 | 17.7 | 17.6 |
| 1978 | Washington Bullets | 23.8 | 23.2 | 21.3 | 18.1 | 11.5 |
| 1979 | Seattle Supersonics | 32.0 | 24.1 | 17.6 | 17.5 | 17.3 |
| 1980 | Los Angeles Lakers | 26.5 | 21.2 | 20.7 | 20.4 | 16.7 |
| 1981 | Boston Celtics | 23.7 | 22.4 | 21.8 | 17.1 | 15.7 |
| 1982 | Los Angeles Lakers | 24.8 | 23.1 | 21.1 | 20.8 | 19.3 |
| 1983 | Philadelphia 76ers | 30.1 | 24.6 | 23.6 | 21.9 | 13.9 |
| 1984 | Boston Celtics | 25.9 | 22.5 | 18.6 | 16.9 | 14.5 |
| 1985 | Los Angeles Lakers | 25.1 | 24.1 | 20.4 | 20.2 | 15.4 |
| 1986 | Boston Celtics | 25.1 | 21.8 | 20.4 | 19.9 | 17.2 |
| 1987 | Los Angeles Lakers | 27.3 | 22.5 | 21.9 | 18.4 | 18.2 |
| 1988 | Los Angeles Lakers | 26.2 | 24.8 | 22.0 | 21.4 | 12.8 |
| 1989 | Detroit Pistons | 29.1 | 24.5 | 20.8 | 15.7 | 10.4 |
| 1990 | Detroit Pistons | 27.3 | 23.4 | 15.4 | 14.9 | 9.5 |
| 1991 | Chicago Bulls | 34.6 | 24.9 | 14.8 | 14.1 | 13.7 |
| 1992 | Chicago Bulls | 36.4 | 24.5 | 13.6 | 13.4 | 11.3 |
| 1993 | Chicago Bulls | 38.8 | 25.5 | 14.7 | 13.8 | 13.5 |
| 1994 | Houston Rockets | 29.7 | 21.2 | 17.8 | 17.8 | 13.5 |
| 1995 | Houston Rockets | 33.7 | 24.0 | 17.7 | 15.7 | 13.3 |
| 1996 | Chicago Bulls | 33.4 | 24.0 | 21.2 | 16.6 | 11.1 |
| 1997 | Chicago Bulls | 35.7 | 24.3 | 15.6 | 14.2 | 9.6 |
| 1998 | Chicago Bulls | 36.7 | 24.2 | 20.8 | 14.1 | 9.8 |
| 1999 | San Antonio Spurs | 25.4 | 22.0 | 21.3 | 18.1 | 15.9 |
| 2000 | Los Angeles Lakers | 30.0 | 26.7 | 17.6 | 17.1 | 15.0 |
| 2001 | Los Angeles Lakers | 30.9 | 30.0 | 15.4 | 13.6 | 12.9 |
| 2002 | Los Angeles Lakers | 30.4 | 29.9 | 16.2 | 13.9 | 12.3 |
| 2003 | San Antonio Spurs | 26.7 | 24.8 | 19.1 | 18.6 | 11.9 |
| 2004 | Detroit Pistons | 27.4 | 23.9 | 20.3 | 16.4 | 15.0 |
| 2005 | San Antonio Spurs | 31.0 | 26.7 | 26.1 | 16.3 | 9.5 |
| 2006 | Miami Heat | 32.8 | 25.5 | 19.9 | 18.9 | 14.3 |
| 2007 | San Antonio Spurs | 29.7 | 28.6 | 28.0 | 18.9 | 9.7 |
| 2008 | Boston Celtics | 25.9 | 25.4 | 22.9 | 19.3 | 10.4 |
| 2009 | Los Angeles Lakers | 34.3 | 18.3 | 16.7 | 15.7 | 15.2 |
| Average Champ |
28.2 | 23.9 | 19.9 | 17.3 | 14.0 | |
| Avg. Non-Champ | 27.6 | 23.5 | 20.4 | 17.7 | 14.0 | |
So, in the playoffs, teams that win it all tend to have a small (but significant) tendency to allocate more possessions to their top two options -- and fewer to their #3 & #4 options -- than teams who lose. In other words, the model of a superstar "Alpha Dog" plus another quasi-star and 3 role players seems to be the dominant usage pattern that differentiates NBA champs from teams that fall short.
