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.
ShareThisTwice the simplicity
Back when I used to write blog posts, I would frequently use this simple rating system as a quick gauge of team quality. To refresh your memory, it's a system that's been around forever and is extremely basic compared some of the other power rating systems out there. But I like it because it's easy to understand. An average team will have a rating of zero. An above average team will have a positive rating while a below average team will have a negative rating. Every team will have a rating that is the equal to their average point margin plus the average of their opponent's ratings, so the teams' ratings are all interdependent: the Colts' rating depends upon the ratings of all their opponents, which depends upon the ratings of all their opponents (some of which are the Colts), and so on. The 2006 Steelers had a rating of +3.4. They outscored their opponents by an average of 2.4 points per game, while their opponents were on average, according to the same kind of calculation, 1.0 points better than average. Hence we estimate that the Steelers must be 3.4 abstract theoretical points better than an abstract theoretical average team.
Another project I was working on --- and that either Chase or I will be writing about soon --- prompted me to do something I've been meaning to do for awhile, which is to separate the simple rating system into an offensive and a defensive component. Doing so, we find that the Steelers +3.4 rating breaks down into +3.0 on offense and +0.4 on defense. Meanwhile, their strength of schedule of +1.0 breaks down as +1.6 for offense and -0.5 for defense (ignore the rounding discrepancy). Here's what all that means:
The Steelers' offense scored 1.4 points per game more than an average team, but they faced defenses that, all things considered, were 1.6 points better than average. So morally the Steelers' offense was 3.0 points per game better than average.
The Steelers' defense allowed 0.9 points per game fewer than average, but the offenses they played were on average 0.5 points below average. So morally, the Steelers D was only 0.4 points per game better than average.
When you put these two facts together (and again ignore the rounding discrepancies), you get the same numbers as in the first paragraph. So what we've done is to break down the Steelers' rating --- and their strength of schedule --- into two pieces. Here are the offensive and defensive ratings for each team last season:
offense defense total
rating SOS rating SOS rating SOS
==========+================+================+===============
nwe 2006 | +4.3 +0.9 | +5.9 +0.1 | +10.2 +1.0
sdg 2006 | +10.0 -0.1 | +0.2 -1.5 | +10.2 -1.6
bal 2006 | +1.5 +0.1 | +7.8 -0.3 | +9.3 -0.2
