≡ Menu

Game Scripts, Part II: Analyzing team seasons

Yesterday, I rolled out Game Scripts, a way to measure the flow of every game since 1940. The sum of each team’s Game Script in each game can be used to give us an average Game Script score on the season. You might think that this number would be a good proxy for how dominant a team was, and that’s largely true: the teams with the highest game script scores tend to have been the most dominant teams. However, there are some reasons to be cautious with this approach: game scripts are not adjusted for strength of schedule and in any given game, the losing team can end up with a better score than the winning team. That said, here are the teams with the highest Game Scripts since 1940:

RkYearTeamLeagueW-L-TPFPASCRIPT
11942CHINFL11-0-03768413.5
21948CHINFL10-2-037515111.3
31941CHINFL10-1-039614710.4
41948SFOAAFC12-2-049524810.4
52007NWENFL16-0-058927410.3
61968BALNFL13-1-040214410.1
71948PHINFL9-2-137615610.1
81947CLEAAFC12-1-141018510
91946CLEAAFC12-2-042313710
101949PHINFL11-1-03641349.5
111969MINNFL12-2-03791339.4
121954CLENFL9-3-03361629.2
131999STLNFL13-3-05262429.1
141973MIANFL12-2-03431509.1
152001STLNFL14-2-05032738.9
161961HOUAFL10-3-15132428.8
171951CLENFL11-1-03311528.8
181972MIANFL14-0-03851718.7
191998MINNFL15-1-05562968.6
201973RAMNFL12-2-03881788.5
211983WASNFL14-2-05413328.4
221984SFONFL15-1-04752278.4
231948CLEAAFC14-0-03891908.3
241949SFOAAFC9-3-04162278.2
251998DENNFL14-2-05013098.1
261968DALNFL12-2-04311868
271966KANAFL11-2-14482767.9
281995SFONFL11-5-04572587.7
291962GNBNFL13-1-04151487.7
301953CLENFL11-1-03481627.6
311971DALNFL11-3-04062227.6
321944PHINFL7-1-22671317.6
331948CRDNFL11-1-03952267.6
341960CLENFL8-3-13622177.5
351980RAMNFL11-5-04242897.4
362010NWENFL14-2-05183137.4
372011GNBNFL15-1-05603597.4
381976BALNFL11-3-04172467.4
391975MINNFL12-2-03771807.3
401975PITNFL12-2-03731627.3
411992DALNFL13-3-04092437.3
421969KANAFL11-3-03591777.3
431964BALNFL12-2-04282257.2
441997DENNFL12-4-04722877.2
451968OAKAFL12-2-04532337.2
461945RAMNFL9-1-02441367
471943CHINFL8-1-13031577
481967OAKAFL13-1-04682337
491963NYGNFL11-3-04482807
501994SFONFL13-3-05052966.9

The teams with the highest game scripts last year? Green Bay (7.4), New Orleans (5.6) and Houston (5.4), while the Rams (-6.4), Colts (-7.2), and Bucs (-8.7) were at the bottom of the league. But let’s get to the real point of using Game Scripts — to help put passing and rushing ratios in context.

Last year, the Buccaneers had the second highest effective pass/run ratio in the league (defined as total pass attempts divided by rushes plus total pass attempts, but with all kneels and spikes excluded). But that’s misleading, because Tampa Bay had the worst Game Script in the league. Conversely, were Houston and San Francisco really the second and third most run-heavy teams in the NFL last year? The table below lists each team from highest to lowest pass/run ratio:

TmPassRushP/RSCRIPT
DET70235066.7%0.7
TAM61433664.6%-8.7
TEN60636462.5%0.2
NOR68242461.7%5.6
WAS62739261.5%-1.9
ARI60237861.4%-3.5
GNB59037361.3%7.4
NYG61639061.2%-1
BUF59638460.8%-2.1
DAL60539660.4%1.1
NWE64342560.2%4.3
IND56437560.1%-7.2
CLE60740560%-3.9
STL60440260%-6.4
SDG61041659.5%1.2
ATL61443258.7%3.9
PIT57941858.1%4
NYJ58343657.2%-0.8
PHI58243757.1%3.7
BAL57743956.8%4.1
CAR55243555.9%1.4
SEA55543855.9%-0.4
MIN55944255.8%-2.8
CIN56044555.7%-0.6
OAK54645954.3%-1.7
CHI51944953.6%1.9
MIA52245653.4%2.6
KAN53247652.8%-4
JAX51247451.9%-3.5
SFO49647751%4.7
HOU49453348.1%5.4
DEN46953746.6%-3.7

But we need to consider the passing ratios in light of the Game Scripts. Teams like Green Bay and New Orleans, and to a lesser extent New England, Atlanta, and Pittsburgh would have passed much more frequently if they weren’t racing out to large leads. Think back to the ’99 Rams, the first edition of the “Greatest Show on Turf” and the only team to win the Super Bowl. You surely think of them as a pass-heavy team, with Kurt Warner, Isaac Bruce, Torry Holt, Marshall Faulk, Az-Zahir Hakim and Ricky Proehl. In 1999, NFL teams passed on 57.1% of all plays. The ’99 Rams passed on 56.6% of their plays that season. That’s right: if so inclined, you could paint a picture that the ’99 Rams were more of a running team than a passing team.

An even more recent example would be the ’07 Patriots. On average, teams passed on 56.6% of plays in ’07; the high-flying, 16-0 Patriots passed on 57.4% of their pass plays, barely more than league average. These odd results happen every season, including 2011. Would you have guessed that the Cardinals and pre-Robert Griffin III Redskins both passed on a higher percentage of their plays than the Packers?

Controlling for a team’s pass/run ratio based on their Game Script is difficult, since they involve completely different units. There is a solution to this problem, although watch out: it involves math. What I did was calculated the standard deviation for each of these metrics to come up with ratings that are similar in meaning to PFR’s passing index numbers. A score of 115 means a team is one standard deviation above average; a score of 85 means a team is one standard deviation below, 70 means two standard deviations below, etc.

How do we calculate them?

  • For game scripts, the average is always zero. For each league and for each season (so as to not combine the NFL and AFL) since 1940, I computed the standard deviation of game scripts among teams. This was necessary because, as you would expect, the standard deviation used to be considerably higher than it has been over the past 30 years. For each team, the formula is: 100 + 15 * (Game Script Average / Standard Deviation for all teams that season).
  • For pass percentage, we use a similar formula, except the average is not zero. So the formula is: 100 + 15 * ( (PassPercentage – LeagueAveragePassPercentage)/(Standard Deviation for all teams that season) ).

Let’s re-look at the 2011 standings, but this time, I’m going to include their Pass Index and Game Script Index numbers (based on standard deviations above average), and sort the table by a column I’m calling Pass Identity.

TmPassRushP/RSCRIPTPASS IndexSCRIPT IndexPass Identity
GNB59037361.3%7.411113243
NOR68242461.7%5.611212437
DET70235066.7%0.712810331
NWE64342560.2%4.310811826
ATL61443258.7%3.910311720
PIT57941858.1%410111719
TEN60636462.5%0.211510116
BAL57743956.8%4.19711815
PHI58243757.1%3.79811615
DAL60539660.4%1.110910514
SDG61041659.5%1.210610511
NYG61639061.2%-1111957
WAS62739261.5%-1.9112924
BUF59638460.8%-2.1110911
CAR55243555.9%1.4951060
SFO49647751%4.7791210
MIA52245653.4%2.687111-2
ARI60237861.4%-3.511285-3
CHI51944953.6%1.987108-4
NYJ58343657.2%-0.89997-5
HOU49453348.1%5.470123-7
SEA55543855.9%-0.49498-7
CIN56044555.7%-0.69497-9
CLE60740560%-3.910783-10
TAM61433664.6%-8.712262-16
MIN55944255.8%-2.89488-18
OAK54645954.3%-1.79093-18
STL60440260%-6.410772-21
IND56437560.1%-7.210769-24
KAN53247652.8%-48583-33
JAX51247451.9%-3.58285-33
DEN46953746.6%-3.76584-51