This is borne out by the logistic equation that best models the data above:
p(Championship) ~ 1 / (1 + exp(3.23 - 0.01*MSA#1 - 0.10*MSA#2 + 0.10*MSA#3 + 0.01*MSA#4 - 0.03*MSA#5))
All else being held equal, the effect is strongest when looking at the #2 and #3 options in the primary lineup; the closer the two numbers are, the worse the probability is that the team in question will win the championship. For instance, a team with an allocation of 27%/21.5%/20.5%/17%/14% will be expected to win the title 6.8% of the time, but for a team with an allocation of 27%/24.5%/17.5%/17%/14%, the expected p(C) becomes 11.7% -- almost doubled from the first lineup, simply by allocating 3% more MSA to the #2 and 3% less to the #3.
The 2010 playoff team that seemed to fit this model best was the OKC Thunder, whose Kevin Durant-Russell Westbrook combo was reminiscent of great 1-2 punches of days past. However, it's important to remember that the model only works when holding everything else equal, and the Thunder's overall talent level (#9 in SRS) was not equal to that of the best teams in the league this year. However, among the Final Four teams (who should theoretically be roughly equal in talent), the Lakers are the best fit for the model -- their Alpha Dog's ability is undisputed, and perhaps more importantly, the gap between their #2 (Pau Gasol) and #3 (Ron Artest) is 4.6 points of MSA%. The teams the formula likes the least are Phoenix and Boston, who have essentially no difference between their #2 and #3 options in terms of MSA% (the Celtics, with Rajon Rondo + the Big 3, take this to a real extreme: less than 4 points of MSA% separate their #1 option from their #4).
May 19th, 2010 at 1:06 pm
Much better that part I, I'd say.
May 19th, 2010 at 2:57 pm
The 1973 Knicks might never be beaten for flatness. The 1978 Bullets total under 100, which I suppose says you can win if you keep your mad bombers mostly on the bench.
The 2010 Celtics look a lot like the 1982 Lakers, only older.
May 19th, 2010 at 4:47 pm
I'm not sure what to think here. Averages are ok but there is a lot of variance and the sample size is pretty small. Caveat: these are generic criticisms, and I'm not trying to say that statistical studies of NBA finalists are futile.
In this case though, it might be good to compare these teams to NBA average, or teams with a losing record. Either way the variance makes me hesitant to draw conclusions from the averages. There are a good number of teams with the third guy at 15-16%, and a good number with the third guy over 20%. We sort of know intuitively that champions usually have a dominant player, a second all star, and beyond that it varies. These numbers confirm that idea but I don't see a lot else there.
May 19th, 2010 at 5:12 pm
Remember, for every data point in that table, there are 10-15 that I didn't show (i.e., teams that didn't win the championship). The general trend is apparent from the regression on 700 teams -- between two teams with equal offensive skill levels, the one that's more balanced between its #2 & #3 options is less likely to win a championship.
May 19th, 2010 at 5:20 pm
This really doesn't have a whole lot to do with this post, but I thought of it after watching last night's game and wanted to mention it somewhere. Can we please stop referring to Pierce, Garnett, and Allen as "The Big 3"? I think the 2010 Celtics can much more adequately be described as "Rajon Rondo and the Medium 3"
May 19th, 2010 at 9:57 pm
Sort by column number one and you find the Alpha Dog being the Alpha Dog again.
May 19th, 2010 at 9:57 pm
Meh, Column three (#1 MSA%)
May 20th, 2010 at 4:53 pm
I'm new to this advanced stat stuff, but do find it interesting (always thought the sexy stats really don't say much).. So who are the best players within their msa% and what would be the ultimate nba team and/or us national team based on this theory?
May 20th, 2010 at 5:13 pm
Hmm, how about weighting things? I think the 2001 LAL squad had the most dominant playoff run ever. And they had 1a and 1b + "fluff". So the top tier playoff runs should maybe more inform the average? Also, by eyeball the "good" championships teams seems to have the a total of ~60 for the top 2...
May 26th, 2010 at 5:23 pm
Looking at the average post-merger, from '77 to present, the two top dogs split looks even more pronounced.
29.56 24.22 19.58 17.19 13.57
Not sure that making a split there (or somewhere else) is valid, but the game has changed since the days of Russell's Celtics so averaging over all champs would seem to flatten what's happening now.