chi 2006 | +4.9 -1.1 | +3.0 -1.7 | +7.9 -2.9
jax 2006 | +2.6 +0.1 | +4.9 +1.3 | +7.5 +1.4
ind 2006 | +6.9 +0.9 | -1.1 +0.8 | +5.9 +1.7
cin 2006 | +4.0 +1.4 | +0.0 +0.1 | +4.1 +1.5
nor 2006 | +4.9 -0.2 | -0.9 -1.4 | +4.0 -1.6
dal 2006 | +5.0 -0.9 | -1.3 -0.1 | +3.7 -1.0
pit 2006 | +3.0 +1.6 | +0.4 -0.5 | +3.4 +1.0
phi 2006 | +3.2 -1.0 | +0.2 +0.0 | +3.4 -1.0
buf 2006 | -0.2 +1.7 | +2.4 +1.2 | +2.2 +2.9
nyj 2006 | +0.4 +1.3 | +1.7 -0.5 | +2.0 +0.7
den 2006 | -0.8 -0.1 | +2.1 +0.5 | +1.3 +0.4
kan 2006 | +0.4 +0.4 | +0.6 -0.4 | +1.0 -0.0
mia 2006 | -3.3 +1.1 | +4.0 +1.1 | +0.7 +2.1
nyg 2006 | +1.2 -0.3 | -1.1 +0.8 | +0.1 +0.5
ten 2006 | +1.0 +1.4 | -2.3 +2.0 | -1.3 +3.5
car 2006 | -4.2 -0.4 | +1.5 -0.1 | -2.7 -0.5
atl 2006 | -2.8 -0.4 | -0.2 -0.4 | -3.0 -0.8
sea 2006 | -1.6 -1.9 | -1.9 -1.3 | -3.6 -3.2
stl 2006 | +0.8 -1.5 | -4.7 -1.6 | -4.0 -3.1
was 2006 | -2.1 -0.6 | -2.0 +0.9 | -4.0 +0.3
min 2006 | -3.7 -0.6 | -0.4 -0.7 | -4.1 -1.3
gnb 2006 | -2.3 -0.4 | -2.1 +0.1 | -4.4 -0.4
hou 2006 | -3.2 +0.8 | -1.3 +0.9 | -4.5 +1.7
cle 2006 | -4.5 +1.3 | -1.3 +0.3 | -5.8 +1.5
det 2006 | -1.8 -0.2 | -4.6 -0.4 | -6.4 -0.5
ari 2006 | -2.6 -1.6 | -4.3 -0.6 | -6.9 -2.2
tam 2006 | -7.2 +0.3 | -0.8 +0.7 | -7.9 +0.9
sfo 2006 | -3.5 -1.4 | -5.2 -0.2 | -8.7 -1.6
oak 2006 | -10.3 -0.2 | +0.7 +0.8 | -9.6 +0.6
Some basic SRS trivia:
Best and worst offenses since 1970
offense defense total
rating SOS rating SOS rating SOS
==========+================+================+===============
min 1998 | +13.2 -0.2 | +1.6 -1.1 | +14.9 -1.4
stl 2000 | +12.6 -0.4 | -9.5 -0.7 | +3.1 -1.2
was 1983 | +11.7 -0.2 | +2.2 +1.1 | +13.9 +0.9
ind 2004 | +11.7 +0.6 | -0.3 +0.1 | +11.4 +0.7
was 1991 | +11.7 +0.4 | +4.9 -0.0 | +16.6 +0.3
sdg 1982 | +10.5 -1.3 | -5.4 -1.0 | +5.1 -2.3
stl 2001 | +10.4 -0.8 | +2.9 -0.2 | +13.4 -1.0
buf 1975 | +10.4 +1.0 | -3.4 +1.4 | +7.1 +2.4
kan 2004 | +10.0 +1.3 | -4.7 +1.0 | +5.3 +2.3
sdg 2006 | +10.0 -0.1 | +0.2 -1.5 | +10.2 -1.6
sfo 1993 | +9.9 -0.9 | -0.3 -0.5 | +9.7 -1.5
sfo 1994 | +9.9 -1.4 | +1.7 -0.0 | +11.6 -1.4
den 1998 | +9.5 -0.5 | -0.6 -2.6 | +8.9 -3.1
bal 1976 | +9.3 -1.3 | +0.4 -1.1 | +9.8 -2.5
kan 2003 | +9.2 -0.2 | -0.9 -1.0 | +8.3 -1.2
sdg 1981 | +9.2 -0.0 | -4.7 -1.0 | +4.4 -1.1
mia 1984 | +9.1 -1.8 | +1.5 -1.1 | +10.6 -2.9
sfo 1998 | +8.8 +0.2 | +1.8 +1.0 | +10.6 +1.2
sfo 1987 | +8.8 -0.2 | +4.2 -0.5 | +13.1 -0.7
oak 1977 | +8.8 +0.9 | +1.6 +0.9 | +10.4 +1.7
stl 1999 | +8.7 -3.4 | +3.2 -2.5 | +11.9 -5.9
gnb 1996 | +8.7 +0.6 | +6.6 -0.7 | +15.3 -0.1
nyj 1982 | +8.7 +1.6 | +1.6 -0.1 | +10.3 +1.5
dal 1983 | +8.6 +0.5 | -0.1 +0.6 | +8.5 +1.1
kan 2002 | +8.4 +0.9 | -2.3 +1.0 | +6.1 +1.9
sdg 1985 | +8.3 +0.7 | -7.2 -1.6 | +1.1 -0.9
dal 1971 | +8.3 -1.4 | +1.6 -1.9 | +9.9 -3.3
ram 1973 | +8.2 -0.1 | +5.2 -1.5 | +13.4 -1.6
ind 2003 | +8.2 +1.0 | -1.2 -1.0 | +7.0 +0.0
.....
