Quantifying Longevity

[ijspeelman]

For the Top 100 Project, I and others have referenced era-adjusting longevity, but I had been personally doing it without actually knowing how many games/years the average player played in different eras.

I went ahead and found all data for those drafted by each year and used them as my sample to give average games per players and average years per player (with inclusions of three and five year bands).













I would like to start using this data specifically for referencing longevity.

Unformatted Data Here:

Spoiler:

Code: Select all

Year   GM/Player   YR/Player   3Y GM/Player   3Y YR/Player   5Y GM/Player   5Y YR/Player
1947   143.18   2.91            
1948   158.42   2.90   168.97   3.11      
1949   205.32   3.52   193.06   3.38   165.78   3.07
1950   215.45   3.73   175.76   3.18   160.29   2.95
1951   106.52   2.28   145.90   2.78   166.73   3.05
1952   115.73   2.33   137.63   2.67   186.80   3.35
1953   190.63   3.40   204.01   3.58   204.37   3.53
1954   305.67   5.00   266.53   4.35   238.75   4.01
1955   303.30   4.65   295.80   4.76   251.75   4.18
1956   278.43   4.64   254.16   4.17   289.31   4.67
1957   180.74   3.21   279.20   4.57   306.93   4.90
1958   378.42   5.85   317.65   5.06   308.37   4.94
1959   393.78   6.13   360.89   5.62   304.54   4.85
1960   310.46   4.88   321.19   5.07   337.29   5.26
1961   259.31   4.19   304.75   4.77   312.44   5.00
1962   344.47   5.25   285.98   4.66   319.65   5.07
1963   254.17   4.56   342.82   5.42   352.98   5.55
1964   429.83   6.47   387.04   6.10   376.16   5.90
1965   477.13   7.28   427.39   6.56   354.87   5.69
1966   375.21   5.94   363.45   5.80   355.91   5.65
1967   238.00   4.19   290.86   4.83   330.01   5.34
1968   259.36   4.35   265.90   4.49   298.00   4.87
1969   300.34   4.93   292.26   4.74   285.85   4.64
1970   317.09   4.93   310.63   4.89   293.67   4.70
1971   314.46   4.81   302.88   4.75   299.64   4.80
1972   277.08   4.50   293.60   4.71   302.54   4.83
1973   289.25   4.83   293.72   4.80   302.65   4.89
1974   314.82   5.08   307.23   5.05   302.63   4.95
1975   317.62   5.25   315.61   5.15   321.46   5.26
1976   314.38   5.11   334.42   5.46   326.87   5.32
1977   371.25   6.03   333.96   5.43   337.59   5.51
1978   316.26   5.15   351.98   5.73   337.96   5.55
1979   368.44   6.00   334.73   5.53   355.49   5.86
1980   319.49   5.44   363.32   6.03   350.73   5.87
1981   402.03   6.66   356.32   6.06   367.04   6.15
1982   347.43   6.09   382.43   6.43   371.39   6.23
1983   397.83   6.53   378.48   6.36   394.76   6.57
1984   390.19   6.45   408.11   6.69   378.39   6.37
1985   436.30   7.09   382.22   6.42   384.67   6.45
1986   320.18   5.71   378.45   6.42   387.79   6.58
1987   378.87   6.47   370.82   6.46   392.79   6.72
1988   413.41   7.19   402.50   6.93   388.60   6.85
1989   415.21   7.13   414.66   7.35   404.67   7.14
1990   415.35   7.75   410.36   7.34   418.43   7.42
1991   400.52   7.16   421.17   7.59   418.74   7.45
1992   447.65   7.85   421.04   7.45   419.18   7.56
1993   414.95   7.35   426.68   7.63   422.01   7.51
1994   417.44   7.69   420.62   7.51   440.42   7.80
1995   429.46   7.48   446.50   7.94   431.91   7.70
1996   492.60   8.64   442.38   7.81   437.37   7.84
1997   405.09   7.32   446.64   8.02   444.57   7.88
1998   442.25   8.09   433.59   7.77   432.91   7.85
1999   453.43   7.89   422.29   7.76   427.50   7.77
2000   371.18   7.30   430.06   7.81   422.59   7.66
2001   465.55   8.24   405.76   7.44   444.83   7.89
2002   380.54   6.77   466.51   8.09   445.04   7.92
2003   553.43   9.26   462.82   8.02   460.51   8.02
2004   454.50   8.04   485.48   8.35   435.69   7.64
2005   448.51   7.76   414.83   7.39   442.14   7.73
2006   341.48   6.37   400.93   7.11   427.45   7.53
2007   412.80   7.20   411.41   7.28   427.79   7.50
2008   479.96   8.27   449.65   7.79   400.17   7.15
2009   456.18   7.88   415.52   7.39   409.24   7.25
2010   310.41   6.02   384.48   6.92   390.33   6.97
2011   386.83   6.87   338.51   6.24   361.71   6.56
2012   318.29   5.82   347.33   6.30   328.61   6.10
2013   336.86   6.20   315.27   5.87   332.06   6.11
2014   290.66   5.60   318.39   5.96   296.21   5.61
2015   327.64   6.07   275.30   5.34   274.76   5.34
2016   207.60   4.35   248.76   4.97   247.74   4.91
2017   211.05   4.48   206.80   4.29   219.53   4.47
2018   201.75   4.04   187.48   3.98   178.55   3.78
2019   149.62   3.41   158.04   3.36   154.91   3.29
2020   122.74   2.64   120.58   2.65   122.03   2.60
2021   89.36   1.89   86.25   1.84      
2022   46.66   1.00            


[SHAQ32]

But the 2010s/20s are so much more competitive and advanced than past eras

[Owly]

But the 2010s/20s are so much more competitive and advanced than past eras

I haven't properly looked at or understood the data but isn't the downward trend likely to be caused by including active players with incomplete careers?

