Layups: Weighting Team Age by Win Shares
Posted by Neil Paine on January 13, 2011
As a follow-up to their post about minute-weighted team age, Hoopism took the advice of our commenters and re-ran team ages, this time weighted by Win Shares:
Mapping Average Age to Success in the NBA
Comparing side-by-side with the raw roster averages, this has the effect of allowing you to see which teams' most productive players skew especially young (Miami, Orlando, LA Clippers) or old (Phoenix, Boston, Houston).
January 13th, 2011 at 1:47 pm
Great stuff. Is it fair, though, to rate players equally just on age, though? Shouldn't we consider the number of career games played or career minutes rather than just the age? Some players played 4 years of college while others played none. Others had significant injuries that kept them out for a season or more. Consider Wade and Lebron. Several years apart in age but entered the league at the same time. In fact Lebron has more wear on the tires due to injuries for Wade.
I also wonder if you can project the length of player's careers based on position, injury history, minutes played, etc. Then you could use the percentage into their career instead of age. I think this could be interesting.
January 13th, 2011 at 1:47 pm
This is great - however, I think meeting in the middle would be a better representation of 'team age'.
That is, weight their ages by UsageMinutes (i.e. Poss% times TotalMin%). This is a better representation of each player's involvement in the game. Weighing by Win Shares will marginally ignore inefficient players. It might also be beneficial to average Poss% with the 20% estimate for each player's defensive impact.
This, for example, makes the Celtics 30.1 years old.
January 13th, 2011 at 2:05 pm
I agree with Alex Z, unfair to rate just on age alone. There are many other factors to take into consideration.
January 13th, 2011 at 4:54 pm
This gives a good look at the age of the good players on the team. In other words, whether the team will likely continue to regress or to improve.
January 13th, 2011 at 6:20 pm
For what it's worth, I think weighting by winshares is exactly the right way to look at aging teams. I agree that usageminutes would also be an interesting weight, and it would be interesting to look at zero-order correlations between unweighted average age, age weighted by minutes, age weighted by usageminutes, and age weighted by by winshares.
But, what I think would be most interesting is to look at the historical evidence on this. So, take team age weighted by winshares in some baseline year (say, 2005) and look at how it predicts winning % (relative to 2005) in years 1, 2, 3, 4, and 5. Obviously you have all kinds of exogenous factors and might want to control for subsequent-year injuries, but it would be an interesting way to get a general feel for the conventional wisdom about how old teams are doomed and young teams are rising.
One interesting thing implication would be how much teams "adapt" to aging by making trades, selectively draft picks to fill in older player positions, etc.
January 14th, 2011 at 12:52 pm
I suppose it depends on what we are trying to measure here, but if we are simply looking at the age of the team, I don't think looking at the positive parts of a team (Win Shares) tells us the age distribution in a team. I guess I think that usage minutes would be the most telling because it shows the level of involvement per player age.
For example: If Shaquille O'Neal used 40% of the Celtics' offensive & defensive possessions on 70% of their minutes, with poor efficiency, the Celtics would not be considered extremely old on Win Shares. But the plurality of their game performance would be from the oldest player in the NBA.
Maybe I'm trying to measure something different here?
January 14th, 2011 at 3:03 pm
Using career games to determine who has more miles on their career is not completely appropriate. Players who entered the league earlier are not necessarily at a disadvantage.
Lebron hit his prime around age 24 like Jordan. Kobe's early prime also started at 24 and his best season was at age 27, his tenth season. Wade went to College and his career path peaked at a similar interval as well. Going by career games this wouldn't make nearly as much sense.