cin 1993 | -7.4 -0.4 | -1.0 +0.2 | -8.4 -0.2
chi 1974 | -7.4 -0.1 | -2.9 -1.2 | -10.3 -1.2
sea 1982 | -7.4 -1.4 | +2.0 -1.8 | -5.4 -3.2
tam 1983 | -7.5 -0.7 | -1.6 +0.4 | -9.0 -0.3
nwe 1990 | -7.5 +1.3 | -7.1 +0.7 | -14.6 +2.0
cle 1999 | -7.6 -0.4 | -6.5 +0.0 | -14.1 -0.3
dal 1989 | -7.6 +0.2 | -2.8 +1.2 | -10.4 +1.4
sdg 1998 | -7.7 -1.4 | -0.7 -0.6 | -8.4 -2.1
nwe 1991 | -7.9 -2.1 | -0.8 -0.7 | -8.7 -2.8
nyj 1995 | -7.9 -1.0 | -3.3 -0.8 | -11.2 -1.7
atl 1987 | -7.9 +0.0 | -5.9 +1.5 | -13.9 +1.5
cin 2000 | -8.1 +1.0 | -2.4 -0.6 | -10.5 +0.4
dal 2002 | -8.2 -0.1 | -0.3 -1.4 | -8.5 -1.5
nor 1975 | -8.2 +0.6 | -5.9 -0.7 | -14.1 -0.1
hou 2002 | -8.4 -0.0 | -1.1 -0.5 | -9.4 -0.5
chi 2004 | -8.5 -1.4 | +0.3 -0.5 | -8.2 -2.0
cle 2000 | -9.1 +1.5 | -5.5 +0.0 | -14.6 +1.5
gnb 1977 | -9.1 -1.5 | +0.4 -1.1 | -8.7 -2.6
bos 1970 | -9.1 -0.5 | -6.8 -0.2 | -15.9 -0.7
phi 1972 | -9.2 +0.7 | -5.3 -0.4 | -14.6 +0.2
buf 1985 | -9.5 -0.5 | -0.2 +2.0 | -9.8 +1.5
atl 1974 | -9.7 +0.6 | -2.6 -1.5 | -12.3 -0.9
sea 1992 | -9.8 +0.2 | -0.2 +0.6 | -10.0 +0.8
oak 2006 | -10.3 -0.2 | +0.7 +0.8 | -9.6 +0.6
tam 1976 | -10.8 -0.6 | -8.8 +1.4 | -19.7 +0.8
tam 1977 | -11.1 -1.3 | +0.3 -0.9 | -10.7 -2.2
ind 1991 | -11.5 -1.5 | -5.7 -0.9 | -17.3 -2.4
phi 1998 | -12.3 -1.1 | -0.5 -0.2 | -12.8 -1.3
Best and worst defenses since 1970
offense defense total
rating SOS rating SOS rating SOS
==========+================+================+===============
min 1971 | -3.4 -1.5 | +9.9 +0.5 | +6.5 -1.1
tam 2002 | -1.0 -0.9 | +9.8 +0.4 | +8.8 -0.6
pit 1976 | +5.7 +0.4 | +9.6 +0.3 | +15.3 +0.8
chi 1985 | +6.5 -0.5 | +9.4 +0.3 | +15.9 -0.2
den 1977 | +1.9 -0.5 | +9.4 +2.8 | +11.3 +2.3
min 1970 | +5.9 +1.2 | +9.2 +0.1 | +15.1 +1.4
pit 1975 | +5.4 -0.7 | +8.8 -0.2 | +14.2 -0.9
ram 1975 | +0.5 -1.2 | +8.7 -2.3 | +9.1 -3.5
bal 1971 | +2.0 -1.0 | +8.5 -0.9 | +10.4 -1.9
mia 1973 | +5.0 -0.1 | +8.2 -0.5 | +13.2 -0.6
bal 2000 | +0.0 -0.1 | +8.0 -2.3 | +8.0 -2.5
bal 2006 | +1.5 +0.1 | +7.8 -0.3 | +9.3 -0.2
chi 1986 | -0.1 -1.6 | +7.7 -1.1 | +7.6 -2.7
nyg 1990 | +0.0 -0.8 | +7.6 +0.7 | +7.7 -0.1
buf 1999 | -0.2 +0.6 | +7.4 +0.9 | +7.1 +1.5
mia 2000 | +0.0 +0.5 | +7.1 +0.5 | +7.1 +1.0
nor 1992 | +0.7 -1.2 | +7.1 +1.0 | +7.8 -0.2
chi 2001 | +0.8 -0.1 | +7.1 -0.5 | +7.9 -0.5
min 1973 | +1.7 -0.0 | +7.0 -0.5 | +8.6 -0.5
den 1978 | -1.9 -1.2 | +6.9 +1.0 | +5.0 -0.2
kan 1995 | +0.8 -0.1 | +6.9 +0.4 | +7.6 +0.3
phi 1980 | +2.9 -0.6 | +6.8 +0.2 | +9.7 -0.4
ten 2000 | +1.5 +0.6 | +6.8 -1.9 | +8.3 -1.3
phi 2001 | +0.9 -0.3 | +6.8 -0.4 | +7.7 -0.7
bal 2004 | -0.6 +1.0 | +6.8 +2.0 | +6.1 +3.1
tam 1999 | -4.1 -0.2 | +6.7 +0.6 | +2.6 +0.4
oak 1973 | +0.1 -1.3 | +6.7 -0.3 | +6.8 -1.6
phi 1981 | +2.0 -0.3 | +6.7 -0.2 | +8.7 -0.5
den 1984 | +1.0 +0.1 | +6.7 +0.5 | +7.7 +0.7
.....
