Why do teams run the ball?
Posted by Chase Stuart on February 12, 2008
Note: It was pointed out in the comments that a similar discussion has been going on over at bbnflstats.com. Thanks to Brian, in comment 7, for pointing that out. His posts are well worth reading if you've got the time.
Everyone knows that teams average more yards per pass than per rush. In the 2007 season, NFL QBs averaged 6.85 yards per pass attempt, and NFL RBs averaged 4.17 yards per rush attempt. This past season wasn't an aberration.
YR RUSH RYD Y/R PASS PYD Y/P Diff 2002 12,016 49,984 4.16 17,087 115,054 6.73 +62% 2003 12,698 53,345 4.20 16,330 108,403 6.64 +58% 2004 12,654 53,028 4.19 16,288 114,979 7.06 +68% 2005 12,726 51,822 4.07 16,411 111,468 6.79 +67% 2006 12,734 53,413 4.19 16,358 112,092 6.85 +63% 2007 12,414 51,786 4.17 17,018 116,580 6.85 +64%
So why do teams run the ball? The most common explanation I've heard is that rushing plays are more consistent, and the larger variance that comes with a passing play makes passing plays less attractive. I don't want to minimize the value of consistency. If a team could run for three yards every play (average yards per play = 3.0), that team would score a touchdown every drive. If a team passed for 10 yards on 50% of its plays, and 0 yards on the other half (average yards per play = 5.0), that offense would score less points; that team would have to punt a non-zero amount of times. The ability to consistently gain yardage is crucial in the NFL.
But I don't think that really explains why teams run as often as they do, considering that pass plays average over 60% more yards per play. After all, rushing plays are a lot less consistent and passing plays are more consistent in real life than in the hypothetical example given. A rush offense that goes for 2 yards, 3 yards and 4 yards a pop an equal number of times is a lot less valuable than the one in the previous paragraph, and more closely resembles the real world. Would you rather have rushing plays of 2, 4, 6 or passing plays of 10, 10, 0? The latter, of course. In the passing group, a team has just a one in twenty-seven chance of being forced into a fourth down situation; in the rushing group, a team has a four-in-twenty-seven chance of being in a fourth down situation (runs of 2, 2, 2; 2, 4, 2; 2, 2, 4; and 4, 2, 2). And of course, the passing group averages 6.7 yards per play while the rushing group averages just four yards per play.
So there has to be more to this run-calling business than consistency. It turns out, there is.
Let's start with yards per pass -- what are we forgetting to calculate? QBs throws interceptions (which are bad) and touchdowns (which are good). Well, adjusted yards per pass solves that. We subtract 45 yards for every interception and add 10 yards for every TD thrown. Here's a table of the league average adjusted yards per attempt by QBs for every year since the merger:
2007 5.85 2006 5.85 2005 5.79 2004 6.07 2003 5.57 2002 5.75 2001 5.65 2000 5.68 1999 5.63 1998 5.79 1997 5.71 1996 5.54 1995 5.79 1994 5.74 1993 5.59 1992 5.51 1991 5.69 1990 5.83 1989 5.81 1988 5.54 1987 5.73 1986 5.59 1985 5.58 1984 5.73 1983 5.64 1982 5.47 1981 5.51 1980 5.38 1979 5.20 1978 4.69 1977 4.31 1976 4.86 1975 4.69 1974 4.52 1973 4.59 1972 4.85 1971 4.51 1970 4.81
That brings things a bit closer together, but we're still pretty far off. After factoring in interceptions (and touchdowns) thrown, passing plays still average in the high fives and rushing plays in the low fours. What else are we missing? When you pass the ball, you can get sacked. That doesn't happen on running plays. If we add sack yards lost to the numerator and sacks to the denominator, we get adjusted net yards, and adjusted net yards per attempt can be calculated. This closes the gap considerably.
Last season was the second best passing season ever. QBs attempted 17,009 passes and were sacked 1,096 times (18,105 plays), and recorded 116,490 passing yards, 715 TDs, 534 interceptions and lost 7,133 yards due to sacks (92,477 adjusted, net yards). That 5.11 yards per pass play by QBs trailed only 2004 as the most prolific passing season ever.
