SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.

Also, our existing PFR blog rss feed will be redirected to the new site's feed.

Pro-Football-Reference.com ยป Sports Reference

For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.

Talent dispersal

Posted by Jason Lisk on August 24, 2010

Last month, Doug wondered what the NFL equivalent of the Miami Heat would look like, and there were several great comments about how that team might perform. Today, I'm going to take a look at teams judged similar in career AV of the starters, but which had different distributions of career value. To do this, I looked at team average career AV of the 22 primary starters, and the standard deviation for the career AV of the 22 primary starters. I'll refer to these as "talent balanced" teams, and "talent dispersed" teams. Now, there is a correlation between overall team quality and standard deviation. Good teams tend to have superstars, and those superstars have career AV's that cause the standard deviation, or talent dispersal, to rise. Bad teams tend to lack star players, and have lower standard deviations because, well, very few players turn out to be very good.

To combat that, I paired teams within groups of similar average career AV. I took all teams who played a 16 game regular season from 1978 to 2002, and grouped them, so that, for example, all teams that had an average starter AV of between 44.0 and 45.9 were lumped together. Then, within each group, I found pairs that were separated by at least 12.0 difference in standard deviation. I would pair the highest standard deviation within a group with the lowest, 2nd highest with 2nd lowest, etc., until I could pair no more groups separated by at least a standard deviation of 12.0.

The result was 54 pairs of teams, which I present below:

year team career AV st dev wins srs year team career AV st dev wins srs
2002 cle 36.0 13.9 9 1.2 1978 sfo 34.4 28.5 2 -9.1
1980 tam 36.4 16.4 5.5 -5.1 1998 crd 36.6 30.1 9 -7.4
1990 gnb 36.7 15.6 6 -4.1 1986 phi 37.9 33.9 5.5 -1.5
1981 tam 36.7 17.1 9 3.0 1996 sdg 37.0 29.2 8 -1.8
1999 det 37.4 16.6 8 1.9 1989 atl 38.0 29.3 3 -9.1
1994 cin 38.1 12.9 3 -6.9 1984 oti 39.2 33.3 3 -12.0
2002 nor 38.6 16.5 9 2.4 1997 sdg 38.3 28.9 4 -8.9
1995 chi 39.7 16.4 9 2.3 1986 nor 39.7 29.0 7 0.0
1989 gnb 40.5 17.4 10 -1.0 1983 nyg 40.6 30.9 3.5 -2.1
1998 chi 40.6 15.7 4 -3.7 1985 oti 40.8 33.9 5 -8.9
1994 chi 41.1 16.3 9 -1.9 1981 nyg 41.3 33.7 9 3.4
1991 clt 41.2 16.7 1 -17.3 2001 sdg 41.1 32.0 5 -0.3
1994 ram 41.7 18.0 4 -5.3 1994 crd 40.8 30.1 8 -2.3
1981 kan 41.9 16.7 9 3.2 1979 sea 41.7 31.2 9 2.6
1986 crd 43.1 17.4 4.5 -8.1 1985 phi 42.9 31.4 7 -3.4
1981 atl 43.6 18.9 7 4.0 1985 sdg 43.5 31.3 8 1.1
1980 det 43.7 12.0 9 1.7 2002 was 43.3 35.1 7 -4.5
1981 det 43.8 13.2 8 5.3 1988 pit 43.9 33.7 5 -5.3
1984 crd 44.6 19.0 9 5.4 2001 dal 45.0 32.2 5 -6.0
1996 nyj 44.7 19.3 1 -10.1 1981 rai 44.9 31.7 7 -3.6
1980 atl 44.8 19.5 12 7.8 1997 clt 45.7 31.6 3 -4.3
2000 atl 44.9 18.1 4 -8.6 2002 dal 45.2 33.1 5 -8.5
1986 nyj 46.1 20.4 10 -1.4 1996 rav 48.0 33.4 4 -2.9
1978 cle 47.5 18.9 8 -2.0 1997 mia 47.9 38.1 9 1.3
2002 car 47.5 19.3 7 -3.3 1998 clt 47.6 37.0 3 -5.4
1993 mia 48.9 21.8 9 -0.5 2001 clt 48.2 34.8 6 -3.8
1998 atl 49.2 21.3 14 10.0 1999 clt 48.8 37.0 13 6.1
1986 min 49.8 20.9 9 6.5 1979 chi 49.7 37.9 10 0.0
1996 sea 49.9 21.5 7 -1.7 1980 chi 50.0 36.9 7 1.4
1981 mia 51.2 19.2 11.5 2.9 1999 rav 50.7 38.0 8 0.9
1981 buf 51.3 21.5 10 1.3 1978 chi 50.5 34.6 7 -2.4
1981 nyj 51.6 19.9 10.5 1.8 1994 den 51.0 35.7 7 -3.3
2002 new 52.6 22.7 9 4.0 1998 tam 53.3 36.6 8 1.3
1981 phi 53.3 20.7 10 8.7 2002 clt 53.3 37.6 10 1.2
1980 buf 53.5 20.2 11 3.9 2002 rai 52.3 40.3 11 10.6
1978 phi 53.5 22.8 9 2.3 1995 gnb 53.6 36.5 11 6.0
2000 den 53.5 23.0 11 5.0 1990 phi 52.7 36.2 10 5.6
1999 jac 53.7 20.4 14 6.4 1999 sfo 52.1 37.8 4 -11.3
1985 sea 54.5 23.0 8 4.3 1999 mia 55.6 35.9 9 1.9
1996 rai 54.8 23.3 7 2.4 2000 sfo 54.3 35.6 6 -3.8
1979 cle 54.9 21.5 9 3.0 2000 tam 55.6 36.8 10 7.3
2001 sea 55.0 22.9 9 -1.9 1992 phi 54.5 36.3 11 8.9
1979 phi 55.8 19.9 11 2.7 1981 chi 54.3 40.0 6 -2.9
2002 pit 56.7 22.0 10.5 2.7 1993 gnb 56.7 36.5 9 3.2
2001 pit 56.8 21.5 13 7.4 1998 gnb 56.5 37.5 11 5.0
1981 den 56.9 23.1 10 1.3 1989 phi 57.3 36.3 11 5.1
1980 phi 57.5 20.0 12 9.7 1996 gnb 56.2 37.8 13 15.3
1979 mia 58.7 21.8 10 4.3 1990 buf 59.2 37.7 13 8.6
2000 pit 58.8 21.3 9 3.9 1998 sfo 59.0 38.2 12 10.6
1978 den 59.7 22.6 10 5.0 1997 gnb 58.6 36.4 13 7.7
1998 sea 61.3 23.3 8 1.5 1994 gnb 60.5 36.3 9 6.0
1993 dal 63.0 25.7 12 9.6 1980 pit 62.7 37.9 9 2.3
1981 dal 65.2 22.4 12 6.1 1993 sfo 65.9 35.3 10 9.7
1978 dal 68.7 23.2 12 11.0 1996 sfo 69.9 35.2 12 8.0

