Week 3 Quick Reads

Some back-of-the-envelope estimates.

There are 3 main costs of a sack: loss of yardage, loss of down, and loss of possession (turnover on a sack-fumble).

Yardage: the average sack loses about 6.5 yards.

Down: I'd guess that a down is worth something like 4 yards on average (e.g., a 4-yard gain on 1st & 10 is basically a tie rather than a win for the offense or a win for the defense).

Possession: Brian Burke has estimated that about 8.5% of sacks result in a lost fumble, and I'd guess that a lost sack-fumble is worth about 65 yards, which means that the average sack costs another 5.5 yards or so due to risk of turnover.

That means that each play involving a sack costs about 16 yards on average (i.e., it's slightly worse than a 15-yard penalty where you repeat the down). Each additional sack that you take is an additional cost of about 16 yards, as it's another down lost, another 6.5ish yards lost, and another 8.5ish% chance at a lost fumble.

If a QB takes unusually deep sacks that means that he loses maybe an extra yard per sack (less than that if we're talking about whole careers, more than that if we're talking about the most extreme QB in a single season). That's a 15% increase in yards lost due to sacks (7.5/6.5 = 1.15), but just a 6% increase in total cost due to sacks (17/16 = 1.06).

Another way to look at it: If a QB gets sacked 32 times over the course of a season, then adding 1 yard lost per sack (due to higher depth of sack) is about as bad as getting sacked 2 additional times (34 instead of 32); both add the equivalent of about 32 more yards lost. And 1 additional yard lost covers most of the variation in depth of sack between average and pretty bad, whereas 2 additional sacks (for a fixed number of dropbacks) does not cover most of the variation in sack rates between average and pretty bad.

So it seems like variation in depth of sack should only account for a small portion of the total variation in cost of sacks.

"There were some bad fourth quarters this week, is what that means."

That sounds like the way Gil Brandt would have put it.

All lost this week. I would think that almost be impossible to happen.

From the Article
"Luke Faulk -92 DYAR
Falk only threw for five first downs against New England, and four of them came after the Patriots had built a 20-0 lead. He did not convert a third- or fourth-down play, going 4-of-7 for 14 yards with an interception and two sacks. He did not throw a single pass in the red zone; in fact, he did not throw a single pass within New England's 40-yard line. He only had three plays in New England territory: two incompletions and a sack."
I watched that game, this is a pretty good summary; I'll add that the Jets were 0-12 on 3rd down and 0-1 on 4th down, plus a turnover (int). When I read the article I note that Kyler Murry somehow accumulated 48 more negative DYAR (inclusive of his positive rushing plays) then Faulk did and I'm asking myself how a QB could possibly be less effective than Faulk was. From a mathematical point of view it's fairly obvious that it's a result of the greater number of opportunities Murry had to suck (51 drop backs vs. 27). looking at those raw numbers you might think that the Cards non-QB rushing game was effective, but that's not the case, many of the third down conversions were Murry's (the Cards were 9-17 on 3rd down, mostly by scramble or short pass). I guess this day is mostly a consequence of the outsized effect of the 8 sacks Murry took, but I still don't think DYAR is a correct comparison of the ineffectiveness of these two QB games.

I'm starting to just wonder (worry) if this is what Aaron Rodgers is now.

I watched Derek Carr and Case Keenum this week, and doing so made me think, once again, that even the most advanced metrics for qb play overvalue risk aversion. I'll offer no defense of Keenum other than this; unlike Carr, Keenum was actually trying to help his team win, into the 4th quarter. Carr was just trying to get a light workout in, from about midway through the 2nd quarter.

Obviously, comebacks while trailing by 20 plus points are always going to be rare, but I suspect they are made more rare by making avoiding more turnovers a high priority. I suspect that a high variance/risk approach greatly increases the incidence of losing by 35 or more, but also slightly increases the incidence of successful comebacks. It's worth investigating, it seems to me.

Watching the Bears game, where Chicago caused three Keenum fumbles by basically just driving the pile straight back into him, popping the ball up hilariously off his helmet each time, I wonder if Keenum's deep sacks aren't a byproduct of his retreating to get more launch space. It seems stepping up into the pocket is no boon for him.

What happens if you regress sack depth against the triangle plot of QB personality type? Do you seem the same agglomerations of Klingons that you saw there?

"As noted frequently in Audibles, Mayfield has a tendency to drift to his right under pressure ... which is why it's so strange that his numbers on throws to his right were so poor: 7-of-19 for 47 yards with an interception, plus a 15-yard DPI on a 20th throw."

On the broadcast, I thought I caught Collingsworth comment at some point that the Rams were playing towards that tendency of Mayfield's. I took that to mean something like shifting the infielders in baseball against a pull hitter. I'm not sure how you do that in a football defence, but Wade Phillips has been around a long time. If the tape showed Mayfield always escaping to his right, maybe the defensive scheme was designed to expect that, which may explain the poor results.

Another example of Phillips outmatching the young QB came on a throw to the right on the Rams side of the field that should have been a pick, possibly a pick 6, except the DB dropped it. Collingsworth did a good job showing on replay how the zone scheme changed after the snap, and Mayfield threw the ball directly at a guy he wasn't expecting to be there based on the pre-snap read.

I feel like even though Lamar's elusiveness has lead to a great yards per sack statline so far in his career, it doesn't even really do justice to his (or other elusive QBs) value. So far in his young career, he's been in a position where he's about to get sacked and turn it into a positive yardage rush, and some of these have been really really big runs of 10 yards or more. If he breaks a tackle or sidesteps a tackle that would surely have taken down an average QB and almost makes it back to the LoS, it goes down as a 1 yards sack in this stat, but if he gains 10 more yards this stat doesn't even count it. The most obvious fix I think would be tracking yards following a QB hit, but even that doesn't capture when he straight up sidesteps a defender. I don't really have a fix to offer, just pointing out an area for potential future research I guess.

I ran the numbers for Aaron Brooks, and it was better than I expected. 5.4-6.7 and 6.2 average. He's a guy of think of for his willingness to move relentlessly backwards, but maybe outrunning the pass rush backwards gave him a chance to avoid sacks (or maybe my memory is biased by extreme plays)

I have a question about the definition of a sack. I'm envisioning five scenarios.

1. A QB is standing in the pocket, his protection fails, and he gets clobbered.
2. A QB is standing in the pocket, his protection fails, and he scrambles toward the LOS looking for a receiver downfield until he gets clobbered behind the LOS.
3. A QB is standing in the pocket, sees nothing downfield, and runs with the ball tucked but is tackled behind the LOS.
4. A QB takes the snap and runs a draw, showing pass before running, and gets tackled behind the LOS.
5. A QB takes the snap and immediately runs but is tackled behind the LOS.

Scenarios 1 and 2 are definitely sacks. With scenario 3, I think it's generally called a sack but I can see a reasonable argument that it isn't. I don't know whether scenarios 4 and 5 are sacks but I don't see why they should be, other than the position that the player tackled for a loss is listed at.