Cumulative Effect: Individual QBs

Taking his inspection to the individual player level, Nathan Jahnke contiues his look at hits on the QB.

| 2 years ago

Cumulative Effect: Individual QBs

cumulative-effect-indLast week I looked into how quarterbacks were effected by getting sacked and hit multiple times over the course of a game. This was followed by going into more detail by controlling for the quality of the quarterback, the depth of the pass, and the quarter of the game. Through all of that, the general conclusion was: the more a quarterback gets sacked or hit, the more their accuracy decreases.

The next step is looking at some individual quarterbacks. The problem is that the sample sizes for looking at a single quarterback in a single season is too small. The average full-time starter averaged just 128 attempts in a season where they had yet to be sacked or hit in a game. The sample size of how many passes a quarterback has after getting sacked or hit once is 94, and the samples just gets smaller from there. Therefore, any single pass attempt can have a large effect on their Accuracy Percentage.

However, we have six full years of PFF data, and looking at that six-year span greatly increases the sample size and leads to some interesting results. For the majority of quarterbacks, there is a little noise in the data, but there is a general downward trend. For a typical example here is Kyle Orton from 2008-2013:

Over the past six years Orton has been more accurate after getting sacked or hit once than he has been without getting sacked or hit. After a few additional sacks/hits his accuracy decreases significantly, but then he is a little more accurate after five sacks/hits than he is at four. I would expect given an infinite number of passing attempts that from one point to another there would always be a decrease. Unfortunately, when looking at various subsets of data like this, sample size will always be an issue.

There are some quarterbacks who show a very clear decline. On average a quarterback will see their Accuracy Percentage decrease by 0.5% after each sack/hit they take. These quarterbacks show a decline of over 1% for each sack/hit they take.

For these teams, their offensive line’s pass blocking is more valuable than other team’s offensive lines. Last year while Drew Brees was amongst the top few quarterbacks in the league, his overall PFF rating was much lower than it had been the previous two years. He also had Charles Brown as his starting left tackle for most of the year. He was also sacked and hit more last year than the previous year.

This could potentially help explain part of why Matt Ryan declined in 2013. In 2012, Ryan had Sam Baker, Justin Blalock, Tyson Clabo and Todd McClure as four of his pass blockers, and the team allowed 68 sacks/hits on the season. In 2013 he was without all four of those blockers, and the team allowed 85 sacks/hits. While in general the better the quarterback, the less getting sacked/hit multiple time effects them, it is interesting to note that quarterbacks of varying quality can see such a dramatic change in accuracy.

While the majority of quarterbacks follow trends like the first two graphs, there are a few exceptions to the rules. Below are two quarterbacks who have been an exception to the rule.

My best guess to the reason of this is these quarterbacks needs some game time before they reach their peak performance. It could be that these two quarterbacks are different from the others and truly get better after taking a few hits. It could also be that these quarterbacks have more noisy data than others, and if we go back to Michael Vick’s career before 2008, or in the future see more of Tannehill’s career, we might see these graphs become more similar to the other quarterbacks.

While the general trend is quarterbacks gradually getting worse as they get sacked/hit, like most statistics there are exceptions to the norm. For some of the younger quarterbacks on this list, it will be interesting to see if these trends continue over time, or if they become more like the other quarterbacks in the league.


Follow Nathan on Twitter: @PFF_NateJahnke

  • statsman

    Very interesting followup article. Vick and Tannehill’s plots are particularly fun. To minimize the effects of noise in these data, I think it would be helpful to just show a best-fit linear trajectory (weighted for number of snaps at each point) through the points instead of a line connecting each noise fluctuation. It’s also pretty straightforward to compute a measure of how statistically significant each correlation is based on the amount of data. Even if there’s only a 1% probability that each correlation you find is due to chance (noise), for example, if you collect enough QBs, you’ll always end up with a couple that are spurious.

  • Fintasy

    I can totally understand T-hill and Vick’s graphs. Just like some professional fighters say they don’t really wake up until they’re hit. Something snaps inside them telling them “they’re in a fight” and then they actually perform better.

  • Happysour

    These QB’s are terrible!!! 0.7% accuracy?! I wouldn’t want my QB completing less than 60% of his passes….but less than 1%?! What a disaster!

    • Jesse Reynolds

      I have the same question, what is the % based on. Tannehill completed 60% of his passes last year…

      • Nathan Jahnke

        It is based on Accuracy Percentage rather than Completion Percentage. Accuracy Percentage takes into account dropped passes, passes thrown away, passes spiked, passes hit as thrown and batted passes.

        • Happysour

          Gotcha 😉 Was just being a pain
          .% Is the title not the unit.

      • LightsOut85

        Happsour was likely joking. Whoever formatted those tables labeled it “Accuracy %”, but showed a decimal on the axis (where it shouldn’t have said %, or have written 70.0, 80.0, etc, if they said percent).

  • LightsOut85

    IDK if the writers would know this – but why can’t I view these new graphs you’ve taken to using, in Chrome? (I can see them fine in Fire Fox). Is it related to using an ad-blocker?

    • Actinn

      Looks like it. I can’t see the graphs when AdBlock is on in Chrome but I can see them fine when I turn it off.

      • LightsOut85

        Weird (I hate when sites find work-arounds to AB). I added to my white-listed domains (AB+) and still don’t see them.

      • Kaedwon

        I’m using Chrome w/ABP and the grafix look fine. Suggest a different issue at work.

    • Mathew Starner

      Same here, I had to switch to Safari to see them, and even then I had to reload the page a few times.

  • Scott@Seattle

    It would be interesting to see what Colin Kaepernick and Russell Wilson look like.