2007 5.11 2006 5.02 2005 4.98 2004 5.23 2003 4.83 2002 4.97 2001 4.82 2000 4.85 1999 4.81 1998 4.91 1997 4.79 1996 4.77 1995 5.04 1994 5.02 1993 4.78 1992 4.53 1991 4.86 1990 4.89 1989 4.88 1988 4.69 1987 4.71 1986 4.59 1985 4.47 1984 4.63 1983 4.56 1982 4.38 1981 4.58 1980 4.45 1979 4.20 1978 3.66 1977 3.17 1976 3.74 1975 3.58 1974 3.55 1973 3.41 1972 3.78 1971 3.47 1970 3.76
Quarterbacks also fumble the ball, but I'm going to put the discussion of fumbles off for another day (hopefully one day soon). For reasons to be explained later, I think it's in our best interest to ignore them.
So now we have a true measure of what happens on a pass. We don't just look at the yards, we look at the interceptions (and touchdowns), the sacks, and the sack yardage lost. But we can't compare that to rushing yards just yet -- because on a running play, you can score a TD, too. Since we looked at passing touchdowns, we can't ignore rushing touchdowns. (Since we ignored QB fumbles, we will ignore RB fumbles as well. It turns out, the rates are pretty similar.)
So if we really want to compare what happens every time a QB goes back to pass to what happens every time a QB hands the ball off to a RB, we need to add in rushing touchdowns to yards per rush. If we add ten yards to each RB's rushing yards total for every rushing TD, we'll get adjusted yards per rush. Here's a look at the historical numbers:
year rb rsh rb ryd rb ypc rb td rb aypc 2007 12012 50153 4.18 328 4.45 2006 12236 51464 4.21 365 4.50 2005 12407 50507 4.07 372 4.37 2004 12287 51637 4.20 370 4.50 2003 12219 51340 4.20 345 4.48 2002 11746 49044 4.18 360 4.48 2001 11242 45698 4.06 275 4.31 2000 11059 44302 4.01 325 4.30 1999 11405 44489 3.90 289 4.15 1998 11669 46546 3.99 314 4.26 1997 11535 46480 4.03 318 4.31 1996 11539 45117 3.91 302 4.17 1995 11282 44886 3.98 318 4.26 1994 10293 38759 3.77 267 4.02 1993 10344 40195 3.89 246 4.12 1992 10083 40698 4.04 280 4.31 1991 10104 39679 3.93 293 4.22 1990 10249 41586 4.06 313 4.36 1989 10866 42423 3.90 315 4.19 1988 11291 45090 3.99 339 4.29 1987 11057 43173 3.90 285 4.16 1986 11435 44996 3.93 343 4.23 1985 11767 49017 4.17 374 4.48 1984 12242 49712 4.06 351 4.35 1983 12553 51373 4.09 380 4.40 1982 6866 26348 3.84 200 4.13 1981 12798 51628 4.03 377 4.33 1980 12565 49785 3.96 357 4.25 1979 13414 53755 4.01 423 4.32 1978 14596 57859 3.96 404 4.24 1977 13147 50448 3.84 287 4.06 1976 13115 53004 4.04 342 4.30 1975 12022 47265 3.93 361 4.23 1974 11062 42423 3.84 300 4.11 1973 11156 45045 4.04 261 4.27 1972 10758 43721 4.06 290 4.33 1971 10015 39698 3.96 263 4.23 1970 9618 35852 3.73 250 3.99
The first thing that jumps off to me is that 1985 was a highwater mark for RBs -- the 4.17 ypc average was higher than any season from 1970-2001, and the same is true for the league-wide adjusted yards per rush. Gerald Riggs had a huge year, and Marcus Allen, Walter Payton, Tony Dorsett and Earl Campbell were all running strong. James Wilder, Freeman McNeil, Joe Morris and Curt Warner all had big seasons, too. It just seemed to be a perfect storm of young and old RBs playing at elite levels.
