APBRmetrics

[94by50]

When I was still updating the weekly bell curve standings, Mountain saw something that led me to research this subject. Specifically, he saw that five of the oldest teams in the league this season were all among the league's leaders in winning percentage over their bell curve expectation. To study this further, I took every team from the merger onward and figured their team age, weighted by minutes played. I then gathered the following data:

* Their actual wins
* Their expected wins based on the bell curve method (prorated to 82 games in the lockout year)
* For kicks, the team's home court advantage, as well as the league's HCA for that season

I'm just an amateur, so the ways in which I know to approach this question are limited, but hopefully some more knowledgeable than I can see this from other angles that I may miss.

First thing I see: there's an odd trend in the league home court advantage. I remember Bill Simmons making remarks about how home court "mattered" more in the past, and I think this is what he meant. The league HCA actually jumped from around 3.7 points per game in the early 80s, to a high of 5.84 in 1987-88. For comparison, over the last ten years or so, it's been somewhere in the 3s. I call it an odd trend because the overall pattern is slowly downward, but there's a sharp, temporary spike in the late 80s.

The main question: are older teams likely to outperform their actual winning percentage?

Test #1: I wouldn't expect a trend like this to be linear, but I ran a linear regression to start. Based on that, there's only a 6% connection (r^2 = .06) between team age, and how much the team exceeds its actual winning percentage.

x = team age, y = actual wins - expected wins
.4498x - 12.05 = y: r^2 = .0608

Adjustment: the average age of teams is slightly below 27 (26.957), so instead of raw team age, I'm subtracting 27. Second adjustment: I'm adding a second degree to the equation. I wonder if the effect peaks anywhere.

Test #2: a second-order regression, subtracting age from 27. Not a thing changes - 6% connection.

x = team age - 27, y = actual wins - expected wins
-.0009x^2 + .4552x + .1132 = y: r^2 = .0609

Just a cursory glance at the subject. It appears to me that age and/or experience has little if any measurable impact on a team's ability to outperform its expected winning percentage. What about this year, then? A fluke? I dunno. There's probably something else I should account for that I've not thought of yet.

I'm going to post the workbook on the Yahoo group, to make the dataset available. If anyone has any suggestions, they'd be much appreciated.

[Harold Almonte]

What about the correlation between team age and expected injuries?

[Mountain]

Thanks for pursuing the subject 94by50.

If you wanted to go further, given the discrepancy between a very long timeframe and this season, maybe it would be good to look at last 5-10 years by itself? If it matches up more with this season then maybe we found a meaningful trend in the current phase. If it matches the long-term results then I'd be more ready to drop the notion. Maybe the surge of new blood in the 70s and 80s made youth more of a win producer back then? The period since the merger is so long it might contain several trends within it that might offset.

Another angle might to focus on just starters, players likely to be in at the end of the game where the close games (the main source of the actual - expected gap) are decided.
On a quick check of recent good teams I see at least 3 older key guys but don't know the average situation immediately.

Or check the correlation just for playoff teams. Maybe age only correlates strongly with teams that work, with obviously other terms of fit mattering too.

[94by50]

[94by50]

[Mountain]

Year by year would be comprehensive detective work. I am personally mainly interested in the recent time period but a longer detailed study would give more context if you are so motivated. I see basketball-reference has close to 30 years of games started information.


Would be interesting to check inter-relationship between wins, age and pace by year.

70s and 80s saw a more wide open door and a faster game than the 90s or recently.

Pistons / Bulls slowed things down. making easier for the old guys and old guys style & tricks?

Recent years a bit faster right? Is the best win age moving back from above peak back topwards 27? Lakers Cavs and Magic are right there by your numbers. Is speend being rewarded more? http://sonicscentral.com/apbrmetrics/viewtopic.php?t=1962&start=45
3 of the 4 oldest teams slipped in wins and more so in contending status.
I think the Lakers aged up some from last season to get closer to the player and maybe team sweet spot.
Celtics not far off and a reminder that the age mix is worth looking at beyond the minutes weighted age.
How does 3 point shooting frequency and accuracy trend by age? Maybe that has been presented before? I am not going to look for it but maybe somebody else knows or knows where it is. Lots of factors. Does the best age for today vary by position? Maybe some.

The interplay of these things affect cap strategy, draft pick use (keep or trade), coaching, etc.

[94by50]