You Are Here > Basketball-Reference.com > BBR Blog > NBA and College Basketball Analysis

SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all Basketball-Reference content, so you can quickly and easily find the content you want.

Also, our existing Basketball-Reference blog rss feed will be redirected to the new site's feed.

Basketball-Reference.com // Sports Reference

For more from Neil, check out his new work at BasketballProspectus.com.

Championship Usage Patterns and “The Secret”

Posted by Neil Paine on May 18, 2010

In basketball perhaps more than any other sport, the concept of team-building -- creating a cohesive group that fits together and may be greater than the sum of its parts -- is phenomenally important. In baseball, a sport dominated by one-on-one matchups, not a whole lot of consideration has to be made for how teammates work together; to make a great team, you basically grab the 25 best players you can, throw them together, and watch them produce. But in basketball, teammates have to work together while simultaneously "competing" for touches & shots. Throw together a baseball lineup of 9 guys who each create 100 runs, you'll probably score 900 runs; throw together a basketball lineup of 5 20 PPG scorers, you probably won't score 100 PPG. There's no upper limit on the number of runs the baseball lineup can produce, but there is an upper limit to the points the basketball lineup scores, because teams are limited by a finite number of minutes in a game, and as a result, lineups are limited by a finite number of touches & shots to be allocated to the individual players.

That's why a stat like Possession% (the % of team possessions a player uses while on the floor) is important in looking at how the pieces of a team fit together. A lineup of All-Stars would be interesting, but perhaps a less-talented lineup with one 26% usage guy, two 20% guys, an 18% guy, and a 16% guy would be even better if the All-Stars are not happy with the way they fit together or are unable to operate at peak efficiency in lesser roles, while the less talented lineup features players who are all at their optimal usage levels. The whole of the latter would be greater than the sum of the former's parts.

I thought of this when I was cruising Bill Simmons' archive and came across this old piece on the 2004 Olympic Team. Now, let's be honest, in The Book of Basketball Simmons kinda made too much of "The Secret of Basketball" -- look, it really wasn't ever a secret, and the only "stats" that end up getting sacrificed by embracing "the secret" are ones like PPG that APBRmetricians don't care about anyway. In fact, the Poss% framework is a nice way of quantifying "the secret": we can actually measure the sacrifice made by players on winning teams vs. losing ones (witness Ray Allen dropping from 28% possession usage on Seattle in 2007 to 21% for Boston in '08, and raising his ORtg from 112 to 116 as a result).

But just because "the secret of basketball is that it's not about basketball" is a really stupid sentence to build a book around, it doesn't mean that Simmons and Isiah aren't right. In that Olympic article, Simmons writes this about how a great U.S. team should be formed:

"The Summer Olympics [is] a blank canvas... A chance to build a superior basketball team from scratch -- not an All-Star team, a basketball team. Choosing from 300 of the greatest players in the world, we would want one dominant big man; one quality point guard; one great scorer immediately designated for Alpha Dog Status, two other good shooters, two other rebounders, one athletic swingman who can defend the other player's best shooter, a backup point guard, two energy guys, and a 12th man who will hustle in practice and just be happy to be on the team. If we pick the right guys, we know we're winning the tournament and possibly ending up on ESPN Classic. It's just a fact."

That's a pretty nice formula, not just for Team USA, but for successful NBA playoff teams as well. And it's also one that fits in with the idea of the 26%-20%-20%-18%-16% mix (or whatever is optimal) I mentioned above. Instead of grabbing all 30% usage guys, you deliberately take players who aren't necessarily as talented, but will perform with better efficiency when they are asked to play that 18% role.

But one question that pertains to the NBA playoffs is this: what exactly is the optimal combination? Is it the percentages I listed above? Or should the Alpha Dog take away more possessions from the mid-usage guys? Or maybe our role players are taking too big a % of the possessions? To find the answer, I looked at the postseason Modified Shot Attempt %s (same as poss%, but without turnovers) for the top 7 playoff minute-earners on every NBA champion since 1952:

Modified shot att% by team rank in MP
Year Team #1 MP #2 MP #3 MP #4 MP #5 MP #6 MP #7 MP
1952 Minneapolis Lakers 31.6 15.7 16.3 18.1 22.9 18.8 13.4
1953 Minneapolis Lakers 28.4 23.1 15.6 21.1 15.6 18.9 17.0
1954 Minneapolis Lakers 19.3 17.1 29.3 16.2 18.2 28.5 16.8
1955 Syracuse Nationals 20.7 18.5 15.0 25.8 19.8 18.2 18.5
1956 Philadelphia Warriors 26.6 16.4 23.0 17.9 22.7 14.8 13.3
1957 Boston Celtics 24.2 16.6 22.5 25.8 17.2 17.9 18.3
1958 St. Louis Hawks 25.7 25.0 17.2 14.3 18.0 20.1 15.9
1959 Boston Celtics 13.6 23.2 24.0 23.1 24.7 16.1 22.6
1960 Boston Celtics 15.5 26.0 17.9 25.8 22.9 12.8 14.7
1961 Boston Celtics 17.3 24.2 20.7 26.8 20.0 18.2 13.7
1962 Boston Celtics 17.5 22.2 25.4 26.8 12.0 16.5 15.4
1963 Boston Celtics 16.8 23.1 27.0 26.0 12.9 13.8 19.9
1964 Boston Celtics 15.6 22.7 14.5 25.7 14.0 25.3 24.4
1965 Boston Celtics 13.8 24.2 25.3 14.5 17.3 26.4 20.5
1966 Boston Celtics 16.8 24.4 27.1 12.3 17.7 20.9 17.8
1967 Philadelphia 76ers 18.0 24.1 20.2 13.1 22.3 29.5 11.8
1968 Boston Celtics 15.8 24.5 24.1 20.5 19.6 18.2 14.9
1969 Boston Celtics 23.4 13.2 16.3 19.1 27.0 23.4 23.2
1970 New York Knickerbockers 17.6 24.0 19.2 20.6 18.3 23.1 18.5
1971 Milwaukee Bucks 25.5 22.8 24.7 16.6 14.9 16.2 15.7
1972 Los Angeles Lakers 12.5 18.7 30.5 14.8 25.3 18.9 14.1
1973 New York Knickerbockers 22.0 21.5 19.8 22.0 20.6 16.8 19.1
1974 Boston Celtics 26.0 23.1 20.8 12.9 14.1 18.1 21.1
1975 Golden State Warriors 30.8 21.6 23.8 9.7 18.9 10.7 20.3
1976 Boston Celtics 22.5 25.0 11.9 24.1 19.2 20.0 18.4
1977 Portland Trail Blazers 20.9 22.8 25.2 17.6 17.7 16.3 20.7
1978 Washington Bullets 21.3 23.2 11.5 18.1 23.8 19.3 25.7
1979 Seattle Supersonics 24.1 17.6 32.0 17.3 17.5 8.8 27.1
1980 Los Angeles Lakers 20.4 21.2 20.7 26.5 16.7 14.0 15.4
1981 Boston Celtics 23.7 21.8 17.1 15.7 22.4 19.8 18.3
1982 Los Angeles Lakers 19.3 24.8 20.8 23.1 21.1 17.7 11.6
1983 Philadelphia 76ers 24.6 23.6 21.9 30.1 13.9 14.1 10.6
1984 Boston Celtics 25.9 16.9 22.5 14.5 18.6 22.0 16.6
1985 Los Angeles Lakers 25.1 20.2 24.1 20.4 15.4 21.6 10.3
1986 Boston Celtics 25.1 21.8 20.4 17.2 19.9 16.5 13.3
1987 Los Angeles Lakers 21.9 27.3 18.2 22.5 18.4 14.1 16.9
1988 Los Angeles Lakers 26.2 21.4 24.8 12.8 22.0 16.1 13.8
1989 Detroit Pistons 29.1 24.5 15.7 20.8 10.4 14.5 30.8
1990 Detroit Pistons 27.3 23.4 15.4 9.5 14.9 23.7 22.8
1991 Chicago Bulls 24.9 34.6 13.7 14.8 14.1 17.7 18.8
1992 Chicago Bulls 36.4 24.5 13.4 11.3 13.6 19.3 12.3
1993 Chicago Bulls 25.5 38.8 13.8 14.7 13.5 13.1 11.1
1994 Houston Rockets 29.7 21.2 13.5 17.8 17.8 24.6 18.6
1995 Houston Rockets 33.7 24.0 15.7 17.7 13.3 24.6 13.6
1996 Chicago Bulls 24.0 33.4 11.1 16.6 21.2 16.8 15.2
1997 Chicago Bulls 35.7 24.3 9.6 15.6 14.2 20.6 13.0
1998 Chicago Bulls 36.7 24.2 9.8 20.8 14.1 16.5 12.2
1999 San Antonio Spurs 25.4 21.3 22.0 18.1 15.9 21.7 13.8
2000 Los Angeles Lakers 30.0 26.7 17.6 17.1 15.0 11.4 16.8
2001 Los Angeles Lakers 30.9 30.0 15.4 13.6 12.9 13.6 15.6
2002 Los Angeles Lakers 30.4 29.9 12.3 13.9 16.2 16.8 11.9
2003 San Antonio Spurs 26.7 24.8 19.1 11.9 18.6 20.0 14.2
2004 Detroit Pistons 27.4 15.0 23.9 20.3 16.4 22.9 15.2
2005 San Antonio Spurs 31.0 26.1 9.5 26.7 16.3 12.1 13.7
2006 Miami Heat 32.8 19.9 25.5 18.9 14.3 11.8 13.3
2007 San Antonio Spurs 29.7 28.6 9.7 28.0 18.9 10.9 11.4
2008 Boston Celtics 25.4 25.9 19.3 22.9 10.4 12.9 10.5
2009 Los Angeles Lakers 34.3 18.3 16.7 15.7 15.2 17.8 15.3
Average 24.5 23.2 19.2 18.9 17.6 18.0 16.5