When looking at the RB data, there's a general increase in rushing efficiency each year. The correlation coefficient between "year" and "rb ypc" is very close to 0.50, so we can feel comfortable in noting that running games are improving slightly each season. The addition of the touchdown bonus to the numerator boosts up league-wide running back yards per rush by about seven percent. But now, let's compare the RB adjusted rushing yards per rush to QB adjusting passing yards per play:
year rb aypc qb nay/a diff 2007 4.45 5.11 0.66 2006 4.50 5.02 0.51 2005 4.37 4.98 0.60 2004 4.50 5.23 0.72 2003 4.48 4.83 0.35 2002 4.48 4.97 0.49 2001 4.31 4.82 0.52 2000 4.30 4.85 0.55 1999 4.15 4.81 0.65 1998 4.26 4.91 0.66 1997 4.31 4.79 0.49 1996 4.17 4.77 0.60 1995 4.26 5.04 0.78 1994 4.02 5.02 0.99 1993 4.12 4.78 0.66 1992 4.31 4.53 0.22 1991 4.22 4.86 0.64 1990 4.36 4.89 0.53 1989 4.19 4.88 0.69 1988 4.29 4.69 0.40 1987 4.16 4.71 0.55 1986 4.23 4.59 0.36 1985 4.48 4.47 -0.01 1984 4.35 4.63 0.28 1983 4.40 4.56 0.16 1982 4.13 4.38 0.25 1981 4.33 4.58 0.25 1980 4.25 4.45 0.20 1979 4.32 4.20 -0.12 1978 4.24 3.66 -0.58 1977 4.06 3.17 -0.88 1976 4.30 3.74 -0.56 1975 4.23 3.58 -0.66 1974 4.11 3.55 -0.55 1973 4.27 3.41 -0.86 1972 4.33 3.78 -0.56 1971 4.23 3.47 -0.76 1970 3.99 3.76 -0.22
As you can see, for most of the '70s, running the ball actually was superior to passing the ball. Had you just compared yards per pass to yards per rush, here's what you would have seen:
year rb ypc qb ypa diff 2007 4.45 6.85 2.40 2006 4.50 6.85 2.35 2005 4.37 6.79 2.42 2004 4.50 7.06 2.55 2003 4.48 6.63 2.15 2002 4.48 6.72 2.24 2001 4.31 6.76 2.46 2000 4.30 6.74 2.44 1999 4.15 6.75 2.60 1998 4.26 6.85 2.59 1997 4.31 6.68 2.38 1996 4.17 6.68 2.51 1995 4.26 6.77 2.51 1994 4.02 6.77 2.74 1993 4.12 6.69 2.56 1992 4.31 6.86 2.55 1991 4.22 6.89 2.68 1990 4.36 7.00 2.64 1989 4.19 7.14 2.95 1988 4.29 6.94 2.64 1987 4.16 6.99 2.83 1986 4.23 7.02 2.78 1985 4.48 7.06 2.58 1984 4.35 7.12 2.77 1983 4.40 7.15 2.76 1982 4.13 6.98 2.85 1981 4.33 7.00 2.67 1980 4.25 6.98 2.74 1979 4.32 6.84 2.51 1978 4.24 6.70 2.46 1977 4.06 6.51 2.45 1976 4.30 6.67 2.37 1975 4.23 6.67 2.44 1974 4.11 6.48 2.37 1973 4.27 6.51 2.24 1972 4.33 6.76 2.43 1971 4.23 6.69 2.46 1970 3.99 6.71 2.72
Pretty interesting, eh? The difference in yards per rush and yards per pass was the exact same in the early '70s as it is now! Yet take a look at pass attempts and rush attempts in the modern era:
year pass rush ratio 2007 17009 12012 1.42 2006 16355 12236 1.34 2005 16429 12407 1.32 2004 16299 12287 1.33 2003 16438 12219 1.35 2002 17239 11746 1.47 2001 16137 11242 1.44 2000 16294 11059 1.47 1999 16726 11405 1.47 1998 15453 11669 1.32 1997 15708 11535 1.36 1996 15939 11539 1.38 1995 16667 11282 1.48 1994 15032 10293 1.46 1993 14384 10344 1.39 1992 13341 10083 1.32 1991 13602 10104 1.35 1990 13481 10249 1.32 1989 14176 10866 1.30 1988 13773 11291 1.22 1987 12747 11057 1.15 1986 13835 11435 1.21 1985 13687 11767 1.16 1984 13753 12242 1.12 1983 13422 12553 1.07 1982 7595 6866 1.11 1981 13643 12798 1.07 1980 13218 12565 1.05 1979 12765 13414 0.95 1978 11594 14596 0.79 1977 9412 13147 0.72 1976 9602 13115 0.73 1975 9359 12022 0.78 1974 9159 11062 0.83 1973 8200 11156 0.74 1972 8450 10758 0.79 1971 8944 10015 0.89 1970 9415 9618 0.98
In the early '70s, teams passed about 80% as often as they ran the ball. Now, the average team passes about 40% more often than they run the ball. Meanwhile, yards per attempt and yards per rush haven't changed very much. But considering how the adjusted numbers show the pass/run effectiveness gap increase significantly is very strong evidence that the adjusted data is what matters.
