Super Bowl LVII DVOA/Quick Reads

And remember, one of their drives started at the Philly 5.

Yes, that's why I said 7 "real" drives. A drive that starts at your opponent's 5 isn't a "real" drive - kicking a FG there is negative points, not positive.

Philly actually generated more points/drive than KC did

Not really, because of the turnover. It's either 30 or 27 points on 9 real drives, depending on how you cost the turnover (-4, its average, or -7, its actual). I do agree that the massive imbalance between the two is due to rebalancing it in per-play terms rather than per-drive.

but I think the main takeaway from what DVOA is saying is that converting third downs at that rate is not sustainable and if your defense is not even slowing down the opponent, you're usually going to get crushed.

This is why I'm saying that the problem here is that it's a small-sample size problem. The underlying assumption for play-based metrics, like DVOA, EPA, S&P, or even yards/play is that the plays are independent.

They're not. We know they're not. Every coach on the planet will tell you they're not. You run tempo because they're not independent. You use play-action because they're not independent. You script plays because they're not independent. And the Chiefs scored 2 touchdowns on Sunday because they're not independent. We talk about "halftime adjustments" because, again, we know they're not independent. 

In the second half, the Chiefs ran a drive that got down into the red zone. They ran Jet motion on a play and watched how the Eagles reacted. And the Eagles reacted like they thought they would, so they faked the Jet motion, got a guy open, and scored a touchdown.

Then they stopped the Eagles quickly and got a long punt return, putting them straight back in the red zone. There were literally no other defensive plays there - it went straight from "red zone" to "red zone," and so the Chiefs just... did it again, to a different player - and everyone's being like "why didn't the Eagles DC adjust?" Because it happened one right after the other! It's entirely possible that the Eagles DC was busy trying to fix other things that went wrong on the prior drive, then the punt return happened, and crap, they're back there again.

If the game had been a normal length, it's likely that things would've shifted back towards Philly's defense as they adjusted. Kansas City wasn't a step ahead of them the whole game (at least not like they were in the second half). But it wasn't. It was super short, and amplified by that punt return.

Kansas City ran a grand total of 53 plays, compared to Philly's 70. A typical team game is 63 plays. So yeah, this is a more limited sample size than normal.

Agree.  This game had 125 plays.  That should be more than enough for a decent sample size.  If the professed win probability doesn't agree with our intuition about what happened, that suggests a model problem, not a sample size problem.

The Eagles had more total yards, more plays, more first downs, dominated time of possession, and lost the game by 3 points.  That shouldn't mean a 98% postgame win probability for the Chiefs. 

I am interested to see a breakdown by half.  The Eagles completely dominated the first half, and then produced nearly nothing in the second half.  The Chiefs got a lucky fumble recovery for a TD early, but completely dominated the second half.  

DVOA consistently does a pretty good job of identifying which teams are and aren't very efficient in their play. But football games are inherently unpredictable and the data set of "an entire season of NFL plays" is large, so there are bound to be statistical anomalies. 

Earlier this year, we had Vikings fans posting here wondering when DVOA would be changed to account for their team's dominance. Of course, those fans were delusional ... advanced metrics and traditional metrics alike pinned Minnesota as a below average team that experienced unprecedented luck in close games. And when the playoffs came around, the Vikings stopped winning and those Vikings fans stopped posting here. Minnesota was an anomaly! DVOA was right!

This game, too, looks like a bit of a statistical anomaly. But it seems a lot less anomalous if you think about it this way:

Over the course of a few decades' worth of data, DVOA sees that teams whose offenses face a lot of third and fourth downs usually lose. In contrast, teams who can consistently convert first downs on first or second down usually win.

That makes sense, right? 

Let's look at one of Philly's drives ... their first drive of the second half. They took 17 plays to move the ball 60 yards.

• First they had to convert a 3rd and 6.
• Then they had to convert a 3rd and 2.
• Then they had a penalty on a 3rd and 9, pushing them back to 3rd and 14. Miraculously, they converted that into a first down.
• After that, they faced a 3rd and 1. No problem, right? WRONG. They were stuffed for no gain on the play.
• But then they kept the drive alive by converting a 4th and 1.
• The drive finally stalled out in the red zone as the Eagles opted to kick a field goal facing 4th and 6 from the 16.

Does that seem like a team that's crushing their opponent? Or does it seem like an overmatched team that's pulling out all the stops in order to keep up with a superior team?

Watching the game, it didn't FEEL like the Eagles were desperate or overmatched. They LOOKED like they were in control for the first half. DVOA couldn't watch the game and make that judgment, though; instead, DVOA watched the stats and said, "Oh, this data matches up pretty nicely with data from teams that lose!" And guess what? The Eagles lost! DVOA was right!

Inevitably, at some point next year there will be a game in the NFL where a team faces 18 third downs. There will be a game where a team gives up 6.1 yards per carry on the ground. There will be a game where a team takes eight minutes and 19 plays to score a field goal. DVOA will look at those games and say, "Hey! I've seen this before! This is what an inefficient team looks like when they lose a game!" Maybe that inefficient team will luck their way into a last-second victory as did the 2022 Minnesota Vikings on so many different occasions. Maybe that inefficient team will lose a nail-biter like the Super Bowl runner-up 2022 Philadelphia Eagles. Or maybe that inefficient team will get blown out à la the (pick a year, any year!) Cleveland Browns.

Regardless of the outcome, DVOA is telling us that efficient, winning teams almost never perform that way. Or, more accurately, DVOA is telling us that efficient, winning teams perform that way about 2 out of every 100 times.

So, you know ... do with that information what you will.

