NFL Week 8 Passing Yards Projections + Week 7 Recap

Getting back into it after my bye week

Due to some traveling I did last week, I never was able to run the NFL passing yards model or recap week 6’s predictions. I’m going to treat that as a bye week for this blog.

It turned out to actually be a great time to take a break from making predictions, as week 7 was absolute chaos in the NFL. First of all, in New England, Mac Jones returned from injury only to get replaced by Bailey Zappe in the first half after some bad play. As a result, the model’s backfilled prediction for Mac Jones last week was off by about 240 yards (had it made a prediction for Zappe, it would’ve been off by about 40).

Also in week 7, Tom Brady got outplayed by PJ Walker, Joe Burrow threw for about a billion yards in the first half versus the Falcons, and the group of Trevor Lawrence, Davis Mills, Jimmy Garoppolo and Andy Dalton all threw for over 300 yards. It was an absolute explosion on offense for some mid-tier QBs, while other veterans like Dak Prescott, Lamar Jackson, and Aaron Rodgers all underperformed relative to expectations.

Week 7 (backfilled) Recap:

I ran the passing yards model today with only data going up to week 6 to "backfill” week 7 predictions without hindsight bias.

There were definitely a lot of misses this week, whether due to injury or benching, some players’ predictions were wildly off. Taking out injuries and benchings, the errors are still a bit larger than in previous weeks:

Let’s see what we got right first:

(Would’ve) Nailed it:

Zappe and Walker are both up strictly because of their massive standard deviations, but if Zappe had played the whole game against Chicago, he might’ve been closer to his 229 yard prediction. For Walker, being 12% off seems like a lot, but being 20 yards off in the aggregate is only the difference of a few completions. Had Tom Brady put the Buccaneers in a better position to compete against the superior Walker, perhaps PJ would’ve thrown for more yardage.

(Would’ve been) Near misses:

In this group we see a mixture of quarterbacks having the games of their lives (Garoppolo, Mills, Lawrence), quarterbacks who should probably retire (Rodgers, Ryan), and quarterbacks who have largely underperformed this year, and whose predictions are probably buoyed a bit by last year’s stats (Goff, Carr, Murray). Jacoby Brissett is an interesting odd=one-out in this group. It seems like in his case he probably benefitted from a few big gains from Amari Cooper and Donovan Peoples-Jones as well as from the porous Baltimore passing defense.

(Would’ve been) Not even close:

Even taking Mac Jones off of this list, there are some bad misses in here. In Burrow and Mahomes’ cases, they are both elite quarterbacks who routinely seem to throw for over 300 yards. They don’t always get predicted to do so, however, largely because it is very rare for a passing defense to give up that many yards. There’s kind of a dragging effect on breakout yardage games for elite quarterbacks simply because there aren’t a lot of them. The prior distribution for the model - which is a probability distribution of passing yards allowed by the opponent’s defense, is not going to have a lot of 300+ yard observations, and certainly very few 400+ yard observations, since there are so few quarterbacks who can achieve those feats. It is also worth noting that both Burrow and Mahomes’ performances came against non-divisional opponents. Against teams which see Mahomes or Burrow twice as frequently per year (divisional opponents) we would, all else equal, expect the predicted passing yards to be greater for those quarterbacks.

Nearly everyone else on this list had a massive underperformance or is just a streaky quarterback. Dak Prescott is also coming back from a long-term injury so it might take a few weeks to regain form.

Finally - congratulations to Andy Dalton on doing his best Jameis Winston impression. 361 passing yards, 4 TDs, and 3 INTs is truly a feat. I still miss Jameis but appreciate Andy’s hard work.

Andy1of1

Week 8 Projections:

The Ironclads:

This is a group of QBs that the model has generally done pretty well with this season as a whole. Lamar Jackson is coming off of one of his worst passing performances in a while, and against a Buccaneers defense which allowed PJ Walker to show his greatness in week 7 it looks like Lamar could have a bounce back.

The Relatively Stable:

I’m including Russell Wilson in here because, at the time of writing, the Broncos have not announced if they’re going with Rypien (also projected later) or Wilson this week. Wilson and Rodgers both seem to still be overrated a bit by the remaining 2021 stats which are factored into the model. Buffalo has one of the best pass defenses in football so it will take a season-best effort for Rodgers to get to 254 passing yards.

Will he or won’t he?

Jones’ 13 yard first quarter performance last week definitely boosted his standard deviation a bit, but now that the Patriots have shown a willingness to pull their QB if he gets off to a cold start, this kind of uncertainty in the prediction is probably a good thing. The rest of the occupants of this list make intuitive sense whether due to only having a few starts this year or due to uneven play.

These guys might not play:

Pretty self explanatory. Rypien would be making his second start so there isn’t a lot of data on him. Zappe has shown flashes of great and mediocre play, and also didn’t get a full start last week. If he does come in the game it will probably be in relief of Mac Jones and, depending on the amount of time he is given, this could impact his totals a lot.

The unpredictable:

The Colts announced this week that they’re benching Matt Ryan in favor of rookie Sam Ehlinger. Since this is his first start, we don’t really have a lot to go off of. One topic I want to investigate in the future would be predicting rookie QB performance using the previous starter’s stats and the rookie’s college stats. Since I don’t have time for that before the game tonight, I’m going to do some back of the envelope math here.

Here we have the Commanders/Football Team’s allowed passing yards over the past season and a half. The mean allowed is 253 yards, while the mean passing yards accrued by the Colts this season is 259 yards. If Matt Ryan were to start this weekend, then here would be his projections:

To do a very rudimentary analysis of how backup quarterbacks have performed this season as compared to the true starters, we calculate average yards/game for PJ Walker, Brett Rypien, and Cooper Rush, and divide that by the yards/game average for Baker Mayfield, Russell Wilson, and Dak Prescott.

On average, in 2022, backups have thrown for 171.3 yards per game, while their starters have thrown for 211.2 yards per game. If we use that ratio to estimate Sam Ehlinger’s performance, his predicted passing yards would be 272.36 * 171.3 / 211.2, or 220.9 yards this week against the Commanders.

Mid-Off of the Week (aka the NFC South Game of the Week): PJ Walker vs. Marcus Mariota

As much as I wish I could say Colts-Commanders here, there was a clear choice

Mariota has only thrown over 200 yards twice this season, even when he has faced poor defenses. PJ Walker, on the other hand, just has talents that aren’t quite captured by the box score.

My prediction: PJ Walker will take home his first NFL (he might have an XFL title already) Mid-Off of the Week!