In the early '70s, rushing was indeed superior to throwing. Teams averaged about two feet more per rush than per pass, when factoring in sacks and interceptions. That's why the Dolphins with Larry Csonka, Mercury Morris and Jim Kiick were able to dominate the early '70s, and the Steelers with Franco Harris and Rocky Bleier were dominant in the mid '70s. The running game was paramount then, because passing wasn't a very attractive option. If you could run and stop the run, you could win. And that's why we probably still hear that phrase today.
But then the rule changes of 1978 kicked in. Defenders were no longer allowed to mug receivers. The Mel Blount rule prevented defenders from making contact with receivers once the offensive player was five yards past the line of scrimmage. Of equal importance, offensive lineman were now allowed to extend their arms and open their hands to pass block. Can you imagine what passing offenses would look like today if those rule changes never happened? We could certainly imagine a large bump in sacks allowed and interceptions. Take a look:
year sk/att int/att 2007 6.4 3.1 2006 7.1 3.2 2005 7.2 3.1 2004 7.3 3.2 2003 6.6 3.2 2002 6.8 3.0 2001 7.4 3.3 2000 7.6 3.2 1999 7.4 3.4 1998 7.8 3.3 1997 8.0 3.0 1996 6.9 3.4 1995 6.4 3.0 1994 6.2 3.1 1993 7.3 3.2 1992 8.5 3.9 1991 7.1 3.5 1990 7.9 3.5 1989 7.6 3.9 1988 7.4 3.9 1987 8.1 3.9 1986 8.4 4.0 1985 9.2 4.2 1984 9.1 4.0 1983 8.7 4.4 1982 8.4 4.3 1981 7.2 4.3 1980 7.5 4.5 1979 8.1 4.6 1978 8.6 5.4 1977 9.7 5.8 1976 10.0 4.8 1975 9.3 5.4 1974 8.4 5.2 1973 9.8 5.4 1972 8.7 5.3 1971 8.2 5.9 1970 8.9 5.1
As recently as 1985, rushing the ball was more effective than passing the ball. As recently as 2003, the difference in true yards per rush and pass was just 0.35 per attempt. The two rules enacted in 1978 severely diminished the impact on the two biggest negatives associated with passing the ball -- sacks and interceptions. In 1978, the average pass play netted just 3.66 yards per pass; three years later, the average pass play was worth 4.58 yards per pass. Teams passed 80% as often in 1978 as they ran; by '81, teams passed 7% more often than they ran.
The rule changes of 1978 answered the question: Why do teams pass the ball? There's a lot of game theory involved in the decision to run or pass, but it's clear that running was a more efficient option and had a lower variance. Now, running is less efficient (but with still a lower variance). Looking at true yards per pass overstates the passing option by about 55% from 1970-2007, and by about 37% over the past ten years. Combined with how not counting rushing touchdowns in yards per rush (understating the average run play by about 7%), and you can see, finally, why teams run the ball. Once you include the lower variance, the only question left is why don't teams run the ball more often?
Thanks for this very interesting post! I would add a time component to this as well. Winning teams want to use more time and losing teams want to use less. On a large scale this probably cancels out over a season. What may not cancel out is that the winning team is willing to take less yards (the defense knows they are going to run the ball) to run the clock down. It would be interesting to see the breakdown by quarter, I have to believe that the average yards/rush drops in the 4th quarter compared to the average. Teams do get into known passing situations but they are trying for more yards not less and the defense is more willing to allow short passes. Also, penalties may be a small factor. More holding penalties and more false starts would be my guess.
I would also add that teams run more in crappy weather usually for less yards.
Why don't teams run the ball more often? Your analysis factors in many things, but not the clock. Passing can get you toward the goal line faster, it's easier to work the sidelines, and incompletions stop the clock. This often comes into play toward the end of each half and beyond that too for teams with a sizeable deficit to overcome.
Oh, and you're more likely to pick up a very handy defensive penalty on a pass play.
