Saturday, April 03, 2010

Passing, rushing, and winning in the NFL since 1969

In January, after having a little fun with team stats for the 2009 NFL season, I looked at the correlations between success (wins) and per-play passing and per-play rushing throughout the early years of this century. The trend is clear--a strong passing game does a top-tier team make. Running is far less determinative.

But rule changes and a paradigmatic shift away from raw power to pure speed (the fullback, a linchpin position as recently as Daryl "Moose" Johnston of the nineties' Cowboys, is all but dead today) have increasingly led to the predominance of passing over running. This has not always been the case.

That has been my conception, anway. To see how well it meshes with reality, I ran the numbers extending all the way back to the 1969 season. The following table traces the last four decades, showing the correlations between winning and both passing yards per attempt and running yards per attempt by regular season:

YearPassRun
2009.80

.09*

2008.48.15*
2007.76.24*
2006.44.10*
2005.60.40
2004.56.45
2003.67.07*
2002.50.11*
2001.39.13*
2000.47.51
1999.53.12*
1998.75.32*
1997.37.27*
1996.40(.06)*
1995.46.28*
1994.55(.05)*
1993.55.34*
1992.70.08*
1991.63.33*
1990.56.05*
1989.64.04*
1988.56.16*
1987.52.15*
1986.49.03*
1985.46.13*
1984.57.09*
1983.32.42
1982.51.22*
1981.21*(.07)*
1980.58.19*
1979.54.11*
1978.72.09*
1977.63.30*
1976.70.29*
1975.77.34*
1974.60.16*
1973.66.47
1972.50.51
1971.34*.12*
1970.57.24*
1969.48(.24)*

* Not statistically significant at 95% confidence

The data dispute the preceding narrative. There is a lot of variation in the strength of the correlations from season to season, as is of course expected given that defense, special teams, penalties, schedules, etc are not taken into account. Still, passing has consistently been tied more tightly to success than running has been. In only three of the last 41 seasons has the relationship between rushing yards per attempt been more closely related to winning than passing yards per attempt has, and in each of those years--1972, 1983, 2000--running's advantage has been small.

The 2009 season, deemed the year only passing mattered, added an exclamation point to this observation. Over the last three decades, passing's importance has risen marginally, correlating with wins at .57 on average durings the 00s, from .52 in the eighties (.55 in the nineties)--but in the seventies, the average was .62. It's nothing new.

If, despite the reliably delivered claims coming from professional commentators during the playoffs to the contrary, passing matters most, why isn't the pass-to-run ratio more lopsided than it is?

Turnovers are one reason. In recent years, interceptions constitute about two-thirds of them, and some portion of the remaining one-third, coming in the form of fumbles, are lost by quarterbacks getting hit or receivers losing the ball after the catch. In 2009, a team's turnover balance (takeaways minus giveaways) correlated with games won at .69.

A one-dimensional approach putatively gives a big advantage to defenses as well--color commentators regularly describe running plays mid- and late-game as attempts by offenses to keep the opposing defenses "honest".

That said, I do wonder if, over time, there has been a shift toward more yards earned through the air relative to those on the ground. If passing's primacy is nothing new, maybe throwing the ball more is. Yet the diminution of the 1,000 yard rusher and the new benchmark of 2,000 yards in a single season suggests more, not less, yardage is being earned on the ground than before. I'll see what the history shows and present here.

9 comments:

Steve Sailer said...

Pro Football Reference has a couple of statistics that incorporate interceptions and touchdowns into passing yards to give an adjusted yards per attempt stat.

In general, interceptions are down since 1969 as a percentage of passes attempted. Guys like Joe Namath would heave the ball downfield and see what happened.

Steve Sailer said...

One question is whether yards per attempt is equally subject to diminishing returns per number of attempts per game.

Say a team throws the ball 20 times for 200 yards or 40 times for 400 yards. The latter is more impressive, but mostly because it's a larger sample size over which 10 yards per pass attempt is maintained, and because the defense gets more chance to set up for the passing game. It's not like the quarterback's arm gets tired throwing 40 times per game instead of 20 times. (Granted, he gets beat up by the pass rush.)

On the other hand, compare a team that rushes 20 times for 100 yards or 40 times for 200 yards. Carrying 40 times for 200 yards seems more impressive than 20 carries for 100 yards for three reasons: first, larger sample size, second, defensive adjustment, and third, wear and tear on the running backs.

This might explain in part the low correlations between yards per carry and victories -- successful NFL teams might run the ball enough that diminishing marginal returns on yards per carry set in with a vengeance because they are wearing down their best backs.

Top Arguments said...

I really dig this stuff, eager for more.

Audacious Epigone said...

Steve,

I wonder if that is manifest in the pass:run ratio of total yards from scrimmage, or if picking up lots of yardage on the ground facilitates picking up lots of yards through the air as well. If the later is the case, I'm not sure how to detect it statistically, other than looking at total rushing yards. But that doesn't reveal much--it seems like a safe bet that teams that average 200 yards on the ground do better than teams that only average 50 yards per game, even without taking other stats into consideration.

Audacious Epigone said...

TA,

I'm glad you find it to be interesting.

Steve Sailer said...

I'm starting to think that the low correlations between yards per carry and victories are kind of inevitable, perhaps due to differences in clock rules.

Say you are up by 21 points, so you decide to grind out a lot of runs to use up the clock. The defense knows you are going to run, so your per carry average is low.

In contrast, say you are down 21 points, so you need to pass. The defense plays deep to avoid giving up a quick touchdown pass, so anytime you rush the ball, you pick up some easy yards, so your per carry average is high.

Most of the time, the team with the big lead winds up winning, but it does so in a way that lessens the correlation between yards per carry and victory.

Steve Sailer said...

My guess is that yards rushing per game has a higher correlation with winning than yards per rushing attempt, at least relative to passing measures.

I don't know about the NFL, but it used to be that in college in the 1970s, a 300 yard passing game was usually the sign of a loser. For example, when I was at Rice in 1976, Tommy Kramer led the country in passing yardage while the Rice Owls went 3-8.

Steve Sailer said...

I think _that's_ a big change -- that throwing for a lot of yards used to be the sign of a team that is going to lose and now it's the sign of a team that is going to win. But I'm mostly thinking about college football, so I don't know about the NFL.

Audacious Epigone said...

Steve,

Great stuff. If that is the case, we should see that a high pass:run total yards ratio was a sign of a lousy team, but over time it has increasingly come to mean the opposite. It's an empirical question that the data might answer.