Week 6 DVOA Ratings

by Aaron Schatz

The Arizona Cardinals blew a 10-3 lead against the Pittsburgh Steelers this week and went home with a 25-13 loss. And with that strong loss, the Arizona Cardinals climbed back into the No. 1 spot in the Football Outsiders DVOA ratings.

That looks strange, doesn't it? Don't worry, it looks strange to us too, but it was a very strange week.

It's probably wrong to say that the Cardinals climbed into the No. 1 spot in DVOA with a loss. It's more accurate to say that the New England Patriots slipped behind the Cardinals with a very close win. Both Arizona and New England see their DVOA ratings fall from a week ago, but New England's rating falls more for two reasons. First, Arizona's single-game rating for a loss to No. 6 Pittsburgh (11.4%) is higher than New England's single-game rating for a win over No. 19 Indianapolis (5.3%). Second, Arizona adds these plays to data set that included five other games worth of plays. New England adds their plays to a data set that included only four other games worth of plays. With a smaller sample, New England's good-not-great performance drags the Patriots' rating down more than Arizona's good-not-great performance drags their rating down.

These results were part of a week that showed how a team's performance can look different depending on the data you are using. There are a lot of ways to determine how well a team played on Sunday. Obviously, the most important variable is the binary one called "winning the game." That's what determines who gets to go to the playoffs and who wins the Super Bowl. But when we are trying to get a more accurate idea of how good a team really is and how good it is likely to be in the future, it's more accurate to look at points scored and allowed. Even more accurate than that is performance broken down at the play-by-play level. Usually, these two methods will agree. The team that scored more points usually was better at the play-by-play level, and when the scoring gap is bigger, the play-by-play gap is usually bigger. But that's not always the case, and we had a lot of those games this week. There were also a lot of close games where opponent adjustments meant the losing team rose in DVOA while the winning team fell.

I'm going to take a look here at each of these games. I'm listing both DVOA and VOA for each game. However, as with Any Given Sunday each week, the VOA listed here only removes opponent adjustments. Unlike the VOA rating listed on the team DVOA pages, this does not also remove weather/altitude adjustments to special teams, and fumbles are still penalized no matter who recovers.

We'll start with the game covered by Andrew Healy in Any Given Sunday. Arizona outgained the Steelers 469-310, but lost on a series of mistakes. Arizona committed the game's only three turnovers and racked up 111 yards in penalties, more than twice as many as Pittsburgh. Right now, the DVOA system does not include most of those penalties, including such things as offensive pass interference, which wiped out a Michael Floyd touchdown right before the half. (Every couple years I go through and test to see if I can figure out a good way to add more penalties to make the DVOA system more accurate, so a chance there is always a possibility for the future.)

A close game where the losing team ends up slightly ahead in DVOA because of the opponent adjustments. In addition, Denver recovered both fumbles in this game. One was a muffed punt, which isn't penalized much because it is usually recovered by the punt return team. The other was on a sack of Josh McCown.

Here's another close game where the losing team ends up ahead in DVOA because of the opponent adjustments. This is closer than it seemed for two reasons. First, onside kicks are not included in the special teams rating because they are so rare, and the Patriots were handed great field position in what was essentially a 50-50 toss-up of the officials determining who had recovered the Colts' surprise onside kick in the second quarter. Second, the backdoor cover touchdown by the Colts means just as much as any other touchdown in DVOA; it was indicative of the quality of the Colts offense and the Patriots defense because the game was still winnable by Indianapolis. The weird failed fake punt is counted in the current version of DVOA as just a rushing attempt on fourth down. (Eventually, I hope to remove fake punts and field goals from the regular pass/run stats.)

This one is odd. Normally, you would expect the Lions to have a better rating when they averaged 7.2 yards per play to just 6.1 yards per play by the Bears. And the Lions do have a slightly better rating, without considering special teams. The difference in special teams comes because the Lions muffed two punts and had zero punt return yards while the Bears got actual yardage on most of their punt returns. And the Bears recovered those muffed punts; doing that twice in a game is pretty rare. But the field-position advantage didn't seem to actually translate into more points for Chicago.