The other question that you might want to consider is how successful is the pass on obvious passing downs. If a team cut their running game out and only passed because it was more successful the line could just pin their ears back and rush all the time making it harder on the QB, teams would be in nickel or dime formation all the time making it harder for any of the WR's to get open. This is why I'm a fan of an offense that runs enough to keep them honest and break off a big one when they're napping.
As I alluded to at the end of the post, a lot of the strategy involving in play-calling can be analyzed using game theory. When a defense is playing the pass, I'd suspect that rushing is more effective than passing. When a defense is playing the run, I'd suspect that passing is more effective than rushing.
For example, on 3rd and 7 or longer over the past five seasons:
NFL RBs rushed 1,756 times for 9,675 yards and 25 TDs, for an average of 5.65 adjusted yards per rush.
NFL QBs attempted 15,979 passes that went for 114,282 passing yards, 518 TDs and 656 INTs. I don't have sack data separated out on a play by play basis, but I'd guess that QBs are sacked a bit more often on these sort of plays than your average pass play. If we assume a sack for every 8 attempts, that would mean QBs were sacked 1,997 times. We could guess that QBs lost about 14,000 yards on those sacks.
That would mean QBs totaled 75,942 net adjusted yards on 17,976 plays, for an average of 4.22 yards per pass play.
On average, you'll gain more yards (after factoring in the penalties for INTs and sacks) running the ball on 3rd and very long than passing. But your odds of actually getting the first down on those situations are better when passing. And interceptions don't matter too much on 3rd and long, especially if it's on a pass far down the field.
I agree that looking at penalty splits on rush/pass plays would be another key ingredient to get a thorough analysis at the effectiveness of each method. Maybe one day. 🙂
Chase-Did you just read my latest posts? Except for the history lesson, it seems like most of your points come straight from here:
https://www.bbnflstats.com/2008/02/passing-paradox-part-1.html
https://www.bbnflstats.com/2008/02/passing-paradox-part-2.html
https://www.bbnflstats.com/2008/02/passing-paradox-part-3.html
Admittedly, no one owns these ideas, but a hat tip is customary, even if someone just sparks your interest.
-Brian
My first comment was going to be that clock control is also probably a factor, but a few posters have already noted this. However, I didn't notice anybody noting that teams might not just want to control the clock for "pure" strategy (e.g., securing the last possession of a half), but also to wear down the defense. I know it's a cliche, but the longer they're out there, the more fatigued they'll get.
Second, my guess is that a teams' propensity to pass and its yards per attempt are not independent. For example, let's say pass plays gain 10-10-0, and running plays gain 0-5-4. Passing every play would seem to be the optimal strategy, but those averages are being calculated with the probability of the offense choosing "run" or "pass" being approximately equal. As a result, it reflects defenses that are preparing equally for run or pass. However, if teams started choosing the more effective option exclusively, defenses would adjust, and the per-play averages would change. So part of the answer is probably that teams run to make the pass more effective.
Finally, part of the answer is that coaching decisions are not always made based on pure rationality. Coaches' reluctance to go for it on fourth down is probably the most well-known example of this.
That said, I'm well aware that a model is supposed to be a SIMPLIFICATION of reality -- it's not necessarily supposed to take all these things into account -- and I thought this was a very cool one.
A logically parallel discussion of this same subject just ran through the last several posts and comments at https://www.bbnflstats.com. That went on to consider two interesting papers by academics that cite team-by-team data but use differing methodology to reach opposite conclusions: one that teams pass too much, the other that teams run too much.
I'll just make a couple observations here:
The minor one is that a rushing yardage adjustment of +10yds per TD seems dubious. Pete Palmer, who originally derived it for rating QBs by AY/A attempt, explained it as reflecting the fact that it is tougher to pass near the goal line, so % completion rate and average-yard-per catch drops. The adjustment equalizes QB efficiency as measured by AY/A wherever the QB is playing on the field (to not penalize the rating of a QB for just being near the goal line). The 10 yards is not a "bonus for scoring."
To do the same thing with rushing, one would determine average yard per carry for TD runs compared to average yards per carry otherwise, and increase the TD-run average by the difference. That'd be a lot less than 10 yards, more like one yard, I'd guess.
And that leaves passing still producing significantly more average yards per play than rushing, as it *must*. The danger in using a high variance strategy that produces higher average return in the long run -- passing v running -- is that the down side of high variance may kill you in short run, so you never get the higher return in the long run. A football game is short-run. E.g.: A QB as good as Peyton Manning can throw six picks in a game. A run-and-shoot team that dominated the league's passing stats all year can blow a 32-pt lead in the second half of a playoff game.
