≡ Menu

The most accurate passer in football

The most accurate passer in football.

It’s Friday, so I thought it might be fun to play around with some stats.  Net Yards per Attempt is probably my favorite predictive statistic to measure quarterbacks, but there are some problems with even that metric.  One issue is that Net Yards per Attempt — which is simply yards per attempt but includes sacks data — is pretty sensitive to outliers.  A quarterback who consistently pieces together short passes with a high completion percentage can be pretty valuable, and may end up undervalued compared to some mad bombers.

There are a couple of ways to deal with this.  One is to use a different measure of central tendency than the average production per dropback; for example, we could look at the median yards gained per pass attempt (including sacks).  Another is to measure the standard deviation on all of a quarterback’s pass plays. I thought I’d compile the data on both and see what you guys found interesting.

No matter how you splice the data, Philip Rivers looks outstanding. After a couple of down years, Rivers is experiencing a career revival under new head coach Mike McCoy.  The Chargers no longer rely on a downfield passing attack (and with Vincent Jackson gone and Malcom Floyd on IR, that may be more out of necessity than design), but Rivers has found a Darren Sproles replacement in Danny Woodhead.  As a result, Rivers has completed an incredible 73.9% of his passes this season.

Rivers ranks 27th in average length of pass (or average depth of target), reflecting the shorter passing attack, but Tony Romo, Chad Henne, Matt Schaub, Sam Bradford, Matt Ryan, and Alex Smith have lower average distances and worse completion percentages (among other stats).  The Chargers star has also been great at avoiding sacks: he’s completing passes on 70.8% of his dropbacks this year, a stat I’m calling Adjusted completion percentage (A_Cmp% in the table). In the table below, I’ve listed each quarterback’s number of attempts and sacks, his Adjusted completion percentage, his Net Yards per Attempt, and his standard deviation on pass plays. Since standard deviation would be biased towards quarterbacks with higher averages, I’ve sorted the table by the Ratio of each quarterback’s standard deviation to his NY/A average. Finally, I’ve also displayed the median number of yards gained for each quarterback on each dropback. All data excludes last night’s Carolina-Tampa Bay game.

1Philip RiversSDG2491170.8%7.929.81.26
2Peyton ManningDEN289969.5%8.4511.21.35
3Robert Griffin IIIWAS2381157.4%6.6191.45
4Matt SchaubHOU2331560.5%5.838.41.44
5Drew BreesNOR2371462.5%7.4310.81.45
6Jay CutlerCHI2251062.1%6.759.81.54
7Matt RyanATL244967.6%7.310.71.55
8Matthew StaffordDET290959.5%6.910.51.55
9Ben RoethlisbergerPIT2152160.6%6.4610.11.64
10Tony RomoDAL2651664.4%6.710.61.65
11Russell WilsonSEA1872055.6%6.6210.51.64
12Andy DaltonCIN2491562.1%6.95111.65
13Colin KaepernickSFO1821552.3%6.8110.91.63
14Andrew LuckIND2241556.9%6.15101.63
15Aaron RodgersGNB2201560.9%7.6612.41.65
16Alex SmithKAN2501854.1%5.591.62
17Sam BradfordSTL2621557.4%5.749.41.63
18Chad HenneJAX1731654.5%5.929.71.63
19Carson PalmerARI2662056.3%5.629.31.64
20EJ ManuelBUF1501352.1%5.599.31.70.5
21Jake LockerTEN1521257.3%5.879.91.73.5
22Christian PonderMIN1001053.6%5.889.91.73
23Brian HoyerCLE96655.9%5.569.41.72
24Eli ManningNYG2681851%6.3310.91.70.5
25Cam NewtonCAR1701857.4%6.310.91.75
26Tom BradyNWE2852051.8%
27Michael VickPHI1321448.6%7.5213.31.80
28Mike GlennonTAM130755.5%4.798.71.82
29Geno SmithNYJ2232552.4%6.1811.31.82
30Joe FlaccoBAL2692055.4%6.1711.41.92
31Ryan TannehillMIA2192654.3%5.7510.71.92.5
32Terrelle PryorOAK1382156%5.77111.92
33Josh Freeman2TM147840.6%4.528.71.90
34Brandon WeedenCLE1952147.7%4.679.220
35Blaine GabbertJAX861242.9%4.2210.22.40

Week 7 Average Field Position

As always, here is the average field position data from week seven.

The Baltimore-Pittsburgh game produced an interesting box score. You might see the final score of 19-16 and think typical Ravens-Steelers defensive battle, but each team had only seven possessions. Pittsburgh punted on their opening drive, and then didn’t punt again for the rest of the game (touchdown, field goal, fumble ending a 9-play drive, field goal, field goal, field goal); Baltimore started three-and-out, and then went FG, 10-play drive that ended in a punt, FG, 8-play drive that ended in a punt, 12-play drive for a field goal, and 16 play drive for a touchdown. The Steelers were 7-12 on third downs, while the Ravens were 7 of 14 on third downs and converted their lone fourth down attempt. I didn’t catch any of this game, but it looks like the offenses may have been better than you think (Pittsburgh also benefited from two long kickoff returns).

Anything stand out to you guys in the AFP data? Would you like me to keep publishing it?

  • Nate

    I wonder about the whole mean yards / mean adjusted yards / median/qartile/quintile yards thing. Median yards gained is theoretically nice, but the resolution clearly leaves something to be desired. One thing that occurred to me a while ago was pushing the play data through Brian Burke’s EPA model to see if that produces anything interesting. (In theory that will provide proper credit for interceptions, first downs, and the like.)