Writers of Pro Football Prospectus 2008

Most Recent FO Features

DarnoldSam18.jpg

» FEI Week 3 Ratings

The key to USC's win over Texas? Despite consistent struggles throughout the game, Sam Darnold and the Trojans offense were at their best in end-of-half scenarios.

31 Jan 2017

QR Bonus: Keys To Beating New England

by Vincent Verhei

Every year at this time, we go back and look at the worst four games by each conference champion in search of common themes that could cost them the Lombardi Trophy. We covered the Atlanta Falcons last week, and today we'll look at New England. That's going to be tricky, though, because the Patriots had a very weird season.

Let's start with the surface-level information. By DVOA, New England's four worst games of the year were:

  • A 23-21 win over Arizona in Week 1.
  • A 31-24 win over Miami in Week 2.
  • A 16-0 loss to Buffalo in Week 4.
  • A 31-24 loss to Seattle in Week 10.

Three of these games came with Tom Brady suspended, and either Jimmy Garoppolo or Jacoby Brissett at quarterback. Neither Garoppolo nor Brissett are likely to see much time in the Super Bowl (at least, the Patriots had better hope they don't see much time in the Super Bowl), so throwing those three games out and picking three new ones would be a sound idea. The problem, though, is that while the offense was certainly worse without Brady than with him, it still wasn't bad, just average. The defense, meanwhile, was one of the worst in the league during Brady's suspension, then improved to a top-10 unit once Brady returned to the field. It's correlation, not causation, but it does make it hard to blame those early performances solely on the absence of Brady.

New England DVOA With/Without Tom Brady, 2016
Status Weeks Offense DVOA Rank Defense DVOA Rank
Brady Suspended 1-4 3.8% 13 12.0% 28
Brady Active 5-17 26.4% 1 -6.5% 10

Were we to wipe those first four games from the slate, we'd be dust-binning some valuable data about New England's defense, and what flaws they may have shown early in the season. At the same time, we already know that Brady is playing and that he makes the offense better and there's no point in evaluating the peaks and valleys of New England's backup quarterbacks. So rather than look at New England's worst four games overall, we're going to split this into two parts: the worst four games for New England's offense with Brady in the lineup, and the worst four games for New England's defense, whenever they may have occurred. This may yield some weird results, and they may not jibe with similar pieces we have run on Super Bowl teams of the past. But like we said: it was a weird season.

OFFENSE

By DVOA, here are New England's worst four games with Brady taking snaps:

  • Week 5: The Patriots beat the Browns 33-13 in Cleveland. While the Browns were alternating Charlie Whitehurst, Cody Kessler, and Terrelle Pryor at quarterback, Brady went 28-of-40 for 406 yards with three touchdowns, no interceptions, and just one sack. Chris Hogan and Rob Gronkowski both went over 100 yards receiving. Yes, this was one of New England's worst offensive games with Brady. That's partly because of an ineffective rushing attack (just 101 yards on 33 non-kneeldown carries, including three failures to convert with 1 or 2 yards to go), and partly due to opponent adjustments for playing Cleveland's defense, which finished next to last in DVOA.
  • Week 13: The Patriots jumped out to a 17-0 lead in Los Angeles and went on to beat the Rams 26-10. Brady completed more than 70 percent of his passes, going 33-of-46, but only gained 269 yards and one touchdown. He was still able to avoid being sacked or intercepted. Julian Edelman was the leading receiver with 101 yards. LeGarrette Blount finished with 88 yards, but nearly half of that came on a 43-yard touchdown run in the first quarter. He averaged 2.6 yards on each of his other 17 carries.
  • Week 15: The Patriots netted only 16 points in Denver, but that was more than enough to outscore a hapless Broncos offense and win 16-3. Brady completed only half of his passes, going 16-of-32 for 188 yards, with no touchdowns or interceptions and two sacks. Blount carried the ball 17 times and gained only 31 yards, with no carry gaining more than 5 yards. Dion Lewis, though, picked up the slack, with 18 carries for 95 yards. Edelman was the leading receiver, with six catches (in 12 targets) for 75 yards.
  • Week 16: The Patriots destroyed the Jets 41-3 in a game that wasn't as close as the final score would indicate. Brady went 17-of-27 for 214 yards with three touchdowns, no interceptions, and one sack. Blount (50 yards on 20 carries) and Lewis (52 on 16) spent most of the day racking up short gains and killing clock. Edelman was again the leading receiver, with five catches for 89 yards.