nwe 1990 | -7.5 +1.3 | -7.1 +0.7 | -14.6 +2.0
hou 1982 | -3.9 +1.2 | -7.1 -0.0 | -10.9 +1.2
nyj 1975 | -1.2 +0.9 | -7.1 +3.2 | -8.3 +4.2
gnb 1983 | +5.2 +0.3 | -7.1 -1.5 | -1.9 -1.2
sdg 1985 | +8.3 +0.7 | -7.2 -1.6 | +1.1 -0.9
nyj 1996 | -2.5 +0.5 | -7.6 +0.3 | -10.1 +0.8
phi 1973 | +3.4 +0.7 | -7.7 +0.9 | -4.3 +1.6
ram 1982 | +1.0 -1.1 | -7.8 -0.1 | -6.8 -1.2
cle 1990 | -5.2 +0.7 | -7.8 +1.0 | -13.0 +1.7
hou 1983 | -3.7 +0.2 | -7.9 -0.9 | -11.5 -0.8
ari 2000 | -7.2 +0.4 | -8.1 -1.1 | -15.2 -0.7
min 1984 | -2.7 +1.3 | -8.2 +0.9 | -10.9 +2.1
cin 1999 | -3.3 -0.2 | -8.2 -0.3 | -11.5 -0.5
sdg 1973 | -3.6 +2.4 | -8.3 -0.2 | -11.9 +2.2
sfo 2004 | -5.1 +0.2 | -8.6 -1.8 | -13.6 -1.6
sea 1977 | +4.2 +1.3 | -8.6 +0.9 | -4.3 +2.2
tam 1976 | -10.8 -0.6 | -8.8 +1.4 | -19.7 +0.8
atl 1996 | -0.2 +0.9 | -8.9 -0.5 | -9.1 +0.4
buf 1971 | -4.5 +1.7 | -8.9 -0.1 | -13.4 +1.6
nor 1980 | -1.1 +1.2 | -9.4 +0.6 | -10.4 +1.8
nor 1977 | +0.3 +0.9 | -9.4 -2.6 | -9.2 -1.7
stl 2000 | +12.6 -0.4 | -9.5 -0.7 | +3.1 -1.2
ind 2001 | +6.1 +0.5 | -9.8 +0.3 | -3.8 +0.8
tam 1986 | -4.6 +1.0 | -10.8 -1.8 | -15.4 -0.8
nwe 1972 | -6.3 +0.2 | -11.1 +0.6 | -17.4 +0.7
bal 1981 | -3.8 +0.7 | -12.0 +0.6 | -15.8 +1.3
hou 1973 | -4.3 +0.9 | -12.3 +0.1 | -16.7 +1.1
sea 1976 | -2.2 +0.6 | -12.9 -1.4 | -15.1 -0.8
Biggest discrepancies between offense and defense
offense defense total
rating SOS rating SOS rating SOS
==========+================+================+===============
stl 2000 | +12.6 -0.4 | -9.5 -0.7 | +3.1 -1.2
sdg 1982 | +10.5 -1.3 | -5.4 -1.0 | +5.1 -2.3
ind 2001 | +6.1 +0.5 | -9.8 +0.3 | -3.8 +0.8
sdg 1985 | +8.3 +0.7 | -7.2 -1.6 | +1.1 -0.9
kan 2004 | +10.0 +1.3 | -4.7 +1.0 | +5.3 +2.3
sdg 1981 | +9.2 -0.0 | -4.7 -1.0 | +4.4 -1.1
buf 1975 | +10.4 +1.0 | -3.4 +1.4 | +7.1 +2.4
cin 1985 | +6.2 +0.2 | -6.9 -1.1 | -0.7 -0.9
sea 1977 | +4.2 +1.3 | -8.6 +0.9 | -4.3 +2.2
gnb 1983 | +5.2 +0.3 | -7.1 -1.5 | -1.9 -1.2
ind 2004 | +11.7 +0.6 | -0.3 +0.1 | +11.4 +0.7
mia 1986 | +6.2 -0.2 | -5.7 -0.9 | +0.5 -1.1
buf 1991 | +7.6 -2.0 | -4.1 -3.2 | +3.6 -5.2
min 1998 | +13.2 -0.2 | +1.6 -1.1 | +14.9 -1.4
phi 1973 | +3.4 +0.7 | -7.7 +0.9 | -4.3 +1.6
sea 1976 | -2.2 +0.6 | -12.9 -1.4 | -15.1 -0.8
kan 2002 | +8.4 +0.9 | -2.3 +1.0 | +6.1 +1.9
was 1999 | +6.8 -0.1 | -3.9 -1.2 | +2.9 -1.2
bal 1996 | +3.9 +1.1 | -6.8 +0.3 | -2.9 +1.4
den 2000 | +7.8 -1.9 | -2.7 -0.3 | +5.0 -2.2
sfo 1993 | +9.9 -0.9 | -0.3 -0.5 | +9.7 -1.5
kan 2003 | +9.2 -0.2 | -0.9 -1.0 | +8.3 -1.2
den 1998 | +9.5 -0.5 | -0.6 -2.6 | +8.9 -3.1
det 1990 | +4.2 +1.0 | -5.9 -0.2 | -1.6 +0.9