To compensate for that risk it is not near enough to subtract yards lost on sacks, picks, etc, from AY/PA to try to equalize it to AY/R. Different methodology is needed.
This is a well known problem in econ and finance and the answer is "portfolio theory". The optimal run/pass mix varies by team (its abilities) time (riskier is safer early) and situation (down and distance, score) etc., to produce the best max-return/minimum risk portfolio. The best portfolio is measured in victories, not average yards per play. To get victory, many specific situations will be highly biased towards the pass or the run, regardless of the "adjusted average yards per play" differnce between them. That's why teams run so little, or run more, or run the wrong amount.
There's team-by-team data on this charted at bbnflstats, and the other two papers with team analysis are reachable through it. BTW, Michael Vick's Falcons were a heck of an outlier! .;-)
Actually, this post (below) is probably most similar to yours here. The other 3 are an extension of the points here:
https://www.bbnflstats.com/2008/01/passing-premium.html
If I'm mistaken, then I apologize. You would feel the same way.
If my prior comment was gobbledygood, it boils down to this: Game situations tend to dominate "average yards per run-or-pass play" when choosing between them.
E.g,: a team may run for 4.5 yards each and every play, no more no less, as an unstoppable scoring machine, but pass for only 3 yards an attempt (think "Navy"). completing 33% of its attempts. Then on 3rd and 1 it gets penalized 15 yards. Although on an "average yards per play" basis running always dominates passing for it, it's gotta pass. A modest chance of getting a first down is better than 0% chance. There's a bias towards passing in the situation.
That's an extreme stylized example, but in reality such biases exist near all the time. Down-and-distance, score, time on clock, need to rest players, tactical opportunities, etc, all throughout a game drive teams to choose between running and passing in a way that can be very different than the choice they make if simply trying to maximize yards per play. Since teams have many goals other than getting yardage, if their final play choices between running and passing look odd in term of not maximizing net yards per play, it's maybe not so odd.
This is very interesting, but you're missing something really major - decisions are optimal at the margin not at the average. This is a point most economists know, but is not widely known or applied by other disciplines. Basically, if a team is rushing/passing at a 1:1 ratio, they should consider the value of one additional pass attempt, not the average value of all the pass attempts that made their ratio 1:1.
Thus, we expect the MARGINAL rush to equal the MARGINAL pass attempt. And there is no particular reason that we should expect the average rush to equal the average pass attempt.
It helps to consider an extreme case - where a team rushes on EVERY play. This is almost definitely not optimal, because then the defense knows you are going to run and put 9-10 players in the box, and thus the marginal value of a pass is enormous (throw a bomb, get a ton of yards). However, as you decrease the amount of runs and increase the number of pass attempts, the marginal value of passing goes down and the marginal value of rushing increases. It is easy to see that the optimal is where these values are equal. However, in our example, it is also clear that the average pass is inflated by the huge value of that very first pass which will go for 50+ yards almost every time.
Statistically, it's a much harder exercise to try to calculate that marginal value of rushing or passing, but it is definitely the theoretically correct approach.
Hey Brian,
I've seen the work of my colleagues over at footballguys.com plagiarized, and I've got no respect for people that steal other's work. I'm not a visitor of bbnflstats except when you post a link in the comments; sadly there's just not enough time in the day to read everything we'd like to read.
I'm sorry for causing you any grief over this, and I hope you'll understand that I never read any of the four articles you linked to here before today. I think they're all very good, and I like part-3 of the passing paradox the best. You're a good commenter here, and I always look forward to reading your stuff.
I put a note at the top to let everyone know of your work.
Take care,
Chase
Chase-My apologies then. I appreciate your response. With all the similarities and the timing between our articles, I hope you understand my reaction. "Great minds" I guess 🙂
Although I don't think it is that easy to measure, not all passes are created equal. Short passes I am guessing don't really lead to too many penalties and/or sacks. After watching Brady pass to Welker time after time, it made me think that not all pass plays are created equal. I can think of at least 5+ kinds, quick passes to a wr (basically a long pitch!), short passes to a wr (short out or a slant), a medium pass to a TE or WR (just over the linebackers, in front of the safeties), long pass (trying to get past the safeties), and like passes to a RB which take more time (I think a short pass to a RB takes about as much time as a medium pass to a TE for example). You could also add drag type patterns which take more time compared to other passes of like yardage. Like I said, I am not sure how it would affect your analysis! 🙂
Andy, I agree with a lot of your thoughts.
