Age is just a variable

Age, with his stealing steps,
Hath clawed me in his clutch.

I’ve spent more than a month working on this. It’s vexed me and frustrated me by being there, tantilizing me, just out of reach. It’s kept me from writing other interesting pieces ( Usage, Smallball, An updated free agent guide to name a few).

I’m finally done.

Was it worth it?

Don’t worry, I’m going to explain that but you’re going to have to have some patience with me while I build my narrative.

Bill Veeck was somewhat of a flim-flam artist. As the owner of four major league baseball teams (yes four) over the course of his life he became famous for being a maverick, always bucking the norm or doing something crazy to try to gain every little edge.

Scene from an actual Major League game

From putting Eddie Gaedel in a major league uni, to attempting to break the Major League’s Color Line with the Phillies in 1943 and stock the team with Negro League All Stars (a plan which the National League threw it’s body in front off at the time) to coming up with Harry Caray’s Take me out to the Ball Game routine he truly put his mark on the game and earned his spot in the Hall of Fame.

His greatest and most important contribution to the business of sport is not as well know and it really might the greatest con ever pulled in sports. You see in 1946 Bill Veeck convinced the IRS that the roster of players on his newly acquired Cleveland Indians was a depreciable asset. What this means in practical terms is that he was able to convince the government that, like cattle, the value of the players on his roster went down over time because of age and that he should be allowed to claim this loss of asset value as a loss on his tax return.

This is the famous Roster Depreciation Allowance (or RDA- see here for full detail). To put it very simply, the RDA allows you to claim the value of your franchise as a loss in your books over a period of 15 years and in essence save 35% of that amount on your tax returns (this is know as the 15/100 Rule of thumb see here for more detail). You can claim that loss in whatever schedule you like. You want to claim 90% in Year 5? Go right ahead. No loss claim in Year 11? Good for you.

This of course is a massive corporate tax shelter and one of the real benefits of owning a sports team. We are not here to argue about that though. The point is, that based on that month of work I just put in, at least for the NBA, Veeck was actually right.

You see I was working on an Age model for productivity (the title and table should have given it away but just in case let’s get it out there before I’m done with the exposition). I’ve been working on projections, value and free agent models. I’ve got most of it done and published but the final piece before the full model is ready is the to figure out how player performance is tied to age.

Cue the age model.

As I said, a month of work went into this so naturally there is a metric ton of data as well as lot of fascinating new pieces of information to share. I’m going to try to give you the insider’s tour. Let’s start from the top. (If you’re new, look here for detail on the model and math used)

The data set chosen really was the most important of all the choices I had to make. When I did this previously, I used:

  • All player drafted from 1977 that played a minimum of ten seasons at greater than >400 MP.
  • I based the analysis on ADJP48 (i.e. non-position adjusted WP48 see here for an explanation)

And I tested that and multiple other scenarios. At the end of the day my best results were achieved by:

  • Focusing on the last decade i.e. 2002-03 thru 2011-12. This takes out the variation I saw over time due to, I suspect, better medical technology and different draft pools.
  • Using ADJP48 (Raw Productivity of players) but focusing on the difference from the current to the previous year or the previous three years. Three years and one year tested higher than other combinations. Three years takes out some of the injury noise I suspect.
  • Using only data from players playing >750 minutes both in the season tested and the previous year. Sample size is important. Too small a sample warps the data.
  • Focusing only on players aged 20 to 36. Players that are older or younger present too small of a sample and tend to warp the data slightly, even though they do hold to the observed pattern for the data.

The data look like so graphically:

The result is a mostly linear model with player improving to about age 25 generally reaching their peaks at around age 26 to 27 and remaining close to it to about age 30 to 31 and then proceeding to fall off with increasing speed after that.

Does that make sense? The empirical data bears it out.

That’s the Age chart by team for the last decade. Age by team is calculated by taking a minute weighted average of the players on each teams roster. The result is that the average age for the league is right at the expected peak age of 27 (actually 27.1 for the last decade). An interesting finding from the data set is that older teams tend to win more and in fact for the final four in the playoffs the average age is 28 and for the Champion it’s 29 years old.

So empirically it makes sense, how about practically?

When I model out every year from 78 to 2012 to now I get a very good level of correlation (around 80% for the three year model for the last five years. This looks like so graphically:

There is some significant year to year variation (particularly when players change teams) which leads me to think there’s some additional variable out there to find. It’s good enough for my purposes though. Editor’s Note: I made sure to bug Arturo about this and he’s clear that he’s on the hook for a post on the impact of changing teams.

This brings us full circle.

That’s the predicted age model based on the average productivity for the previous three years. As I said, Player are improving thru age 27. For a typical four year contract you can expect marked improvement for players signing at under 22, slight improvement thru 25, steady performance thru 29 and rapidly declining value after. What does it mean? When signing players, teams should not be willing to backload contracts for older players. It’s a bit of a tradeoff though. Older players help you win but they also , getting back to the initial analogy, age like cattle and not cheese or wine.

A fun illustration is what would happen if I could put together a team of 18 year olds that were good enough to win twenty games.

Assuming you could keep that team together (are you listening OKC?), you would be contending in 4 years and have an eight year championship contender window. This seems to be very much what OKC did with their roster. Again, the risk is that keeping such a roster together could get very expensive. Editor’s Note: Not quite as expensive though, as keeping together a team of 30+ year olds (are you listening Lakers?)




Comments are closed.