The numbers bring to mind the oddity of the big showdown Jets-Pats game of 2011 in which the Pats basically rolled New York out of the stadium, but New York ended up with the higher DVOA since they essentially either went 3-and-out or scored each time they had the ball. Every once in a while, there is a game that exposes this glaring blindspot in DVOA. I don't think it calls the model into question though. Averaged over the course of a season, which is where DVOA's actual power comes in, the counting stat makes a lot of sense.

Let's put it this way. I don't think Belichick is ever going to be remotely interested in what DVOA says when it comes to evaluating his players or his team. So, in that sense, it's not a useful measure. I think, at best, it's a stat (like many stats) that can pique your interest in exploring a team or player in a more informative way because of 'surprising' things that get highlighted.

Julio Jones was almost perfectly covered on three of his catches. He made great catches, and any stat will say that, but the NE coaches will look at the tape of those catches and know that the defense did what it's supposed to do. Next time those same plays may all be picks, or all defensed.

DVOA is never going to be good as a predictive, or even evaluative stat. It's an historical stat that can be a nice way to traverse football history from a wide-angle view.

I've been watching the Pats all year. And Brady has thrown a good number of missed passes his way. Without going into the numbers too much, about 3/8 of the passes his way have been incompletions. Yesterday was the worst of the season, at 5 receptions in 13 targets. But there were other games, like vs. SF and vs. Bal, where Brady kept going to Edelman without success.

And crossing patterns are very different than deep sideline routes.

Edelman excels at getting open because he changes direction quickly and reads the defense well enough to know where he should go. But he's not great on routes that depend on a receiver's speed to create separation. Mitchell, who is noticeably faster, is better at those. Hogan probably is better at those, too.

And even beyond missed attempts where Brady might be forcing the issue, Edelman seems to drop one very easy completion about every other game, and has done so for years (granted, he receives a lot of attempts, but I recall one analysis from last season which had Edelman leading the NFL in drops by a wide margin over a three-year period). Edelman had one of these in the AFCCG against the Steelers, and another one in this game on 3rd down that killed a viable scoring drive on the Patriots' first possession of the second half with the Patriots still down 21-3. DVOA may not be able to appropriately account for these drops versus any other unconverted attempt, but they're important.

Edelman also has durability issues. Folks here in NE are supremely high on him, I think mostly for his toughness in-game, when he often seems to absorb horrific hits and bounces back up, but I still think that Welker was a much more consistent slot receiver. I imagine Edelman will stay or not depending on his cap impact. Even if his role is reduced, there really isn't a good reason why the team would want to let go of him. The player that befuddles me is Amendola. I imagine he is done in New England based on salary, and the team has been looking to replace him for some time, rarely letting him see the field for long stretches. But he always seems to show up in the clutch at key moments in the playoffs. What gives?

I think Belichick already started doing this a while ago, what with picking up Gronkowski and Michael Bennett.

Also with Michael Floyd late in the season.

I think the Patriots are just happy to have a combination of both small quick receivers and big burly receivers, really. It's just that these big receivers are much, much more expensive on a salary cap if they're wide receivers, and not so expensive if they're TE's.

I kinda expect the Patriots to be drafting more TE's.

For a team that's in this position as often as they are, it's amazing how often the Pats season plays out like this in the Belicheck era. Obviously four of the five SB wins were classics, but so were both losses, and last year's AFC championship and the one they lost to the Colts. More than any other team I've ever seen, they're able to extend the game just to the point where they have a chance (Manning could do this at his best, too, just not as reliably- throwing every down helps). It's true that Belicheck has benefited from lucky bounces in razor-thin margins to burnish his legacy, but, at the same time, he could easily have two or three more rings with just a little more luck

I've long thought that there is a happy medium between play-by-play (DVOA), which gives you lots of data and has useful splits but overlooks the effectiveness of how plays fit together and the higher leverage of certain plays (e.g. 3rd down), versus drive stats (e.g. drive efficiency), which is a good indication of how an offense or defense performed, but smears out the fine details, and is plagued by small sample size.

I think that happy medium would be a series efficiency stat. I.e. each time a team faces 1st down, one of about five things can happen (getting another 1st down, punting, committing a turnover, getting a FG, or getting a TD). By comparing a team's performance to league average for a given context (e.g. starting at your own 10 versus the opponent's 35, for example), you can get something that is probably more resilient to games like this one, and yet still has lots of interesting splits possible and doesn't suffer as much from sample size issues as drive efficiency. I've been meaning to try to work on something like that in the offseason.

My thought is that it is analogous to batting average in baseball--when a batter is at bat, sabermaticians don't look at the success on a per-pitch basis, but a per-at-bat basis. A hitter that always swings (and often strikes) on the first pitch may still be a good hitter if he's good at fouling so he can battle deep into the count, and BA may still like him, whereas a per pitch metric would hate him, just like DVOA won't like a team that throws deep on 1st down every time and fails often, even if they are led by Peyton Manning and can easily convert just about any 3rd down. (Another way of saying that 1st down success matters a lot less to a good offense than to a bad one--DVOA won't pick this up, but a per-series metric will).

But without baselines for comparison, I did a quick analysis just on this game, and it tells quite a different story than DVOA.

The Patriots converted for 1st down 32 times, with 4 punts, 2 turnovers, 2 FG, and 4 TD. That means that they were successful (1st down or TD) on 82% of their offensive series, punting on 9% and committing turnovers on 4%.

The Falcons, on the other hand, had only 14 1st down conversions, with 6 punts, 1 turnover, and 3 TD. So they were successful on only 71% of their series, punted on 25%, and also had turnovers on 4% of their plays.

So both the Patriots and Falcons turned over the ball on 4% of their series, but the Falcons were 2.5X more likely to punt than the Patriots each time they had 1st and 10. They were also 11% less chance of having a positive outcome than the Patriots (and that's neglecting FG's, which are kind of a neutral outcome).

I'd be happy to just see them put points on the board in the first quarter.

I'm getting really sick of them looking unprepared in these games, after being so overly prepared in the regular season.