Looking at the top row, the Packers had 590 pass attempts and ran 373 times. They passed on 61.3% of their plays, and had a game script of 7.4. So this was a team that passed a lot even though it was frequently ahead in games, often by double digits. The Packers were nearly one standard deviation above average in pass percentage while being over two full standard deviations above average in terms of Game Script. That’s what the numbers in the next two columns represent. So how do we judge how pass-heavy an offense is? The Packers pass ratio was 11 units above average, despite a season-long game script that was 32 units above average. Because we’ve converted game scripts and pass percentage into the same units, we can simply add the two, to measure how pass-heavy an offense was. That makes their passing identity 43 units above average, meaning no team was more pass-heavy than the Packers last year. The Saints weren’t far behind, with the Lions and Patriots up next. That passes the smell test.

That’s the point of Game Scripts. We can measure offensive philosophy much more acutely by understanding the flow of the game. When teams are passing even when they’re leading, that’s a team that’s passing by design, not necessity.

In 2007, the Patriots ranked 14th in the league in pass to run ratio. But using the “Pass Identity” formula described above, New England had a score of 49, highest in the league and the fourth highest of all-time. The ’99 Rams ranked 18th out of 31 teams in pass percentage. But they had a Pass Identity grade of 36, highest in the league.

When I developed Game Scripts, the goal wasn’t to create a new stat to tell us something we didn’t know. Instead, my intention was to create a stat to help us empirically show what we already know. The goal isn’t to wow anyone by saying that the ’99 Rams and ’07 Pats passed a lot, but rather to adjust the stats to show how that was the case. By that measure, I consider Game Scripts a success.

What teams had the truest passing identity of all-time? The 1995 49ers were a great team. They had the highest Game Script in the league that season, at 7.7. They outscored their opponents by 199 points, over 50 points more than the team with the second-best points differential. Yet they called 677 passing plays, and finished with the sixth highest pass percentage in the league. [1]Note that for years since 2000, I removed spikes and kneels, but for years prior to 2000, I used the more familiar data available at PFR. This was also the year that Jerry Rice set the single-season receiving mark. With a game script of 138 and a pass ratio of 116, the 49ers had a Pass Identity of 53, just edging out two Run-n-Shoot Oilers teams for #1 on the list. There aren’t too many surprises at the top, as the ’07 Pats, ’79 Air Coryell Chargers,, the ’01 GSOT Rams, a couple more Run-N-Shoot Oilers teams, and last year’s Packers teams are next.

