Week 12 DVOA Ratings
by Aaron Schatz
The Denver Broncos remain No. 1 in the Football Outsiders DVOA ratings this week, and their overall DVOA rating actually gets a nice boost from last week after a nice win over Miami. But in another very important fashion, the Broncos finally fell from the top spot this week. They are no longer our projected No. 1 favorite to win Super Bowl XLIX. That spot now belongs to the New England Patriots.
The Patriots' dominating win over Detroit has moved them up to the No. 2 spot in overall DVOA, trailing only the Broncos. They rank even better in weighted DVOA, because this week their Monday night debacle in Kansas City begins to count less in the weighted DVOA formula. The last three Patriots wins are their three best games of the year by single-game DVOA, and a week-to-week graph of their performance this season shows a very steady and obvious slope of improvement with one big, glaring outlier in Week 4.
Early in the season, the Patriots looked like they had a strong defense, but that was exaggerated by the fact that we had not started using opponent adjustments yet. But over the last few weeks, the Patriots finally have had an actual strong defense. (If we split the season into three four-week periods, the Patriots defense would rank 10th in Weeks 1-4, 28th in Weeks 5-8, and then third in Weeks 9-12 behind only San Francisco and Arizona.) This week, the Patriots move into the top ten for defensive DVOA, making them one of two teams that currently ranks in the top ten for all three phases of the game. (The other is Baltimore.)
As I detail in today's ESPN Insider playoff odds report, the Broncos had been our Super Bowl favorite pretty much straight through since the first preseason projections that I did after the NFL Draft in May. The only exception was a two-week period in late September when we had the Cincinnati Bengals as our favorites, in between the Denver loss to Seattle and the Cincinnati loss to New England. But with the improvement in the Patriots' weighted DVOA, New England now wins the No. 1 seed in 46.4 percent of our simulations and the Super Bowl in 18.7 percent of our simulations. Denver is second at 17.0 percent, Green Bay third at 14.5 percent. No other team is above 7.5 percent. Incidentally, the Super Bowl XXXI rematch of New England and Green Bay that takes place this Sunday afternoon is also our most likely "special" Super Bowl, showing up in 9.0 percent of simulations.
(Also, the ESPN article has an error where I mistakenly referred to Baltimore as the only team in the top ten for all three phases of the game. I wrote that before the Bills-Jets game changed opponent adjustments and moved the Patriots defense into the top ten as well.)
 |
 |
Another interesting note about the Broncos: Denver had ranked first or second in both offense and defense for six of the past seven weeks, but the Broncos drop to fourth in defensive DVOA this week. However, they move back into the No. 1 spot ahead of Green Bay on offense. New England's move into position as our Super Bowl favorite is truly more about the Patriots rising than the Broncos falling. Although tears have been shed and garments rended for the struggles of the Broncos in recent weeks, DVOA suggests we're only really talking here about two bad games, neither of which was really that terrible.
One extra reason we know that New England and Denver have been good this year as that they have had to play strong division rivals... and each other's strong division rivals. Buffalo's "Practice? Who Needs Practice?" stomptacular over the New York Jets gives the AFC East three different teams in this week's DVOA top ten. The whole division isn't great, of course; the Jets are now No. 28 with what I believe is the first below-average defense in Rex Ryan history, ranked No. 19. The AFC West is not quite as strong as the AFC East because of the recent decline of the Chargers, but that division still has the Broncos at No. 1, the Chiefs at No. 7, the Chargers at No. 16, and the "underrated because they aren't the worst team in the NFL" Raiders now at No. 26, moving up four spots after their first win of the season. The AFC North may be the more balanced division, and everyone may end up with a winning record because they all get to play the NFC South, but Baltimore at No. 4 is the only AFC North team in our top dozen.
And oh, the sad, sad NFC South. While the Saints are at least respectable if you consider them only by DVOA (No. 18), we're still stuck in a situation where the NFC South champion now has a losing record in 63.0 percent of simulations and goes 6-9-1 or worse in 15.9 percent of simulations.
Two quick notes on the playoff odds report. First, we realized that we were not listing the Carolina/Cincinnati tie in either team's Projected Mean Wins number. So beginning this week, each team has 0.5 added to its Projected Mean Wins to represent the tie. Second, we've been debating around here whether we should be giving the Arizona Cardinals home-field advantage for the Super Bowl. We finally decided that we would give them "half advantage," which is similar to what Ken Pomeroy does when he does college basketball projections and a team is hosting during March Madness. We weren't able to get that programmed into our simulator in time for this week but beginning next week the Cardinals will get a small boost for playing the Super Bowl in their home stadium.