If we look at those pairs, we get the following averages:

Talent Balanced: 8.8 wins, 1.5 SRS
Talent Dispersed: 7.8 wins, 0.1 SRS

Thus, we see that the teams that were talent balanced in our pairs did tend to outperform the more talent dispersed squads with a similar overall career AV.

Breaking it down into subgroups, here are the numbers:

TEAMS WITH AVERAGE CAREER AV UNDER 44.0

Talent Balanced: 6.9 wins, -1.6 SRS
Talent Dispersed: 6.0 wins, -3.9 SRS

TEAMS WITH AVERAGE CAREER AV BETWEEN 44.1 AND 54.0

Talent Balanced: 9.3 wins, +1.9 SRS
Talent Dispersed: 7.4 wins, -0.9 SRS

TEAMS WITH AVERAGE CAREER AV ABOVE 54.1

Talent Balanced: 10.2 wins, +4.6 SRS
Talent Dispersed: 10.3 wins, +5.8 SRS

I'll caution that this is just one way of looking at it. It could be that Career AV overvalues either star players relative to Hall of Very Good Players, or that the weighting used in Career AV resulted in pairing teams that really were not similar. These numbers do, however, lend support to a chain theory, as advanced by Brian Burke in regard to passing games, and as discussed by Chase Stuart in regard to defenses. For below average and average to slightly above average overall talent teams, those with less talent dispersal (and thus fewer weak links) performed better than their paired teams with greater talent dispersal.

This entry was posted on Tuesday, August 24th, 2010 at 9:37 am and is filed under Statgeekery. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.