YearLgTmPassRushP/RSCRIPTPASS IndexSCRIPT IndexPass Identity
1995NFLSFO67741562%7.711613853
1991NFLHOU69133167.6%2.514111151
1990NFLHOU67832867.4%4.113311851
2007NFLNWE58645158.4%10.310514549
1979NFLSDG57248154.3%4.611912948
2001NFLSTL55141659.1%8.910913848
1993NFLHOU65740961.6%4.512512146
1992NFLHOU60535363.2%3.313011545
2011NFLGNB59037361.3%7.411113243
1981NFLMIN73839165.4%-0.31449942
2004NFLPHI54737662.1%4.812112142
1974NFLPHI50341554.8%2.313011140
2002NFLOAK61941462%5.611612440
2010NFLNOR66138065.1%3.612411639
1983NFLKAN68738764%1.513210739
1950NFLRAM47640454.1%6.511612339
1960NFLBAL41834554.8%4.912211739
1995NFLDET63738762.2%4.511612238
2009NFLIND60136663.7%3.612111637
1978NFLMIN62250555.2%-0.51409737
2011NFLNOR68242461.7%5.611212437
2009NFLPHI55338461.8%4.911512137
1980NFLSDG62650955.2%4.811312436
1999NFLSTL56343156.6%9.19813836
1994NFLGNB64241760.6%4.211412134
1998NFLMIN55845055.4%8.610113434
2003NFLSTL60041161.6%3.312011434
1999NFLCAR62635663.7%2.412411034
2009NFLGNB55343858.9%6.310712734
1995NFLATL64633765.7%1.212810634
1985NFLMIA59544457.3%4.411212234
2007NFLGNB57838861.7%4.311511933
1996NFLGNB58846555.8%6.510113333
2010NFLIND67939364.8%2.312311033
2008NFLPHI60642760.2%4.511311933
1994NFLMIN70441962.7%2.612011333
1986NFLMIA66234965.5%-0.31349933
1996NFLNWE65842760.6%2.911811432
1987NFLMIA59740859.4%1.812410832
1952NFLCLE40839450.9%4.911611532
1947NFLCHI37844845.8%5.611012232
1982NFLSDG35026756.7%4.611112031
1943NFLWAS25432044.3%6.111711431
2011NFLDET70235066.7%0.712810331
2009NFLNWE59246657.5%6.610212931
1995NFLGNB62641060.4%4.211012031
1988NFLMIN56750153.1%5.710113031
1982NFLCIN33726955.6%5.110812331
1981NFLCIN58549354.3%4.411012131
1961NFLPHI45637355%2.811911131
1986NFLMIN56346155%610412730
2009NFLARI59436563.9%1.912210830
1967NFLBAL48244352.1%6.710412630
1953NFLPHI47241053.5%511511530
1963NFLNYG46145350.4%710412630
2000NFLCAR56636364.1%1.312410630
1946NFLCRD26637141.8%4.911111930
1951NFLCLE31741543.3%8.89913130
1994NFLATL66633066.9%-0.41329830
1980NFLCLE57743657%2.411811230
2008NFLARI63034066.9%-1.11349530
2000NFLSTL58738362.2%2.611811230

We can also use game scripts to measure the most run-heavy teams. We often think of run-heavy teams like the ’90s Cowboys or ’60s Packers, but that’s because they were so frequently playing with leads. For bad teams with a low Pass Identity, it may be a better indicator that the quarterback was just really, really bad.

You shouldn’t be too surprised at this list then, either. The most run-heavy team after adjusting for situation was the Tim Tebow 2011 Broncos, followed by the ’72 Bears, with Bobby Douglass at quarterback. The ’72 Bears were a bad team, but they had the fewest pass attempts in the league along with the lowest pass percentage. You probably aren’t surprised to see a Cowboys team with Emmitt Smith high on the list, but I doubt it was one of the ones you were thinking of. Rather it was the ’01 team, in Smith’s penultimate season in Dallas. The Cowboys alternated between Quincy Carter, Anthony Wright, Ryan Leaf and Clint Stoerner at quarterback. Dallas wasn’t very good, and ended the year ranked 30th in points. But Dave Campo’s team ranked last in pass attempts and third in rushing attempts, despite an average Game Script of -3.5.