cin 1986 | +5.7 +0.6 | -4.2 -0.1 | +1.5 +0.5
nor 1977 | +0.3 +0.9 | -9.4 -2.6 | -9.2 -1.7
sdg 2006 | +10.0 -0.1 | +0.2 -1.5 | +10.2 -1.6
was 1983 | +11.7 -0.2 | +2.2 +1.1 | +13.9 +0.9
ind 2003 | +8.2 +1.0 | -1.2 -1.0 | +7.0 +0.0
......
cle 1984 | -6.1 -0.5 | +2.3 -0.3 | -3.8 -0.8
chi 2004 | -8.5 -1.4 | +0.3 -0.5 | -8.2 -2.0
tam 1996 | -5.9 +0.7 | +2.9 +0.8 | -3.0 +1.5
car 2002 | -6.0 -0.5 | +2.8 -0.0 | -3.3 -0.5
tam 1995 | -6.9 -0.3 | +1.9 +1.4 | -5.0 +1.1
ari 1994 | -5.6 +0.0 | +3.3 -0.3 | -2.3 -0.3
den 1978 | -1.9 -1.2 | +6.9 +1.0 | +5.0 -0.2
nwe 1988 | -4.5 +0.1 | +4.4 +1.9 | -0.2 +2.0
chi 1993 | -4.6 -0.5 | +4.3 +0.0 | -0.2 -0.5
nyg 1983 | -5.5 -0.4 | +3.4 +3.3 | -2.1 +2.9
buf 2003 | -5.0 +0.7 | +4.0 +0.6 | -1.0 +1.3
kan 1973 | -4.1 -1.2 | +5.1 -0.6 | +1.0 -1.8
buf 1985 | -9.5 -0.5 | -0.2 +2.0 | -9.8 +1.5
sea 1982 | -7.4 -1.4 | +2.0 -1.8 | -5.4 -3.2
kan 1979 | -4.0 +1.2 | +5.4 +1.8 | +1.4 +2.9
gnb 1977 | -9.1 -1.5 | +0.4 -1.1 | -8.7 -2.6
sea 1992 | -9.8 +0.2 | -0.2 +0.6 | -10.0 +0.8
phi 1983 | -6.7 +0.5 | +3.6 +1.9 | -3.1 +2.4
tam 2002 | -1.0 -0.9 | +9.8 +0.4 | +8.8 -0.6
tam 1999 | -4.1 -0.2 | +6.7 +0.6 | +2.6 +0.4
oak 2006 | -10.3 -0.2 | +0.7 +0.8 | -9.6 +0.6
was 2004 | -7.4 -0.9 | +4.0 -0.9 | -3.4 -1.8
tam 1977 | -11.1 -1.3 | +0.3 -0.9 | -10.7 -2.2
phi 1998 | -12.3 -1.1 | -0.5 -0.2 | -12.8 -1.3
chi 2005 | -5.2 -0.9 | +6.6 -1.4 | +1.4 -2.2
jax 2004 | -5.7 -0.5 | +6.4 +2.5 | +0.8 +1.9
atl 1977 | -6.1 -1.7 | +6.6 -1.4 | +0.5 -3.1
min 1971 | -3.4 -1.5 | +9.9 +0.5 | +6.5 -1.1
This entry was posted on Tuesday, August 21st, 2007 at 4:20 am and is filed under General. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
Do you have a file with each seasons results that I could access? Thanks in advance!
Very cool. Almost too much info to absorb at once.
One question, though. I understand the point of adjusting for opponent strength. By why is the word "morally" used to describe the resulting comparison? I've seen that used before and don't understand the use of "moral" in that context.
dave, I don't at this point, but probably will before too long.
Brian, I honestly haven't given my use of the word "morally" a lot of thought. But I think I use it because a lot of other adverbs would be ambiguous in that spot. And even the lack of any adverb is ambiguous.