* * * * *
 |
Once again in 2014, we have teamed up with EA Sports to bring Football Outsiders-branded player content to Madden 15 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 12 are:
- WR Odell Beckham, NYG (24-HOUR HERO): Led all WR in Week 12 with 76 receiving DYAR (10-for-11, 146 yards, 2 TD).
- RE Chris Clemons, JAC: 3 sacks and a forced fumble in just the first quarter alone.
- K Steven Hauschka, SEA: 4-for-5 on field goals; all six kickoffs into the end zone including four touchbacks.
- MLB Rolando McClain, DAL: Led NFL defenders with 10 stops in Week 12 including 8 run tackles for a combined 0 yards and the tackle to prevent Giants conversion on fourth-and-2 and end the game.
- RT Louis Vasquez, DEN: Allowed no sacks or QB hits vs. Cameron Wake; Broncos had 77 rushing yards on seven runs to the right side.
The elite Odell Beckham has a 94 overall rating and a 100 spectacular catch rating.
* * * * *
All stats pages are now updated with Week 12 information -- or will be in the next few minutes -- including FO Premium, snap counts, and playoff odds. You can also read the new weekly playoff odds report on ESPN Insider to get more commentary on the current playoff odds.
* * * * *
[ad placeholder 3]
These are the Football Outsiders team efficiency ratings through 12 weeks of 2014, 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. WEIGHTED DVOA represents an attempt to figure out how a team is playing right now, as opposed to over the season as a whole, by making recent games more important than earlier games. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.
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>
| TEAM | TOTAL DVOA |
LAST WEEK |
WEIGHTED DVOA |
RANK | W-L | OFFENSE DVOA |
OFF. RANK |
DEFENSE DVOA |
DEF. RANK |
S.T. DVOA |
S.T. RANK |
|
| 1 | DEN | 35.6% | 1 | 35.0% | 1 | 8-3 | 26.1% | 1 | -14.3% | 4 | -4.9% | 29 |
| 2 | NE | 25.7% | 5 | 29.7% | 2 | 9-2 | 17.1% | 4 | -1.7% | 10 | 6.9% | 3 |
| 3 | GB | 24.5% | 2 | 25.4% | 3 | 8-3 | 24.6% | 2 | -0.1% | 12 | -0.2% | 15 |
| 4 | BAL | 22.4% | 3 | 23.2% | 4 | 7-4 | 10.1% | 10 | -4.3% | 9 | 8.0% | 2 |
| 5 | MIA | 19.1% | 4 | 21.3% | 5 | 6-5 | 12.3% | 7 | -10.6% | 6 | -3.8% | 24 |
| 6 | SEA | 18.5% | 7 | 16.0% | 6 | 7-4 | 12.0% | 8 | -8.2% | 7 | -1.8% | 20 |