YearLgTmPassRushP/RSCRIPTPASS IndexSCRIPT IndexPass Identity
2011NFLDEN46953746.6%-3.76584-51
1972NFLCHI20553630.7%-4.67085-46
2000NFLCIN45449550.6%-6.18274-45
1954NFLWAS25742741%-8.17877-45
1982NFLNWE18732438.4%-1.66393-44
2001NFLDAL41350547%-3.57185-44
1983NFLBAL37760141.4%-3.17385-43
2007NFLOAK45150849.4%-4.37681-42
1982NFLBAL28329350.8%-8.39663-41
1978NFLKAN37066337.1%-2.77386-41
2005NFLSFO38942850.8%-5.68576-40
1951NFLWAS22654730.4%-4.57784-39
2008NFLOAK42145950.1%-4.88179-39
1952NFLCRD28947740.7%-4.17487-39
1979NFLNYJ36963438.7%-1.87388-39
2009NFLCLE44349848.9%-3.27586-39
1945NFLPIT16535831.5%-6.68181-38
1984NFLIND41151047.9%-6.58676-38
1981NFLNOR44154646.9%-4.98777-36
1959NFLWAS28442243.3%-8.19371-36
1981NFLBAL47944153.9%-9.310956-36
2010NFLCAR48442855.5%-6.79471-35
2006NFLATL41653746.6%-1.67293-34
2009NFLTEN47649950.1%-3.17987-34
1978NFLCHI35263437.8%-1.97690-34
2011NFLJAX51247451.9%-3.58285-33
1965NFLPIT35440750.5%-6.210464-33
1971NFLNYJ27848538.3%-3.78186-33
1964NFLPIT32351642%-2.57889-33
1941NFLPIT16838130.6%-7.68879-33
1986NFLTAM45945553.1%-6.99869-33
2009NFLNYJ39360741.5%3.552115-33
2011NFLKAN53247652.8%-48583-33
2010NFLJAX46951250.2%-2.27790-33
2006NFLSFO44443852.6%-5.59176-32
1962NFLPIT31957238.8%-0.77197-32
1985NFLCLE41453345.8%-1.37593-32
1983NFLHOU48250251.4%-6.29969-32
1965AFLDEN48245352.8%-6.19475-32
1963NFLPIT36857840.8%-1.27395-32
1978NFLBAL38353244.8%-6.310267-31
1954NFLCRD34941847.6%-9.89672-31
1949AAFCNYY19951028.1%-1.47396-31
2001NFLWAS43249049.2%-2.17891-31
1985NFLATL46256048.7%-2.98485-30
1964AFLDEN45639156.9%-810168-30
1957NFLPHI20442435.4%-3.28684-30
1999NFLTAM44750249.3%-0.47298-30
2006NFLTEN44746950.4%-3.28486-30
1971NFLGNB25450035.2%-0.87397-30
1981NFLCHI48960846.3%-3.18585-30
1997NFLPIT46657245.9%0.369101-30
1988NFLNWE38958841.2%0.369102-30
2009NFLOAK48541056.9%-6.810070-30

If we look at just teams with positive game scripts, we get a more familiar look at the most run-heavy teams. The team with the strongest rushing identity? Rex Ryan’s ground and pound 2009 New York Jets:

YearLgTmPassRushP/RSCRIPTPASS IndexSCRIPT IndexPass Identity
2009NFLNYJ39360741.5%3.552115-33
1997NFLPIT46657245.9%0.369101-30
1988NFLNWE38958841.2%0.369102-30
1973NFLBUF21360528.7%0.768102-29
2004NFLPIT35861840.2%4.452119-29
1974NFLBUF25154534.3%1.564107-29
1998NFLTAM44952347.7%0.171100-28
1997NFLTEN42054145.5%1.168105-27
1968NFLMIN28250038.8%1.269104-27
1994NFLNYG40552546.2%0.372101-27
1981NFLKAN41061042.3%0.273101-26
1993NFLCHI38847747.8%0.173101-26
2003NFLBAL41555245.6%1.170105-25
1986NFLRAM40357842.7%1.668107-25
1984NFLRAM35854141.9%1.670106-24
1949NFLPIT20953529.3%0.874102-24
2010NFLKAN47555647.9%1.670107-23
1961NFLPIT33454340.5%0.874103-22
1970NFLDAL29752239.2%1.273105-22
1996NFLBUF48356348.5%0.675103-22
1957NFLWAS20150030.2%0.774104-22
1995NFLIND43447850.3%0.576102-22
2004NFLATL39552446.7%1.373106-22
1968AFLKAN27053735.4%6.460119-21
1994NFLIND37649544.9%2.168110-21
1984NFLCHI39067438.7%4.762117-21
2000NFLPIT43952748.2%1.174105-21
1960NFLSTL28548438.8%0.777103-21
2008NFLBAL43359244.9%3.365114-20
1967AFLHOU33247642.5%2.771109-20
1975NFLMIA27959433.7%4.266114-20
1980NFLCHI40457943%0.478102-20
2009NFLCAR46552549.1%1.176105-20

References

References
1 Note that for years since 2000, I removed spikes and kneels, but for years prior to 2000, I used the more familiar data available at PFR.
{ 7 comments }