For example, I might say that the Steelers offense was actually (or truly, or really) 3.0 points per game better than average. Well, no. They were actually (as in, in the actual (or true or real) world we live in) 1.4 points better than average. But in a perfect abstract world where every team played a fair schedule, our best estimate is that they would have been 3.0 points per game better. Because we all believe fairness is to be strived for, it's but a step from there to: "The Steelers offense should have been 3.0 points per game better than average." The word "should" implies some sort of moral judgement. Hence my use of the word. But like I said, I'd never given it any thought until now.
What I'm looking for there, I guess, is a word that contrasts with the word "literally."
The Steelers' O was literally 1.4 points above average. But they were __________ly 3.0 points better than average.
Splitting ppg into offense and defense should only be seen as a rough approximation, since:
- defense and special teams also score points
- pace varies
- field position is fluid
The 2006 Bears scoring, for instance, was aided by 6 Hester touchdowns plus a couple more from the defense. Based on FO drive stats, the Bears offense was also among the top 5 in number of drives and average starting field position (with 191 drives, starting on average past their 32 yard line), which means that they had both more chances to score and easier chances to score (thanks, again, to that D and special teams). So Chicago's simple offensive scoring rating of +4.9 may be one of the top 5 for 2006, but that is not because they fielded a top 5 offense.
Interesting how this seems to mesh with your earlier post on offensive and defensive consistency and how offenses tend to be less consistent than defenses, which was at least partially attributed to the importance of the quarterback position.
Best offense: +13.2
Best defense: +9.9
Worst offense: -12.3
Worse defense: -12.9
10 offenses were ranked higher than +9.9. On the other hand, only one defense ranked lower than -12.3. So it looks like the best offenses are significantly better than the best defenses, but the worst offenses aren't really much worse than the worst defenses.
We could use the "QB factor" to attribute the great discrepancy at the top of the rankings, but why is it so minor at the bottom?
Are these ratings perfect? No. Are they awesome? Yes.
Interesting that even when adjusted for era, most of the really good defenses are from the 70s, and most of the really good offenses are from the 80, 90s, 00s.
I like the simple rating system but separating it out makes it lose a bit of its value. The 1991 Eagles who were one of the best defenses ever don't show up int he top 20 using this method! I like for a method to at least pass the "sniff" test.