| 7 | KC | 14.2% | 6 | 14.2% | 7 | 7-4 | 10.3% | 9 | 2.0% | 18 | 5.8% | 4 |
| 8 | PHI | 12.4% | 8 | 13.3% | 8 | 8-3 | -2.2% | 16 | -5.3% | 8 | 9.4% | 1 |
| 9 | BUF | 9.3% | 12 | 9.6% | 9 | 6-5 | -11.1% | 26 | -14.9% | 3 | 5.4% | 6 |
| 10 | DAL | 8.5% | 13 | 8.9% | 10 | 8-3 | 14.5% | 5 | 6.1% | 25 | 0.1% | 13 |
| 11 | SF | 7.6% | 11 | 8.2% | 11 | 7-4 | -3.9% | 18 | -16.6% | 2 | -5.0% | 30 |
| 12 | IND | 6.9% | 14 | 6.8% | 13 | 7-4 | 3.1% | 13 | 2.0% | 17 | 5.8% | 5 |
| 13 | PIT | 6.8% | 10 | 7.4% | 12 | 7-4 | 17.8% | 3 | 9.7% | 29 | -1.3% | 19 |
| 14 | DET | 4.6% | 9 | 3.7% | 14 | 7-4 | -10.0% | 24 | -20.5% | 1 | -5.9% | 32 |
| 15 | CIN | 3.5% | 17 | 0.8% | 17 | 7-3-1 | -0.4% | 15 | 1.3% | 16 | 5.2% | 7 |
| 16 | SD | 1.3% | 18 | 0.2% | 18 | 7-4 | 9.6% | 11 | 7.9% | 27 | -0.4% | 17 |
| TEAM | TOTAL DVOA |
LAST WEEK |
WEIGHTED DVOA |
RANK | W-L | OFFENSE DVOA |
OFF. RANK |
DEFENSE DVOA |
DEF. RANK |
S.T. DVOA |
S.T. RANK |
|
| 17 | ARI | 1.2% | 15 | 1.0% | 16 | 9-2 | -9.6% | 23 | -12.6% | 5 | -1.8% | 21 |
| 18 | NO | 0.9% | 16 | 2.0% | 15 | 4-7 | 14.1% | 6 | 14.1% | 31 | 0.8% | 12 |
| 19 | CLE | -3.7% | 19 | -4.5% | 19 | 7-4 | -3.3% | 17 | 0.3% | 13 | -0.1% | 14 |
| 20 | ATL | -6.5% | 20 | -9.0% | 20 | 4-7 | 6.7% | 12 | 16.5% | 32 | 3.3% | 8 |
| 21 | HOU | -9.7% | 21 | -9.2% | 21 | 5-6 | -4.8% | 19 | 0.6% | 15 | -4.3% | 27 |
| 22 | CHI | -9.8% | 23 | -12.7% | 22 | 5-6 | 0.0% | 14 | 4.2% | 23 | -5.6% | 31 |
| 23 | MIN | -13.8% | 26 | -13.2% | 24 | 4-7 | -13.7% | 28 | 2.6% | 20 | 2.5% | 9 |
| 24 | STL | -15.1% | 24 | -13.1% | 23 | 4-7 | -12.3% | 27 | 4.1% | 22 | 1.4% | 10 |
| 25 | NYG | -15.8% | 25 | -18.1% | 26 | 3-8 | -6.7% | 20 | 8.2% | 28 | -0.9% | 18 |
| 26 | OAK | -17.9% | 30 | -17.1% | 25 | 1-10 | -18.2% | 30 | 0.6% | 14 | 0.9% | 11 |
| 27 | WAS | -18.0% | 28 | -19.6% | 27 | 3-8 | -8.4% | 21 | 5.2% | 24 | -4.4% | 28 |
| 28 | NYJ | -19.3% | 22 | -19.9% | 28 | 2-9 | -16.6% | 29 | 2.5% | 19 | -0.2% | 16 |
| 29 | CAR | -19.7% | 29 | -20.3% | 29 | 3-7-1 | -10.1% | 25 | 7.3% | 26 | -2.2% | 22 |
| 30 | TEN | -21.8% | 27 | -21.2% | 30 | 2-9 | -8.9% | 22 | 9.9% | 30 | -3.1% | 23 |
| 31 | JAC | -30.4% | 31 | -27.7% | 31 | 1-10 | -27.1% | 32 | -0.8% | 11 | -4.1% | 26 |
| 32 | TB | -31.3% | 32 | -28.5% | 32 | 2-9 | -23.7% | 31 | 3.6% | 21 | -3.9% | 25 |
- 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).