First takeaway: the Brady offense, at its worst, was still awfully good. Their total DVOA in those four games was 12.8% -- over the course of the season, that would have ranked seventh in the league, right behind New Orleans and right ahead of Oakland. In those four games, Brady completed 65 percent of his passes for 7.4 yards per throw, with seven touchdowns, no interceptions, and only four sacks. He had 311 DYAR in these games, which was actually in the top five among all quarterbacks in those weeks.

The rushing offense, though, was largely ineffective, with 489 yards on 135 non-kneeldown carries. That's a 3.6-yard average, and remember that 43 of those yards came on one run -- take out that play and the average drops to 3.3 yards per carry. A lot of that, though, has to do with context and game situations. In these four games, the Patriots had 68 runs for 281 yards in the first half (a 4.1-yard average), but 67 runs for 208 yards in the second half (3.1 yards per carry). All 67 of those second-half runs came with New England ahead; 62 of them came with the Patriots up by at least two scores. Their rush offense DVOA in the first half of these games was -4.8%, a number that dropped to -17.2% in the second half. Clearly, this Patriots offense is very happy just to grind out clock with a big lead late in games, but is this information useful to Atlanta? As game plans go, "lure the opposition into a false sense of security by falling behind early" wins lots of points for originality, but little for actually leading a team to victory.

The only individual offensive player who really had bad numbers in these games was Julian Edelman. He dominated the Patriots with 42 targets -- nobody else had more than 23 -- but only caught 24 balls for 300 yards, with a -7.1% DVOA. Behind Edelman, the numbers are a big mish-mash. Two wide receivers (Hogan and Malcolm Mitchell), a tight end (Martellus Bennett) and a running back (James White) each had at least 17 targets, and gained 125 to 179 yards. Hogan had his one big day against Cleveland, but gained only 63 total yards in the other three games. Nobody else had more than nine targets, so the bulk of the passing action in the Patriots' worst games was definitely funneled through these five receivers.

We can gain a little more insight by looking at Brady's DVOA when throwing to those players. This will help us uneven the playing field a bit by removing positional baselines, and just look which players gave Brady the best results, and which gave him the worst. And there we start to get some clarity. Brady's passing DVOA on throws to White was just 6.6%, and his DVOA on throws to Edelman was 23.7%. Meanwhile, Brady had a DVOA higher than 40.0% on throws to Hogan, Mitchell, and Bennett. If there's a glimmer of hope for Atlanta's defense here, it's that they can take away deep completions, and that New England will get over-conservative with short routes, dumpoffs, and runs. And even if all that happens, it will put pressure on an Atlanta defense that missed 122 tackles, one of nine defenses to miss at least 120 tackles this year according to Sports Info Solutions charting. (New England was another, suggesting again that this game will turn into a shootout.)

One more note of weirdness: In these four weeks when the Patriots offense was at its worst, the Patriots defense had a DVOA of -27.9%, best in the league by a wide margin. It sounds like a cliché, but even when Brady and the offense have struggled, the defense has shown a tendency to step up its game.