Pretty much I'm echoing Dan's comment. The Philly offense was just about as bad as the defense was good that year(especially without at least McMahon) contributing to the defense giving up more points.
figuratively?
representatively?
MS Word Thesaurus is the bomb.
adjustedly?
Ok, I made that one up.
"The Steelers D was only 0.4 points per game better than average" Doug said abstract-theoretically.
Hmm... Offense and defense are obviously linked as a better offensive team will be more likely to score and leave the defense behind the 30 on a kickoff, or if they don't, they'll still be in a better position than teams who habitually go 3-and-out. And better defenses give offenses better starting position too. For instance, I think Oakland's defense was probably better than indicated by this system since their offense was so inept. Indy's defense (at least during the regular season) was probably worse than the numbers suggest as well.
I wonder if it'd be possible to somehow credit defenses more who are tied to a bad offense, or penalize ones tied to a good offense, and credit/penalize offenses based on the defense they're tied to. Obviously it's POSSIBLE, but I was trying to think of a "moral" way, without the use of made-up magic numbers, perhaps similar to the iterative system you're using for simple scoring, or some sort of curve fitting... It'd be fun to play with anyway.
Speaking of... I've already snaked the files you have for download so I can crunch numbers, but there doesn't seem to be any files with historic team records, scores, and whatnot. Are those available here, or somewhere else?
Thanks. I'm on board with morally. I think truly might be a good fit though.
11: I agree that there are interaction effects between offense and defense. But I think they are far more complex than what you suggest. It might not be true that a good offense means the opponent will score less, or a bad offense allows the opponent score more.
I'm a Ravens fan, so I think about this a lot. There are lots of situations where a team with a big lead "eases up" allowing the opponents to make completions and score but keeping them in bounds to run out the clock. It's part of what makes trash time so trashy, and it allows the opponent to score more than he 'morally' should have.
And there is the reverse effect too. The Ravens will very often play "field goal ball" instead of passing in the red zone to go for a TD. They know they only need a 9 point lead to seal a game. A great defense allows an offense to play very conservatively and still win, which leads to low point totals.
Also consider that the defense can, by itself, score points. Beyond the occasional safety or interception return TD, consider this example: The Ravens defense gets a turnover on the opponent's 35. On comes Jamal Lewis for 3 up-the-gut runs for a total of 5 yds. In comes Matt Stover for the automatic 3 points. It seems to happen twice a game. Those points really belong to the defense.
I'm not a big fan of using only points and point differentials to compare offenses and defenses for this reason. But I think it is a very fun way to compare teams across eras.
This is all intuitive at this point, but I think it would be interesting to see what the dependency effects are between points allows and points scored. I know in baseball, they've found that teams with big leads allow more runs.
Who are the best and worst teams since 1970?
11: I agree that there are interaction effects between offense and defense. But I think they are far more complex than what you suggest. It might not be true that a good offense means the opponent will score less, or a bad offense allows the opponent score more.
No, it's pretty much always true. Even if the offense chooses to score fewer points (which I doubt they would ever really do - they'd choose the safer points rather than the riskier points, but in the end that averages out to the same, assuming the risk calculation is right), they're given more, and better, scoring opportunities by the defense. The corrections you're thinking of are certainly second order - you can figure that just by imagining the total number of situations in which the team has a choice compared to the total situations.
In any case, it's easy enough to verify. Just look for a correlation between the offensive and defensive scores. You'll probably find a slight one.
Also, Doug:
The schedule strength corrections you're making seem
1) odd in that they're additive
2) first order only.
For point 1, I don't see why you'd expect the opponent strength to be additive rather than multiplicative. If the average team scores 21.5 points per game, you allow 18 points per game, and you faced 25 point per game opponents, I don't know why you'd expect 14.5 points per game versus an average opponent (25-18=7, 21.5-7=14.5) versus 15.5 (18/25=0.72, 0.72*21.5=15.5). There's no real difference for small numbers, so it's not a big deal, but it still seems odd to me.
For point 2, the point there is that your estimate of the strength of opponents is biased by their strength of schedule as well. The simple method there is to take the "schedule-corrected" value and use it to compute a new strength-of-schedule for each team, and a new schedule-corrected-corrected version. That's a second order solution.
(You could go all out and just solve the all-order matrix, too.)