This game was pretty much a tie between two good teams. Differences in opponent adjustments are generally caused by run/pass ratios. Yes, Seattle is still in the DVOA top ten despite being 2-4. The losses have been very close, and the system is probably giving the Seahawks too much credit for destroying the Chicago Bears without Jay Cutler in the lineup.

Here's another one where the difference in VOA is in special teams, and that's almost all because of great punts by Sam Koch. The 49ers gained 7.1 yards but only achieved success on 29 percent of their plays according to our baselines. Most of their offense was three huge pass plays over 50 yards, which tends to be less sustainable over the long term. The Ravens had just 5.6 yards per play but a 53 percent success rate.

This DVOA result, combined with the big loss from Atlanta, helped Baltimore climb into the No. 13 spot in DVOA despite the 1-5 record. The Ravens do have a negative rating, though. Only 12 of 32 teams right now have a positive DVOA. As you would imagine given that there are five undefeated teams and no winless teams, the league looks very top-heavy this year.

I wanted to run this one to show that not all close games are close in DVOA. The Packers gained 7.6 yards per play compared to just 6.2 for San Diego, but the Chargers ran 89 plays and the Packers only 49!

This is the craziest, kookiest result of all. These ratings look so insane that I had to go back a couple times to make sure I had not made any huge errors. But on a play-by-play basis, the Chiefs completely outplayed the Vikings despite losing 16-10. Kansas City gained 5.8 yards per play with a 47 percent success rate. Minnesota gained only 4.7 yards per play with a 33 percent success rate. However, the Chiefs failed on fourth down twice and lost a fumble while the Vikings recovered their own fumble. The Chiefs had 95 yards of penalties while the Vikings had only 50, and 15 of those were a taunting penalty after the game was effectively over once the Chiefs failed on their final fourth-down attempt. It's also remarkable how much the Chiefs did everything after halftime. Kansas City had 2.6 yards per play and -57.1% DVOA before halftime, then 7.5 yards per play and 35.5% DVOA after halftime. It wasn't enough. But this crazy lopsided "wrong team wins" game is a big reason why the Vikings are by far the lowest winning team, ranked just 26th in total DVOA right now.

* * * * *

Of course, it's extra strange to have a 4-2 team at No. 1 in a year where there are still five undefeated teams after six weeks. In the past couple of weeks, DVOA commentary articles have reported on the odds of teams going undefeated, and you may have noticed that we had those odds much higher than any other site that does similar simulations. This week we realized why our odds of an undefeated team were artificially high, and we've made moves to change that.

This year's playoff odds report, going all the way back to the odds given in Football Outsiders Almanac 2015, have been using what we call a "dynamic" simulation. Historically, our simulations had always ended with too many teams finishing near 8-8, with far higher frequency than the real NFL. Part of the problem was that our previous simulations never accounted for the fact that a win often indicates that a team is better than we thought they were going into the game. The new dynamic simulation added 2.0% DVOA to any team after a win, and subtracted 2.0% DVOA after a loss. This change meant that a team that began the season with a projected DVOA of 20.0% but went 14-2 would end up with a DVOA rating much closer to what we usually see from 14-2 teams.

We continued to include this dynamic aspect in the simulations as we got into the regular season. The problem, however, is that actual DVOA ratings have a much larger spread between low and high during the season than they have in preseason projections. This is particularly true early in the season, although incorporating preseason projections in the DAVE ratings blunts the effect somewhat. However, now a long winning streak wasn't adding 2.0% DVOA each week to a team that started out with a projection around 20%. Sometimes a long winning streak was adding 2.0% DVOA each week to a team that started out in the simulation with a DAVE rating of 41.3%, such as the New England Patriots a week ago. As a result, teams that were projected to start 11-0 or 12-0 were getting into the last month of the season with the simulator projecting their chances of winning each game as if they were better than the 2007 Patriots or 1991 Redskins. Thus, the odds we were getting for teams to go 16-0 were artificially high.

We've corrected this problem this week by altering the dynamic simulation. No longer do all teams get 2.0% DVOA added after a win. Now, the amount of DVOA added after a win is dependent on what the chances of a win were before the game, with between 0% and 3% DVOA added. For example, let's imagine that the playoff odds simulator gave the Patriots a 67 percent chance to beat Indianapolis last week. In that case, a win for the Patriots would raise their DVOA by 1.0% and lower the Colts' DVOA by the same 1.0% before the simulator did the projections for the following week. However, a Colts win would raise their DVOA by 2.0% and lower the Patriots' by 2.0% instead. Underdogs that win get a bigger boost to DVOA than favorites.

