The following is a guest post from Kevin(@theDissNBA) from the great NBA blog The Diss, which has been posting angsty NBA analysis since 2011. Kevin kindly offered to give some amazing statistical analysis, which we love around here. Enjoy!
If you’re more worried about Baseball right now (I hear some kind of tournament is going on) then you can also head on over to Freakonomics where our own Dave Berri has a great post on if money can buy happiness in Major League Baseball.
As the NBA offseason dragged on and there was less material to write about, we saw the return of some of basketball’s most boring columns. As writers go on vacation or their creativity is sapped, they return to the same thing they wrote the year before. The most tired of these is the annual “San Antonio Spurs are too old” columns that are written, but almost as bad are the “[insert young team predicted to do well] are too young and inexperienced to win a title.” It seems to be accepted basketball wisdom that title-winning teams exist in a middle ground where the players are just reaching their peak, but not yet over the hill. Rarely, however, is it asked, is this actually true?
I have pulled data* from the last ten years for this analysis. A simple histogram of the average age of each team shows the distribution to be relatively normal, with a slight right-skew. Average team age is 26.73 years, with the 2006 Atlanta Hawks winning the distinction of youngest team in the sample at 22.7 years (the top minute-getter was a 21 year old Josh Smith, and the only player over 30 was Lorenzen Wright) and the 2008 San Antonio Spurs the oldest team in the sample at 31.4 years (35 year old Michael Finley, 37 year old Bruce Bowen and 36 year old Kurt Thomas all got major minutes).
A simple regression shows that average team age is a relatively poor predictor of winning percentage. Regressing age upon winning percent, Pythagorean winning percentage or Simple Rating System gives an r2 between 0.2 and 0.25 (Editor’s Note: This means age on its own is only able to explain around 20-25% of what wins games.) Considering that Roland Beech found (in only a five-year sample) that the previous year win percentage predicted 57% of wins, average age doesn’t particularly add to our ability to predict how well a team will do.
Perhaps, however, age can tell us other things. The average age of an NBA team in the sample was 26.73, while the average age of a playoff team was 27.37 and a non-playoff team 26. The average age of the ten Larry O’Brien Trophy winning teams (small sample size alert!) was 28.68. It seems that this gives some credence to the “team x is too young to win this year” argument, while giving San Antonio Spurs fans hope for the upcoming year. Indeed, the youngest championship winning teams in the sample were the 2004 Detroit Pistons and the 2009 LA Lakers at 27.4 years: no championship winning team has had a below league average age!
To understand the age distribution of good, okay, and bad teams, I separated teams into three tiers based upon their Simple Rating System, a metric that takes into account point differential and strength of schedule. I find this to a better metric at judging the relative strength of teams because of the extreme imbalance of playing in, say, the Southwest Division versus the Atlantic Division. The graph above shows us that while there are both good and bad old teams and good and bad young teams, it is fairly easy to predict how well a team on the extreme ends of the age scale will do. Of the 66 teams that had an average age of 28 or above, only 2 were in the bottom third of the league. The data on young teams isn’t as conclusive, as 3 of the 10 teams with an average age of 24 or below were in the top third of the league.
Now, to build a proper model to use age to predict winning percentage, you must first make a good estimate at minute distribution. The way this age data is calculated is by minutes played, so if a team has a lot of young players getting big minutes, and a lot of old players getting few minutes, the team’s average age will be low. If you can accurately model minutes, there are a few pretty solid claims that can be made before the season starts:
- If a team’s average age is 28 or above, it will make the playoffs (88%)
- If a team’s average age is below 25, it will not make the playoffs (79%)
- If, among the teams that make the playoffs, a team’s average age is in the bottom half, it will not win a championship (100%)
Now, if somebody could just build a model that accurately predicted minutes, including injuries, trades, and coaching changes…
*All data is taken from Basketball Reference and was used unmodified