Player Aging Part 1

This will be the first post of a yet to be determined number of posts as I investigate aging curves for various stats.  Gabe over at Behind the Net has already done quite a bit of work on this topic, so some of this will be a rehashing of things he’s already investigated.  Our methodology differs slightly (honestly his is likely better) so the results, while similar, won’t be the same.  For the first pass I chose to use the delta method with no correction for survival bias.  What does this mean?  I took all players that played both an age 18 and 19 season and calculated their increase/decrease in the various stats.  I then calculated the same for all age pairs.  Once that’s done I found a weighted average of the increases/decreases for each age pair where the weight is equal to the minimum number of games played between the two ages.

The first stat I chose to look at is pts/game.  My input data is 2001-2010.  The following table summarizes the results by general position

Age	All	F	D
18_19	0.140	0.160	-0.108
19_20	0.068	0.072	0.052
20_21	0.026	0.026	0.023
21_22	0.049	0.050	0.045
22_23	0.023	0.025	0.018
23_24	0.010	0.022	-0.018
24_25	0.002	-0.009	0.024
25_26	0.020	0.027	0.007
26_27	-0.022	-0.021	-0.023
27_28	-0.027	-0.039	-0.003
28_29	-0.024	-0.022	-0.028
29_30	-0.012	-0.030	0.017
30_31	-0.038	-0.044	-0.028
31_32	-0.052	-0.069	-0.022
32_33	-0.045	-0.057	-0.021
33_34	-0.057	-0.062	-0.048
34_35	-0.067	-0.079	-0.046
35_36	-0.080	-0.083	-0.071
36_37	-0.058	-0.074	-0.022
37_38	-0.043	-0.042	-0.044
38_39	-0.089	-0.128	-0.002
39_40	-0.070	-0.073	-0.063

The table shows the age pair in the first column and then the respective increase/decrease in the other three columns.  These numbers seem to indicate a peak around 26.  The following graph shows the same data in graphical form

I also ran a similar analysis on average time on ice (ATOI).  Here’s the table

	ATOI
18_19	2.62
19_20	0.93
20_21	0.90
21_22	1.18
22_23	0.82
23_24	0.59
24_25	0.57
25_26	0.32
26_27	0.03
27_28	0.05
28_29	0.03
29_30	-0.10
30_31	-0.30
31_32	-0.54
32_33	-0.48
33_34	-0.78
34_35	-0.70
35_36	-1.13
36_37	-0.97
37_38	-0.83
38_39	-0.96
39_40	-1.16

and the graph

Here the positional difference was smaller (although I may graph it in a late post). Overall there appears to be a fairly flat peak between 26-30.

Next up I’d like to investigate aging of some of the more advanced stats (i.e. GVT, relative +/-, etc.)