[ijspeelman]

But the 2010s/20s are so much more competitive and advanced than past eras

I haven't properly looked at or understood the data but isn't the downward trend likely to be caused by including active players with incomplete careers?

Correct, any year within ~15 years ago doesn't have complete data.

[DraymondGold]

For the Top 100 Project, I and others have referenced era-adjusting longevity, but I had been personally doing it without actually knowing how many games/years the average player played in different eras.

I went ahead and found all data for those drafted by each year and used them as my sample to give average games per players and average years per player (with inclusions of three and five year bands).
...

Really great stuff ijspeelman! Quantifying this sort of thing is super crucial if we are to give an honest era-relative longevity adjustment

How difficult would it be to add a filter for larger value player, e.g. just looking at players who were an all-star at least once, or players who were either all star or all-defense at least once, or players who played 25 minutes per game at least once, or something like that?

I'm worried some of this change is dominated by a change in role players, not stars.

[ijspeelman]

For the Top 100 Project, I and others have referenced era-adjusting longevity, but I had been personally doing it without actually knowing how many games/years the average player played in different eras.

I went ahead and found all data for those drafted by each year and used them as my sample to give average games per players and average years per player (with inclusions of three and five year bands).
...

Really great stuff ijspeelman! Quantifying this sort of thing is super crucial if we are to give an honest era-relative longevity adjustment

How difficult would it be to add a filter for larger value player, e.g. just looking at players who were an all-star at least once, or players who were either all star or all-defense at least once, or players who played 25 minutes per game at least once, or something like that?

I'm worried some of this change is dominated by a change in role players, not stars.

I saw your post in the other thread about this and thinking of a good way to do it.

I like the idea of only including all-stars, but I worry that the sample may be too small year-to-year so maybe just all players who played an average of 15-20 minutes per game for their career?

I can try the all-star method as well and will get back here with it.

Though, I am personally not worried about the difference between longevity of all drafted players versus longevity of all all-stars as I think the data, while different, will provide similar results when it comes to era-adjustment.

[DraymondGold]

For the Top 100 Project, I and others have referenced era-adjusting longevity, but I had been personally doing it without actually knowing how many games/years the average player played in different eras.

I went ahead and found all data for those drafted by each year and used them as my sample to give average games per players and average years per player (with inclusions of three and five year bands).
...

Really great stuff ijspeelman! Quantifying this sort of thing is super crucial if we are to give an honest era-relative longevity adjustment

How difficult would it be to add a filter for larger value player, e.g. just looking at players who were an all-star at least once, or players who were either all star or all-defense at least once, or players who played 25 minutes per game at least once, or something like that?

I'm worried some of this change is dominated by a change in role players, not stars.

I saw your post in the other thread about this and thinking of a good way to do it.

I like the idea of only including all-stars, but I worry that the sample may be too small year-to-year so maybe just all players who played an average of 15-20 minutes per game for their career?

I can try the all-star method as well and will get back here with it.

Oh sure, a minimum career mpg filter would also work! And agreed, I was also worried about the small sample of all stars -- I was just trying to brainstorm a better filter than *literally every player ever* lol.

No worries if this ends up being too much work, but I would definitely be very interested in the results if you get the chance to do this!

Edit: up until about 2005, it looks like you could get a pretty great fit by just assuming it's linear (plus some noise). It looks like the biggest deviation from linear is the bump around 1960. There's some secondary dip in the late 80s, and you get some pretty big oscillations in the mid/late 2000s, though this second one might be that we're starting to get to recent players. I wonder if there's contextual factors to explain the first two....

Out of curiosity, does this include the ABA or is it just NBA?

[eminence]

I want to say f4p had a thread pretty similar to this in the last few months, might save some work if you liked his methods well enough.

[ijspeelman]

Really great stuff ijspeelman! Quantifying this sort of thing is super crucial if we are to give an honest era-relative longevity adjustment

How difficult would it be to add a filter for larger value player, e.g. just looking at players who were an all-star at least once, or players who were either all star or all-defense at least once, or players who played 25 minutes per game at least once, or something like that?

I'm worried some of this change is dominated by a change in role players, not stars.

I saw your post in the other thread about this and thinking of a good way to do it.

I like the idea of only including all-stars, but I worry that the sample may be too small year-to-year so maybe just all players who played an average of 15-20 minutes per game for their career?

I can try the all-star method as well and will get back here with it.

Oh sure, a minimum career mpg filter would also work! And agreed, I was also worried about the small sample of all stars -- I was just trying to brainstorm a better filter than *literally every player ever* lol.

No worries if this ends up being too much work, but I would definitely be very interested in the results if you get the chance to do this!

Edit: up until about 2005, it looks like you could get a pretty great fit by just assuming it's linear (plus some noise). It looks like the biggest deviation from linear is the bump around 1960. There's some secondary dip in the late 80s, and you get some pretty big oscillations in the mid/late 2000s, though this second one might be that we're starting to get to recent players. I wonder if there's contextual factors to explain the first two....

Out of curiosity, does this include the ABA or is it just NBA?

This includes just the NBA as I just took each NBA draft since 1947 (I don't believe the ABA had drafts?).

Here is the same data with career 15MPG filter.

[WestGOAT]

I did something very similar as well recently:
viewtopic.php?f=64&t=2303913&hilit=Longevity

I think only difference being that I included ABA minutes. I think we pretty much have similar results.