DEFENSE

By DVOA, here are the worst four games for New England's defense this year:

  • Week 1: The Patriots beat Arizona 23-21. New England sacked Carson Palmer three times, but otherwise he went 24-of-37 for 271 yards with two touchdowns and no interceptions. David Johnson gained 89 yards on just 16 carries, and another 43 yards on six catches. Larry Fitzgerald added eight catches for 81 yards and two scores.
  • Week 2: New England took a 31-3 lead in the third quarter, then let Miami come back to make it 31-24. The Dolphins got the ball back with 1:09 to go and no timeouts, but the game ended on a fourth-down incompletion at the New England 29. Ryan Tannehill finished 32-of-45 for 387 yards with two touchdowns and two picks, though he wasn't sacked. He also ran six times for 35 yards. Jarvis Landry and DeVante Parker both went over 100 yards.
  • Week 10: Seattle went into Foxborough and upended the Patriots 31-24. Russell Wilson completed 25-of-37 passes for 348 yards and three scores, with three sacks but no interceptions. The real star for Seattle, though, was rookie running back C.J. Prosise, who gained 66 yards on 17 carries, and also caught each of the seven passes thrown his way for 87 more yards.
  • Week 12: The Patriots needed a Malcolm Mitchell touchdown inside of the two-minute warning to beat the Jets 22-17. Ryan Fitzpatrick went 22-of-32 for 269 yards and two scores, with no interceptions and just one sack. Quincy Enunwa caught each of the five passes thrown his way for 109 yards and a touchdown, while Brandon Marshall added six catches for 67 yards and another score.

None of those quarterbacks ranked higher than 15th in DVOA, but the Patriots let them complete 68 percent of their passes for an 8.4-yard average, with 10 touchdowns and two interceptions. New England did register seven sacks in those games, but that hardly seems to have mattered.

More important than how well these quarterbacks threw against New England, though, is how frequently they threw. Accounting for spikes, penalties, and kneeldowns, these teams had 162 total dropbacks against the Patriots, but only 79 runs. New England's defense was much better against the run (-23.7% DVOA, fourth) than against the pass (13.9%, 23rd) this year, and the offenses that gave them the most trouble threw the ball more than twice as often as they ran it. Yes, that is partly because those teams were usually behind. But it's also because putting the ball in the air is the best way to beat New England. This is good news for Atlanta, a team that runs the ball a lot overall, but likes to pass early -- they were ninth in first-half pass plays this year, but 30th in second-half pass plays. To win the Super Bowl, they'll need to stay aggressive and pass late in the game too.

And when they pass, they'll have no shortage of targets to choose from. The Patriots often struggled to cover anyone, though they seemed to focus first on taking big plays away from opponents' top wide receivers.

New England Defense vs. Receivers in Four Worst Games, 2016
Position DVOA Tgt Yds Catch% PYD YAC
WR1 37.1% 40 342 75% 8.8 3.4
WR2 32.2% 33 322 56% 14.0 3.7
Other WRs 15.9% 27 260 54% 15.8 3.3
TE 43.5% 21 189 85% 7.8 4.9
RB 21.5% 30 209 80% 2.7 7.2
Weeks 1, 2, 10, and 12.

In their bad games, the New England defense gave up tons of catches to opposing No. 1 wide receivers, but those receptions were by and large on shorter routes. Meanwhile, all the other wide receivers were running low-percentage, high-reward deep routes through the Patriots secondary. There's a bit of a chicken-or-egg question to this -- is this the way opponents chose to attack the Patriots, or is this the way the Patriots defense forced offenses to play? -- but it's easy to visualize Julio Jones catching a lot of curls and crossers to draw coverage to himself, then Taylor Gabriel and Mohamed Sanu streaking by New England's lesser corners in one-on-one matchups for big plays. This is another reason Atlanta will need to go pass-wacky against the Patriots -- the more often they drop back to pass, the more chances they will have to hit on those low-percentage throws. This also jibes with what we wrote about Atlanta last week: the Super Bowl will likely decided by New England's ability to cover Taylor and Sanu.

The Falcons offense was the NFL's best in 2016 and should have a big edge over New England's defense, but the Patriots offense was second-best and should enjoy an even bigger edge against Atlanta's defense. In a game where the first team to punt likely loses, Atlanta will need to use all of its weapons to the best of their abilities to bring home the Lombardi Trophy.

