Nerds on Sports Where nerds are talking about sports!

May 15, 2007

Revising the Passer Rating

Filed under: Football — Tags: , , , — Perich @ 9:26 am

This post may contain recycled content.

I have a mission for you: revise the QB rating.

The QB rating (technically called the “passer rating,” as that’s all it’s meant to evaluate) is an arcane formula devised when the NFL and the AFL merged in order to standardize statistics. Rather than compare quarterbacks to each other – which provided no metric for season-to-season growth, statistician Don Smith devised a way to measure their progress objectively.

The problem is that (A) it’s not very objective and (B) it’s cryptic.

(A) The rating system is great, in theory. Take the four key areas of a passer’s performance: completion percentage, yards gained per pass, number of touchdowns and number of interceptions. Normalize them against some industry-standard benchmarks (50% pass completion, 55 interceptions per 1000 passes, etc). Total these numbers and voila.

Peyton Manning wishes he had the talent of Drew BreesThe problem is, those benchmarks were set in 1971 and have not been revised since. The “yards per pass” metric, which figures an average of 7, punishes QBs whose coaches use the fire-and-forget “West Coast Offense.” Is a 5.5% interception percentage still average? Look at the numbers yourself and tell me: as of the end of the 2006 season, Dallas’s Tony Romo and the Saints’ Drew Brees led the league in passer rating (89.6, 88.3, respectively), while those jackasses Tom brady and Peyton Manning couldn’t even crack the top 5.

(Cold Hard Football Facts agrees with me, and has some pictures of Pamela Anderson to boot. No, for serious)

(B) In order to normalize against those aforementioned benchmarks, you have to go through some weird hoops. Nothing terribly complicated – just lots of division.

({[(Comp / Att) x 100 – 30]/20 + [(Yards / Att) – 3 x 0.25] + [(TD / Att) x20] + [2.375 – (INT / Att) x 25]} / 6) x 100

Remember your order of operations and show all work.

There has to be a better way – and this wouldn’t be “Harebrained Schemes Tuesday” if I didn’t come up with one.

The NerdsOnSports Alternative (v1.0):

Average together the following two figures:

A: Passes Completed divided by Passes Attempted

B: Touchdowns divided by [Interceptions plus Touchdowns]

Multiply the result by 100.

Step A is pretty obvious – it’s completion percentage. Step B was a judgment call on my part, but I’ve long argued that an interception costs a team just as much as a touchdown benefits them. If anyone wants to dispute that, take it up in Comments.

Please Execute My Dear Aunt SallyThe only problem I’ve seen so far with my new metric is that it tends to cluster everyone around 55 or so. While there’s nothing wrong with that, it’s not a very useful stat if the difference between the best QB in the league and the worst is only 4.28 points. But I’ve only tested it with career stats. Game by game, it might work better.

But I’m sure you have a better way! Enlighten me, Intertron.

  • Well, I’m a better qb than that a-hole Brady, plus I’m not knocking up every dame that comes my way.

  • Serpico

    I think the weighting and merging of “unrelated” statistics creates a result that has less meaning than the components.

    It’s like OPS (On-Base Plus Slugging) in Baseball. If you see an On-Base of .400, you can assume the player is patient at the plate. If you see a Slugging Percentage of .700, you can assume the player routinely hits doubles and homeruns. If, you see an OPS of 1.100, you are not sure where his strength is. Is he an astonishingly patient slap-hitter? Is he a reasonably patient power hitter? Is he an “all or nothing” swing-for-the-fence guy?

    The joy of sports for many come from the debate. “Manning has more TDs, but Brady has a higher completion percentage. McNair has a better TD/INT ratio, but Favre has a superior Yards Per Throw.” If you try to merge them all, or even just two of them, you lose something.

  • While I don’t agree with you re:OPS (and ne’er will, and someday the schism may rend the world), I see what you’re saying re: QB stats.

    Having just put every QB from last season into a spreadsheet with these formulae, I’ve discovered that my formula HEAVILY weights interceptions. One extra pick can drop you 2-5 ranks on my listing. While I still think interceptions are important – and underrepresented in the current metric – I’m not sure I want to make them that important.

    Also: the old rankings (passer rating) and the new (Perich rating) are very similar in spots. Romo’s still #1, Peyton Manning’s still #8, Brady moves down a slot, Hasselbeck moves up a slot, etc. If I’m working this hard to reproduce the current system, it kind of defeats the purpose.

  • OK, but what we really need is a formula for comparing, say, Eli Manning to Damion Easley or Baron Davis.

  • Now whatif, and this is going to sound crazy, you use the same type of scoring used by Fantasy Football to rank/rate your QBs? Only difference is you divide the score by games played or something so that it averages out.

    So (Completions + (yds / 25) + (6 * TDs) + (-2 * Int)) / Games Played.

    I don’t know how standard 1 point per 25 yards is or if you want to modify the points gained/lost on TDs and Interceptions.

  • I like, Will. Since it’s really the fantasy football nerds who care the most anyway, why not borrow what works?

Powered by WordPress

%d bloggers like this: