Posted by Chase Stuart on February 8, 2008
If you've got a few minutes to spare, check out Part I of this series, although it's unrelated to today's post. Thanks to JWL, who pointed out the absence of J.J. Jones in that post, which has since been updated. If you notice any other absences from that list, please post their names in the comments. (I'm not considering players like Antwaan Randle El or Brad Smith as QBs, since they are really WRs who just run some trick plays.)
Pro-Football-Reference introduced sacks and sack yards lost as part of the new database. You can check out which QB has been sacked the most number of times in a single season or career. To date, this blog hasn't used the sack data too often, so I'm going to make an effort to incorporate this new and valuable information a bit more frequently into my posts.
How should we use sacks data when ranking QBs? Well, an incompletion is better than a sack, right? Therefore, a sack should count against completion percentage -- it was a pass attempt that failed. Further, the yards must be taken away from a QB's gross passing yards. So a QB that is 30 for 40, for 300 yards, with 5 sacks for 50 yards would have his stats translated to 30 for 45 for 250 yards. Does that seem right or fair to you? I feel pretty comfortable with taking all sacks and counting them as 1) incompletions and 2) negative yards.
We can use the sack data to calculate net yards per attempt. It's computed by taking an individual QB's passing yards, subtracting his sack yards lost, and dividing by [pass attempts + sacks]. This ratio is already calculated for you at every QB's PFR page. Kurt Warner in 2000 has the highest net yards per attempt in a single season, with a 9.0 average (and averaged 9.9 passing yards per attempt without factoring in sacks).
Why all this buildup? Because I'm going to rank the best QB seasons by a black QB, and this gives me a good excuse to use our new data. Further, we need to incorporate the rushing yards in here somehow. Reader "Boogie Down Enigma" made an interesting observation in a comment to a blog post about Mike Vick, when he compared him to Tom Brady. Please note that this was before the 2007 season, when Brady actually had great statistics for the first time in his career.
When Vick ran, he was essentially throwing screen passes to himself. When you consider that, you see he was essentially catching 7.9 screen passes per game for 57.6 yards per game.
The average game saw Brady complete 20 passes on 32.3 attempts for 227 yards, 1.5 touchdowns and 0.8 interceptions.
The average game saw Mike Vick complete 13.9 passes on 25.8 passes for 171.7 yards, 1.1 touchdowns, and 0.8 interceptions.
The average game saw Tom Brady run 2.5 times for 4.58 yards.
The average game saw Mike Vick run 7.9 times for 57.6 yards.
When you combine their rushing totals and treat the rushing attempts as completions and attempts because that is essentially what they were (Vick completing screen passes to himself), you see that it is appropriate to perhaps consider Vick as a regular season equal to Brady in terms of offensive production from the QB position.
Combining the rushing totals, you get this:
In the average game for Brady, 22 completions on 35 attempts for a 65% completion percentage for 232 yards, 1.6 TDs, and 1.1 turnovers, 6.7 yards per attempt.
In the average game for Vick, 22 completions on 34 attempts, 65% completion percentage, 229 yards, 1.4 TDs, and 1.3 turnovers, 6.8 yards per attempt.
That's some pretty interesting analysis that coincidentally matches up the two QBs almost perfectly. I like it the idea, and think that's a good way to think about Vick. . . but I don't think that's a good way to think about all QBs. Over half of all Peyton Manning's rushing attempts this season were kneel downs. So I don't really want to penalize QBs for essentially a negative one yard pass every time their coach decides the QB should sit on the ball. I don't know if anyone has ever done a study to meaningfully compare a QB's rushing yards to his passing yards. That is, which is better: A QB that passes for 3,000 yards and runs for zero yards or one who throws for 2,000 and runs for 1,000? Are they the same?
Here's an idea I've got, which I've pretty much plucked out of nowhere: QB rushing yards over four yards per carry. Let's say a QB rushes 100 times for 500 yards; I'll add 100 yards to his passing yards total, because he gained an additional 100 yards over four yards per carry. If he rushed 200 times for 1,000 yards, he'll get 200 yards added to his passing yards total. The thinking here is that on a running play, an average team can pick up four yards; the QB only gets credit for the additional yards they get. What I like about this, is that QBs that don't run aren't penalized.
How does this idea work in practice? Remember the Best QB of All Time post? (Which needs to be updated; it's on my to-do list). Well I'm going to use that sort of logic. Here's the thinking:
1) Start with a QB's adjusted net yards per attempt. Adjusted yards per attempt, you might remember, is calculated by adding 10 yards for every TD pass and subtracting 45 yards for every interception, and dividing that number by pass attempts. To calculate adjusted net yards per attempt, we'll subtract sack yards lost from the numerator, and divide by [pass attempts plus sacks].