Posted by: Vincent Verhei on 31 Jan 2017

19 comments, Last at 03 Feb 2017, 3:59pm by nat

Comments

1
by billprudden :: Tue, 01/31/2017 - 3:22pm

Mario Manningham and David Tyree approve of high-variance downfield throws!

(and neither was half the receiver Gabriel and Sanu are)

2
by nat :: Tue, 01/31/2017 - 3:26pm

I was wondering how you were going to find four weak Patriot games to talk about, with just two losses to choose from. Taking the four weakest for each of the two main units, and looking only at the offense's games with Brady in the lineup makes sense.

Perhaps splitting this kind of analysis by unit should be the standard for future years as well.

8
by RBroPF :: Tue, 01/31/2017 - 7:27pm

It makes sense in this case only because the weird season the Patriots had made every other approach make less sense. I don't agree that it's a good idea in general, because offense and defense do not play the game independently of each other. What happens on one side of the ball affects the other side of the ball as well.

In the Miami game, the defense had only allowed 3 points by mid-way through the 3rd quarter. Then Garropolo went down and the defense went to pieces. Watching the game, that didn't feel like a coincidence. And when the defense is dominating the opposing offense, the offense plays more conservatively. That's smart strategy.

3
by sbond101 :: Tue, 01/31/2017 - 3:28pm

This article kind of reads like a write up on why DVOA sometimes stinks as a metric evaluating football team performance rather than how to beat the Patriots. 3 of the four "bad" offensive games were situations in which the Pats built a big lead and then played low variance inefficient football to maximize the chances that they would win the game. That sounds like pretty good football to me. The defensive side of the discussion seems a lot more productive since only one of those results makes me question the usefulness of DVOA (the Miami game, where they played awful on defense in the second half after building a massive lead in the first half and then having Jimmy G get hurt).

The other side of it though is that the Jets, Seattle, and Arizona basically did what Pittsburg did in the conference championship, the only difference being in the Pittsburg game the Pats managed to capitalize on their chances for Picks where they failed to execute earlier in the season. I guess that speaks to how in against a pass-heavy strategy a your success on the couple chances you get to grab the football are sometimes the difference between the a great game and an awful one. Sometimes I'm glad I don't play corner in the NFL, if it weren't for all the money it would really be a terrible job.

4
by Aaron Brooks Go... :: Tue, 01/31/2017 - 4:35pm

The Jets game was one of their bad DVOA games?

Jesus, how bad were the Jets that a 41-3 game was a bad performance?

5
by nat :: Tue, 01/31/2017 - 4:49pm

It occurs to me that these articles should be using VOA, not DVOA. Using DVOA, the "answer" may turn out to be "be a very bad team, so opponent adjustments make the Patriots (or whoever) look bad".

Which isn't what we're looking for here at all.

We want to know what the teams that were the most successful in absolute terms (VOA) did. What was their scheme? What were their strengths? What weaknesses were they able to exploit? And which of their own weaknesses could they hide?

There's a place for DVOA, too. It can help you find the teams that did better than expected. But VOA should be the first stat you look for in this case.

7
by Eddo :: Tue, 01/31/2017 - 6:23pm

Aha, yes! I couldn't quite put my finger one what was bothering me about these types of articles, and you're right - the more important factor in winning is to hold the opponent to as low a VOA as possible (not necessarily DVOA - I can do that by making my own team as bad as possible in the games leading up to playing said opponent).

10
by Rob Eves :: Tue, 01/31/2017 - 11:06pm

Well said; I have a consistent issue with DVOA in that the 'D' part tends to trump the rest of it. As a poster above alluded to, how can a 41-3 win be a 'bad' performance? OK so the Jets are bad and it's not the same as beating a playoff contender by a similar margin, but you can only beat what's in front of you right? What sort of numbers does a team have to put up against a bottom-5 ranked DVOA team to come out with a good rating?