I think it's overstating it to say that the eagles played like the worst team, but it's not unreasonable to say that this was one of their worst outings. If you win the time of possession battle and lose, it's a pretty fair guess that maybe you weren't making the most of your possessions. Couple that with facing a league-average defense and giving up a fumble -6, it's pretty fair to say that they left points on the field, and Reid/Mahomes are going to eat that up all day.

 

First, if that is truly the case, what are we doing here? 

Talking about silly things is fun?

It's the same thing with DYAR. Player value by DYAR has absolutely no solid basis whatsoever. Baselines are built from averages between entirely different players and you pretend Lamar Jackson is playing the same position as Jalen Hurts or Tom Brady. We pretend a guy who played wide receiver his entire career is a running back because a team hands the ball to him a bunch. Guys are labelled tight ends because a team doesn't want to call them a wide receiver and pay the cost associated with it.

Game-by-game total VOA's silly to me. The individual breakdowns are interesting because you can reweight them in your head easily. Like, yeah, Philly's defense looked horrendous in the second half that game, but you can also recognize it's only three drives.

that is either so sensitive to sampling noise

Of course it's sensitive to sampling noise. If there's one punt return in the game and you give up a touchdown on it, your punt return ST DVOA looks like the worst team in the league.

I've made this comment before, and it always seems to get mocked - but it's the difference between a frequentist-style approach and a Bayesian approach. At least for me, my brain tends to "think" more Bayesian instinctively - you come in with a prior understanding of a team, and the events that happen in the game change it... but only a little. You get a punt return near-TD, you think "man, that's terrible timing for that punt" - that's a Bayesian mindset. You've got a prior idea of how often those events happen.

Whereas the pure frequentist measure for that game is that the Eagles give up near-TDs on 50% of their punts. Which is so, so bad.

It somewhat matched the eye test for KC D. It doesn't appear to match the eye test for the Philly O. But when you step back, it kind of does?

If KC doesn't convert that first down and kicks a FG with 1:48 left, do you think KC stops Philly from scoring?

Because DVOA sure as shit does. I think Philly felt super-confident about scoring if they get the ball back.

Especially not if the judgment is based on hindsight. 

They could have run a different play on third down that would have had a chance to get the first down instead.  Instead of a forced rollout and a throwaway.  If you're going to punt on 4th and 3, the 3rd and 3 play call was bad.

Suppose the Eagles don't give up the big punt return.  If the Chiefs drive down and get the TD anyway, you're probably more screwed.  Because that TD comes at 5-6 minute mark.  Now Eagles go down and get TD and conversion, say it's 1:30.  Pretty easy still for Chiefs to go down and get FG at the gun, game over.  Don't even need a def. hold for that one.

So the Eagles actually got a big break when the Chiefs scored so quickly, and they STILL only got one more meaningful possession.

YOU'RE BEHIND IN THE 4TH QUARTER OF A SUPER BOWL.  AND YOU'RE PUNTING.

It shows a tremendous and unwarranted trust in your defense to get stops given what we'd all been shown in the second half.  The Chiefs went down for 2 straight TDs and barely broke a sweat, never a long 3rd down conversion.  It just doesn't seem likely at that point in the game that you're going to get a 3 and out, or FG FG while you get a TD with what wound up being your one remaining meaningful possession.

Or...  you can trust the offense that's already scored 27 points.  Give them a chance to get 3 yards.  Including the one where they punted, they would have actually had two meaningful possessions.  Isn't it possible that Eagles offense could go TD FG or even TD TD, that forces the Chiefs offense to actually perform at a top level for the remainder of the likely possessions.

The punt was a terrible idea.

Brown helped, but mostly in the sense that he wasn't Reagor (who is aggressively useless to the point of counter-productivity) and it kicked every receiver down one notch. You remember Quez Watkins dropping a 39 yard-pass that hit him in the hands? He was their WR2 last year.

So Brown's utility is less in being awesome (he is), it's mostly in being competent at running routes and catching catchable balls. I'm not sure Brown actually raised Hurts' catchable ball ration any, although he was a boon in terms of converting those opportunities into actual reception and TDs, versus the INTs the other guys tended to have occur.

Hurts this year has been startlingly accurate, usually putting a ball into someone's hands. Whether they catch it and/or stay inbounds is on them, but he usually throws a ball that can be caught by a professional catcher of balls.

That Mahomes fumble vs. Cincy has to be the single weirdest fumble I've ever seen.  

do something KC weird like line up for a sneak tight, then everyone flex out and run a shot play and watch the freaking safeties run like hell.

I predicted this the first time PHL lined up all bunched up for the sneak, but they never did it.  Was very disappointed.

Yeah, that's true, but even in those situations you want to have more of a threat of the shot play. The plays they actually ran there were really basic.

It's really just the difference in playcalling between the two teams. I don't think Sirianni/Steichen have enough experience at this point to really force a defense back on their heels constantly - the success in that Super Bowl was far more about just the talent advantage Philly had. Using a center to pull isn't some weird schematic wrinkle, it's just "normally, centers can't do this, but ours can!" And going for it in short yardage situations isn't some grand new strategy when, again, the talent advantage is just so massive.

Whereas Andy, dear God. Andy's staff is just incredible. I still don't think they're getting enough credit, which is crazy. I know everyone's anointed Mahomes as the greatest thing ever, but even if he continues this way after Reid retires, it's still Reid who taught him. It's a similar thing with Travis, too. I wonder if people sometimes forget that Reid was a tight end coach before he was a quarterback coach.

20 *.03 = 0.6

If your numbers are correct, it's off by a zero, it should be 0.3

Yes, I forgot to subtract out the missed FG since that is special teams and not offense. Removing the FG gives me the same answer of 0.16 EPA/play for KC for the first half. 

This increases KC OFF EPA per play for the whole game to 0.41EPA / play.