I don't think I was trying to prove too much. I was mostly focused on how the passing game has really become much more effective relative to the running game. What's most interesting to me is that:
1) Running was more effective when factoring in sacks and INTs than passing in the early '70s. Adjusted yards per rush was greater than adjusted yards per pass. That's really noteworthy, since yards per rush was way lower than yards per pass in the '70s. The adjustment really matters, and...
2) Teams ran a lot more often in the '70s. In fact, the correlation coefficient between the difference between adjusted net yards per pass and adjusted yards per rush was 0.94 with the ratio of passes to rushes. This means, I think, that teams really do care about net adjusted yards per pass and adjusted yards per rush. That's really cool.
Jim G.,
I think the ten yard bonus is appropriate. Remember this post?
The ten yards is given to a TD not because it's difficult to pass near the end zone, but it's a good approximation of the marginal value of gaining a yard from the one yard-line, relative to gaining a yard elsewhere on the field. And in theory, it's the same for rushing and passing (and all types of) TDs.
Not sure how this affects the analysis, but the distinction between passing and rushing plays is a little strange these days, with the prevalence of screens and short dump-offs. I don't really think of most of those as passing plays - they're more like pitch-outs or delayed draws. If we counted these ball-control passes as runs, I suspect the 70s figures (running somewhat better than passing) would still hold.
I think the question "Why don't teams pass more?" in this day and age is--"they run to open up the pass." Look at this past SB. Brady had 48! attempts--yet they were never behind by more than 3 points (i.e., the score never forced them to go pass-wacky except for the last-second drive at the end). The game was always in reach for NE; they never needed to abandon the rush. Yet because the Giants knew that the Pats pass, pass, and pass again, they were able to counter that strategy and win the game. They held maybe the best offense ever to 14 points; the weather did not help them, nor did an injury to a key player (maybe it hindered TB a little, but Matt Cassell wasn't out there). They just played great D against the pass AND the run.
And the last sentence is the key. Teams passed on the Vikes all year because they couldn't run effectively against them. But against great D's like the 85 Bears and 00 Ravens, it doesn't matter. Against most teams, you need to keep a balance because NFL D's are good enough to stop ONE thing if you do that SAME thing over and over, but if you keep them guessing (around a 50/50 pass/run ratio) it becomes MUCH harder to play defense.
To sum it up--run to set up the pass, pass to set up the run. Keep the D guessing, and the whole offense does better. This is why Elway finally won 2 SB's--the D had to play the run honestly with Davis in the backfield, thus allowing Elway to pass even MORE effectively.
Where did you get the sack data? I thought sacks weren't officially recorded until the late 70's/early 80's.
Also, I wonder what red-zone playcalling was like in the 70s before the Mel Blount rule. It might affect those adjusted numbers some.
Derek, sacks weren't an official stat for individuals until 1982, but they were tracked at the team level starting in I think 1950ish.
Defenses still play the run and the pass differently, if teams passed 80% of the time, defenses would specialize and nullify the perceived advantage.
If the majority of pro football teams went to a 4 wide receiver, one back set and passed like mad, several things would happen.
1) Talent would be spread thin. Passers, receivers, and pass blockers would be scattered throughout the league with most teams lacking one of those essential elements.
2) Defenses would forget about run stopping and be able to focus solely on pass rush and coverage. An increase in sacks and interceptions would be sure to result.
3) The running backs that remained would have a high YPC running draws and pitch-outs which would lead people to ask: "Why don't they run more?"
The program I wrote to analyze how QBs perform in different situations can actually analyze any of the box score stats. I had it curve fit rush att/(pass att + rush att) split into 5 categories for two periods. Pre-1978 and post 1978, with the splits being down 8+, down 1-7, tied, up 1-7, and up 8+.
Pre-1978, rush% was around 42% when losing by more than a TD, up to 66% when winning by more than a TD
Post 1978, rush% was around 32% when losing by more than a TD, up to about 60% when leading by more than a TD.
So the Mel Blount rule appears to have had a major effect, and score appears to have a major effect as well.
Graph here:
https://www.flickr.com/photos/mattie_shoes/3744528855/