So for a starting 5, it looks like 24-22-19-18-17 is the optimal championship mix, with a sixth man who can create and a pure role player in the #7 slot. Here are the remaining playoff teams and their patterns, along with the historical probability that a team with that specific usage pattern wins the championship:

Team #1 MP #2 MP #3 MP #4 MP #5 MP #6 MP #7 MP p(C)
Phoenix Suns 24.2 26.6 23.8 15.8 14.4 14.9 24.3 9.9%
Los Angeles Lakers 33.1 21.4 16.8 14.3 16.5 14.9 20.7 9.3%
Orlando Magic 17.4 25.0 28.1 21.3 15.7 18.8 9.8 8.0%
Boston Celtics 24.6 20.7 23.2 22.3 11.1 18.5 17.0 7.9%

Strangely, though, the 2010 playoff team this "Secret" formula would have predicted to win the most often through chemistry? The Bobcats. Proving that sometimes it really is about basketball... Basketball talent, that is. You can have as much on-court chemistry as you like, but if the other team is significantly better, it's probably not going to matter.

ShareThis

9 Responses to “Championship Usage Patterns and “The Secret””

  1. DSMok1 Says:

    VERY intriguing work, Neil. I look forward to more in this vein.

  2. Jason J Says:

    "Strangely, though, the 2010 playoff team this "Secret" formula would have predicted to win the most often through chemistry? The Bobcats. Proving that sometimes it really is about basketball... Basketball talent, that is."

    I think this is another indicator of what a great coach Larry Brown is. I'm sure he doesn't "know" what usage percentage breakdown his players should optimally have, but he seems to have a great sense of how his team should best operate on offense - be it a one man show in Phili or a perfect team split in Detroit or Charlotte.

  3. Neil Paine Says:

    Good point, and that made me think of another baseball example -- Earl Weaver. Weaver didn't know or care anything about sabermetrics (it didn't really even exist for most of his managerial career), but he somehow intuited a strategic model that was really similar to the model that sabermetrics eventually came up with. It seems like that's the mark of a great coach, being able to intuitively make decisions that are later confirmed by quantitative analysis.

  4. Mike G Says:

    Wait a minute.
    You've just averaged in 9 Bill Russell Usg% with 6 Michael Jordan Usg% (plus everyone in between) and conclude the ideal Usg% for your #1 minutes player is somewhere in between?

    And so, Ideally, you'd advocate getting rid of a Jordan or a Russell, in favor of guys who shoot less (or more) because that's more 'ideal'?

    No, Jordan is the guy to keep; compliment his 36% with a 10% Rodman, etc.
    Around Russell's 15%, a bunch of 25's.

  5. Luke Says:

    Well, since each of these teams won, I find it hard to say that any of these combinations don't/won't work. I think you'd have to look at what the makeup of the teams that they beat was (and by how badly) to see which combination works "best."

  6. izzy Says:

    I don't think it's fair to say that by averaging the results of different teams with different contexts creates the ideal usage breakdown. And even if it does, subscribing to the usage breakdown as a team (i.e. The Bobcats) does not mean that you have chemistry. Chemistry goes beyond the usage breakdown--chemistry is knowing exactly where each teammate is on the floor so you can kick it to the corner for an easy three or hit your man for that telepathic alleyoop. My gut tells me there are flaws in this article, but I that's just my gut--I'll save it for others to help articulate my intuitions (which is unfair, but hey, I'm tryin)

  7. Brian Says:

    There might be a bit of an issue with sorting things by minutes played. This confounds players who get minutes due to their role in the offense vs. other reasons, e.g. defense and rebounding. Notice guys like Bruce Bowen, Dennis Rodman etc. getting a lot of minutes despite a small role in the offense. What might make more intuitive sense is to take the top 5 minute-getters on each championship team, then sort those by usage and take that average. The optimal combination, figured this way, is going to be more skewed towards the "alpha dog" formula.

  8. Neil Paine Says:

    Good call, Brian. Here's the sequel to this post, which sets up the study in those terms:

    http://www.basketball-reference.com/blog/?p=6013

  9. Joel Says:

    Neil,

    I think this just demonstrates that the rankings of modified shot percentages have a very weak correlation with a team's ability to win games.