2) Now we want to add in rushing yards over 4.0, too. I don't think we want to call this net adjusted net yards per attempt; let's just call it yards per play for now. So our numerator which previously contained:
Passing Yards [plus] Passing TDs *10 [minus] Passing INTs * 45 [minus] Sack Yards lost
Will now also get rushing yards over 4.0 added to it (RYO4.0). And the denominator will stay the same. Here's the final formula to calculate yards per play:
(Passing Yards [plus] Passing TDs *10 [plus] RYO4.0 [minus] Passing INTs * 45 [minus] Sack Yards lost) / (Pass attempts + Sacks)
2) Then subtract the league average ANY/A from the individual QB's yards per play.
3) You now know how much value each QB added per attempt. Let's use an example.
Look at Daunte Culpepper in the year 2000. He attempted 474 passes and was sacked 34 times; so he had 508 total attempts. He passed for 3,937 yards and rushed for 470 yards on 89 attempts. He gained an extra 114 rushing yards over 4.0 yards per carry, so we'll add that to his passing yards, giving him 4051. He also threw for 33 TDs (+330 yards) and 16 INTs (-720 yards), for a total of 3,661 yards. Finally, he lost 181 yards due to sacks, giving him 3,480 yards. That's an average of 6.85 yards per play. In 2000, all QBs averaged 4.85 adjusted net yards per attempt. Therefore, Culpepper added two yards of value per play, so over his 508 plays, that's 1,018 yards (small difference due to rounding). So for the year 2000, Culpepper added 1,018 more yards than the average QB. Where does that rank among all black QBs? Here's the single season leaderboard:
Quarterback team year Lgavg Yd/Pl Value Daunte Culpepper min 2004 5.23 7.45 1323 Randall Cunningham min 1998 4.91 7.79 1279 Warren Moon oti 1990 4.89 6.75 1148 Donovan McNabb phi 2004 5.23 7.36 1070 Daunte Culpepper min 2000 4.85 6.85 1018 Steve McNair oti 2003 4.83 7.24 1008 Donovan McNabb phi 2006 5.02 7.42 810 Daunte Culpepper min 2003 4.83 6.45 796 Doug Williams tam 1981 4.58 6.15 766 Randall Cunningham phi 1990 4.89 6.26 704 David Garrard jax 2007 5.11 7.12 696 Steve McNair oti 2001 4.82 6.16 625 Michael Vick atl 2002 4.97 6.29 597 Warren Moon oti 1988 4.69 6.59 582 Aaron Brooks nor 2003 4.83 5.85 565 Warren Moon oti 1989 4.88 5.94 528 Warren Moon oti 1991 4.86 5.61 509 Kordell Stewart pit 2001 4.82 5.80 460 Warren Moon oti 1992 4.53 5.79 454 Steve McNair oti 1998 4.91 5.72 423 Warren Moon min 1995 5.04 5.67 406 Doug Williams tam 1980 4.45 5.18 401
Here are their raw stats:
quarterback ryo4.0 att pyd ptd int sack sackyd Daunte Culpepper 54 548 4717 39 11 46 238 Randall Cunningham 4 425 3704 34 10 20 132 Warren Moon 0 584 4689 33 13 36 252 Donovan McNabb 56 469 3875 31 8 32 192 Daunte Culpepper 114 474 3937 33 16 34 181 Steve McNair 0 400 3215 24 7 19 108 Donovan McNabb 84 316 2647 18 6 21 140 Daunte Culpepper 130 454 3479 25 11 37 196 Doug Williams 17 471 3563 19 14 18 135 Randall Cunningham 470 465 3466 30 13 49 431 David Garrard 0 325 2509 18 3 20 96 Steve McNair 114 431 3350 21 12 37 251 Michael Vick 325 421 2936 16 8 33 206 Warren Moon 0 294 2327 17 8 12 120 Aaron Brooks 0 518 3546 24 8 34 195 Warren Moon 0 464 3631 23 14 35 267 Warren Moon 0 655 4690 23 21 23 174 Kordell Stewart 153 442 3109 14 11 29 175 Warren Moon 39 346 2521 18 12 16 105 Steve McNair 251 492 3228 15 10 33 176 Warren Moon 0 606 4228 33 14 38 277 Doug Williams 138 521 3396 20 16 23 194
Regular readers of this blog won't be surprised to see 'Pepper top the list. His curious 2004 season was the subject of a still unsolved blog post from last June.
- I think rushing touchdowns and fumbles should factor into this as well, although I'm not 100% sure how. I'll consider that in a future post.
- What do you guys think about rushing yards over 4.0? Should it be over 3.0? Is 4.0 too high? Alternatively, should the whole idea be scratched? Isn't a QB that runs 100 times for 300 yards still more valuable than one that runs 100 times for 100 yards? But if we look at all the rushing data, how do we avoid penalizing Peyton Manning and Dan Marino for taking lots of kneeldowns in their careers? Is it a reasonable assumption to say that every QB takes the same number of kneeldowns per pass attempt? I don't think so.
- Randy Moss was on the top two, and three of the top five teams that had the best seasons by a black QB ever. If I recall correctly, his QB did pretty well this year, too. Terrell Owens was on another one of the top five teams, and his QBs always seem to do okay, too.