Nevertheless I'm sure more intelligent people than myself will have good justification for why it's calculated exactly the way it is.

11
by tballgame :: Wed, 02/01/2017 - 10:25am

I don't think it is just opponent adjustment. There is also the matter of running low variance plays against a weaker opponent where the opponent would benefit from high variance tactics. Against the Rams, Browns, and Jets, you don't want to get too creative or elaborate. Similarly, against the Broncos, it is hard to envision a loss to the Broncos the way their offense was playing without Denver's defense getting turnovers. A game plan that includes conservative offensive play calling will slow down the offense.

The Super Bowl will not be like any of these games. The only top ten offenses (by DVOA) that the Pats played this year were Buffalo and Pittsburgh (once without Ben and once without a healthy Bell). The defense will be in bend but don't break mode, giving up short plays and preventing big plays, hold the Falcons to field goals.

Having an early lead will be very important. It is easier to be patient and manage low risk plays with a lead. I also think the Pats defense will pay more attention to the Falcons' WR2 and WR3 between the 20s than the running backs, so I would expect both Falcons' running backs to have a big role in the game.

12
by sbond101 :: Wed, 02/01/2017 - 10:55am

Exactly, the Pats "bad" offensive games from an offensive DVOA point of view (DVOA roughly equates to efficiently moving the ball and scoring on offense) were mostly about their strategy to maximize their chances of winning. In reality the worst offensive game the Pats played down the stretch was probably vs. the Seahawks (I'm assuming were excluding the Huston wildcard game form the analysis). That was really the only game where it was obvious watching it that the defense was preventing the Pats from executing as they intended, in view of this it makes sense that when the Pats were faced with a long "gotta have it" situations at the end of the game they weren't able to successfully execute enough of them.

6
by ClavisRa :: Tue, 01/31/2017 - 4:52pm

McCourty is going to have a pick in the first half if Ryan feels like testing too many downfield throws early.

The Falcons leading receivers were Jones and their running back (as a combined entity). Almost without a doubt a big component to Atlanta's game is going to be getting the ball to the running backs in space. The Patriots should anticipate this and their schemes should account for it right from the gun. The Pats are a very strong tackling team, and if they stymie some early attempts to get the ball to Freeman in space, Ryan will try to find advantageous matchups downfield.

Another important component to the Falcons plan will be Julio as a decoy. The high risk play is to try to hit Gabriel over the top when Jones is pulling safety help to the other side of the field. If Atlanta is wise, they will threaten downfield, but then take the deep out, take chunks of yards without the homeruns

9
by Anon Ymous :: Tue, 01/31/2017 - 7:56pm

It gives me hope that the one legitimately bad defensive performance - 31 points to Seattle in Foxboro - has some reasonable qualifications. First off, that was against a mobile QB, where NE historically values containment over pressure. Intentionally not taking advantage of Seattle's biggest weakness really backfired when the coverage was a mess in that game. Then you have the fact that this was the first game after Jamie Collins' trade, which was likely on the players' minds and took several weeks for a LB rotation to emerge out of. Prosise was a beneficiary of this uncertainty, gaining most of his yards against the now primarily run defending Elandon Roberts.

13
by Vincent Verhei :: Wed, 02/01/2017 - 6:05pm

A few thoughts on using VOA instead of DVOA to determine a team's worst games:

* By VOA, only one of the four worst games for the Brady offense would change: we would remove the Cleveland game and replace it with the Week 10 loss for Seattle. Yes, that 41-point outing against the Jets is still one of their worst games of the year with Brady, even without opponent adjustments. They punted four times in that game (twice in the first half), they were fortunate to recover their one fumble, five of their seven scoring drives started at their own 46 or deeper, they only went 11 of 18 on third downs. Now, mind you, this is still a very good offensive day! Their DVOA was 18.2%, their unadjusted VOA was 22.5%. Only two teams had full-season offensive VOAs higher than 22.5%, and they're both playing in the Super Bowl. It's just very, very, VERY hard to find games for the Brady offense this year.