In the long run, of course, a better simulator would simulate not just wins and losses but also how often teams would see their DVOA ratings change in different ways after a win or a loss. However, we won't have time to test and program anything like that until next offseason. For now, we think this fix does a good job of giving a much more accurate picture of how likely teams are to have a 16-0 season. Odds of making the playoffs or winning the Super Bowl don't change very much either way; this really only affected how often the simulations were ending with teams having records on the extremes.

The new playoff odds report uses this new method and the "CHANGE" column is based on the numbers from a simulation of Week 5 that also used this new method, not from the simulation actually posted on the site last week. Our total odds of at least one team going 16-0 are now listed at 11.0 percent. The odds after Week 5, which we listed on the site as 16.4 percent last week, now would be 10.1 percent. Here's a look at how the new method gives the odds for a perfect regular season after Week 4, Week 5, and right now:

Those numbers use the DAVE ratings as they looked as of each of these weeks; the Patriots chances of a perfect season have dropped between Week 5 and Week 6 because their overall DVOA (and DAVE) rating has dropped.

* * * * *

Once again in 2015, we have teamed up with EA Sports to bring Football Outsiders-branded player content to Madden 16 Ultimate Team. Each week, we'll be picking out a handful of players who starred in that week's games. Some of them will be well-known players who stood out in DVOA and DYAR. Others will be under-the-radar players who only stood out with advanced stats. We'll announce the players each Tuesday in the DVOA commentary article, and the players will be available in Madden Ultimate Team packs the following weekend. We will also tweet out images of these players from the @fboutsiders Twitter account on most Fridays. One player each week will only be available for 24 hours from the point these players enter packs on Friday.

The Football Outsiders stars for Week 6 are:

• QB Philip Rivers, SD (24-HOUR HERO): Led all quarterbacks with 259 DYAR, best single game by a quarterback so far this season. 503 passing yards against the No. 4 DVOA pass defense in Green Bay.
• LOLB Shaquil Barrett, DEN: 1.5 sacks, forced fumble, PD, and three run TFL. 8 of his 10 defensive plays stopped offenses short of a successful play.
• RG Willie Colon, NYJ: No sacks allowed. Helped Jets running backs gain 88 yards on 12 carries up the middle. Key blocks included a 23-yard screen pass to Chris Ivory and a 54-yard Ivory run.
• WR DeAndre Hopkins, HOU: Fourth among all wide receivers in Week 6 with 59 DYAR (10-for-15, 148 yards, 2 TD).
• ILB Christian Kirksey, CLE: 9 total tackles including 3 that prevented conversions after third-down completions.

* * * * *

All stats pages are now updated with Week 6 information including FO Premium, snap counts and playoff odds.

* * * * *

These are the Football Outsiders team efficiency ratings through six weeks of 2015, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)

OFFENSE and DEFENSE DVOA are adjusted for strength of schedule and to consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season. Because it is early in the season, opponent adjustments are only at 60 percent strength; they will increase 10 percent every week through Week 10. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE. 

DAVE is a formula which combines our preseason projection with current DVOA to get a more accurate forecast of how a team will play the rest of the season. Right now, the preseason projection makes up 19 percent of DAVE for teams with six games played, and 27 percent of DAVE for teams with five games played.

To save people some time, please use the following format for all complaints:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>

• NON-ADJUSTED TOTAL DVOA does not include the adjustments for opponent strength or the adjustments for weather and altitude in special teams, and only penalizes offenses for lost fumbles rather than all fumbles.
• ESTIMATED WINS uses a statistic known as "Forest Index" that emphasizes consistency as well as DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close. It then projects a number of wins adjusted to a league-average schedule and a league-average rate of recovering fumbles. Teams that have had their bye week are projected as if they had played one game per week.
• PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
• FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
• VARIANCE measures the statistical variance of the team's weekly DVOA performance. Teams are ranked from most consistent (#1, lowest variance) to least consistent (#32, highest variance).