I just did some VERY quick and dirty work with the drive finder at PFR. New England had 11 non-kneeldown drives in that Jets game. Based on field position alone, an average offense playing against an average defense would have scored about 27 points that day. Yes, New England was still two touchdowns better than that, but they still got a LOT of field position help in that game.

* Similarly, only one game changes for the defense by using VOA: you would take out the Week 12 Jets game, and replace it with the Week 8 game, a 41-25 win over Buffalo when the Bills ran for 167 yards.

14
by nat :: Wed, 02/01/2017 - 9:11pm

It's not too surprising that only a couple of games swap out when you go by VOA. VOA is still the right stat for this purpose. It's more about modeling the proper use of the stats then it is getting the a different answer.

I poked around in that Jets game, too. Foremost, we have to accept that it was a decent day, just not as good as most of the Brady-led offense's days. But they certainly could have done more with the field position they had. It was the defense's victory. Or maybe it was the Jets offense's loss.

The Bills game is a good one to include for the defense. It highlights an interesting dilemma: if the Pats invite you to run all over them, should you take the invitation?

15
by Vincent Verhei :: Wed, 02/01/2017 - 10:37pm

VOA is still the right stat for this purpose. It's more about modeling the proper use of the stats then it is getting the a different answer.

I disagree. I don't think it's cut and dry. I think both stats have their pros and cons. If the flaw in DVOA is that it will under-rate New England's performance against bad defenses, then the flaw in DVOA is that it would give us less useful info. "Have a defense as good as Seattle's" is not a strategy Atlanta can use right now.

The Bills game is a good one to include for the defense. It highlights an interesting dilemma: if the Pats invite you to run all over them, should you take the invitation?

This is a good question, and I think it goes for both defenses. If you're Atlanta, and you give up a 6-yard run, don't you shrug and say "probably better for us than a Brady pass would have been?"

16
by nat :: Thu, 02/02/2017 - 11:03am

The title of your article is "Keys to Beating New England". So you need to find the teams that did the best (on each side of the ball) at doing just that. Adjusting for expectations (which is what DVOA does) isn't the way to go. It doesn't reduce the lessons to learn that a team was also good at beating the rest of their schedule. If a team did badly against New England, it doesn't really help their story that they did even worse against the rest of their schedule.

I agree that "Have a defense as good as Seattle's" isn't a strategy. But "study the successful schemes and match ups that Seattle used" is a good start for getting a strategy. It's a much better piece of advice than "suck less than usual" and "copy the teams that sucked less than usual", which is what you get by using DVOA instead of VOA.

Now, I'll admit that the opening text of your article seems to switch the emphasis to looking at what New England did wrong (or at least less well), rather than what the opponents did right. For that purpose, DVOA is a fine stat, because adjusting for expectations makes sense there. But "Hope that New England screws up" is even less of a strategy than "Suck less than usual".

It comes down to this distinction:

1) Common threads to New England under-performing expectations (use DVOA), hoping that New England under-performs in the same way in the Super Bowl
vs.
2) Common threads to your team beating New England (use VOA), trying to emulate the teams that did well against New England during the season

The title promises (2), but using DVOA really only delivers on (1).

17
by Bright Blue Shorts :: Thu, 02/02/2017 - 3:57pm

If he changes the title would it become an acceptable article?

18
by MC2 :: Fri, 02/03/2017 - 12:36am

I agree that "Have a defense as good as Seattle's" isn't a strategy. But "study the successful schemes and match ups that Seattle used" is a good start for getting a strategy.

Don't worry. The Falcons are already doing all they can to copy Seattle's defense. (Unsuccessfully, of course.)

19
by nat :: Fri, 02/03/2017 - 3:59pm

I'll bet they are.

Still, it's a better idea than trying to copy the Browns' defense, I think we'd all agree. You can see that VOA points to a more useful game than DVOA in that particular case.