| TEAM | TOTAL DVOA |
W-L | NON-ADJ TOT VOA |
ESTIM. WINS |
RANK | PAST SCHED |
RANK | FUTURE SCHED |
RANK | VAR. | RANK | |
| 1 | DEN | 35.6% | 8-3 | 32.2% | 9.5 | 1 | 3.8% | 7 | 2.1% | 15 | 13.5% | 10 |
| 2 | NE | 25.7% | 9-2 | 24.9% | 7.2 | 5 | 2.9% | 9 | 7.0% | 4 | 19.4% | 25 |
| 3 | GB | 24.5% | 8-3 | 27.3% | 7.4 | 4 | -2.8% | 24 | 0.4% | 18 | 16.1% | 17 |
| 4 | BAL | 22.4% | 7-4 | 27.8% | 7.7 | 2 | -5.0% | 25 | -4.7% | 22 | 8.9% | 6 |
| 5 | MIA | 19.1% | 6-5 | 21.1% | 7.5 | 3 | 6.0% | 3 | -0.9% | 21 | 19.1% | 24 |
| 6 | SEA | 18.5% | 7-4 | 15.1% | 7.1 | 6 | -0.1% | 18 | 2.8% | 12 | 16.0% | 16 |
| 7 | KC | 14.2% | 7-4 | 13.0% | 6.9 | 8 | 3.9% | 6 | 5.4% | 6 | 23.1% | 30 |
| 8 | PHI | 12.4% | 8-3 | 13.5% | 7.0 | 7 | -8.2% | 30 | 0.3% | 19 | 15.3% | 15 |
| 9 | BUF | 9.3% | 6-5 | 9.1% | 5.8 | 17 | 1.1% | 14 | 12.9% | 2 | 7.0% | 4 |
| 10 | DAL | 8.5% | 8-3 | 11.9% | 6.3 | 12 | -8.9% | 32 | 0.8% | 17 | 16.4% | 19 |
| 11 | SF | 7.6% | 7-4 | 6.3% | 6.3 | 10 | -0.1% | 17 | 4.3% | 8 | 11.7% | 8 |
| 12 | IND | 6.9% | 7-4 | 11.3% | 6.4 | 9 | -0.2% | 19 | -8.9% | 27 | 13.3% | 9 |
| 13 | PIT | 6.8% | 7-4 | 8.8% | 5.9 | 16 | -8.0% | 29 | 3.1% | 10 | 16.1% | 18 |
| 14 | DET | 4.6% | 7-4 | 6.4% | 6.1 | 13 | 0.5% | 16 | -8.0% | 26 | 8.6% | 5 |
| 15 | CIN | 3.5% | 7-3-1 | 6.3% | 6.3 | 11 | -1.2% | 22 | 2.8% | 11 | 23.7% | 31 |
| 16 | SD | 1.3% | 7-4 | 5.4% | 6.1 | 14 | -0.2% | 20 | 21.1% | 1 | 17.6% | 20 |
| TEAM | TOTAL DVOA |
W-L | NON-ADJ TOT VOA |
ESTIM. WINS |
RANK | PAST SCHED |
RANK | FUTURE SCHED |
RANK | VAR. | RANK | |
| 17 | ARI | 1.2% | 9-2 | 2.0% | 5.9 | 15 | 2.0% | 11 | 3.7% | 9 | 5.1% | 2 |
| 18 | NO | 0.9% | 4-7 | 2.0% | 5.2 | 19 | -0.3% | 21 | -12.1% | 30 | 17.8% | 21 |
| 19 | CLE | -3.7% | 7-4 | 5.5% | 5.4 | 18 | -7.0% | 28 | 4.5% | 7 | 15.3% | 14 |
| 20 | ATL | -6.5% | 4-7 | 1.5% | 4.8 | 21 | -8.5% | 31 | 2.8% | 14 | 13.9% | 11 |
| 21 | HOU | -9.7% | 5-6 | -3.1% | 3.9 | 25 | -2.7% | 23 | -10.6% | 28 | 6.7% | 3 |
| 22 | CHI | -9.8% | 5-6 | -13.2% | 5.0 | 20 | 1.8% | 12 | 1.0% | 16 | 11.7% | 7 |
| 23 | MIN | -13.8% | 4-7 | -15.0% | 4.6 | 22 | 0.8% | 15 | -5.0% | 23 | 15.0% | 13 |
| 24 | STL | -15.1% | 4-7 | -18.2% | 3.6 | 28 | 5.6% | 4 | -6.4% | 24 | 20.7% | 27 |
| 25 | NYG | -15.8% | 3-8 | -23.0% | 3.6 | 27 | 3.1% | 8 | -14.6% | 32 | 18.8% | 23 |
| 26 | OAK | -17.9% | 1-10 | -25.9% | 3.1 | 30 | 7.7% | 2 | 10.3% | 3 | 4.8% | 1 |
| 27 | WAS | -18.0% | 3-8 | -14.5% | 3.7 | 26 | -6.8% | 27 | -0.6% | 20 | 21.8% | 28 |
| 28 | NYJ | -19.3% | 2-9 | -25.4% | 4.1 | 23 | 9.4% | 1 | 5.6% | 5 | 20.0% | 26 |
| 29 | CAR | -19.7% | 3-7-1 | -24.4% | 4.1 | 24 | 4.2% | 5 | -10.9% | 29 | 14.0% | 12 |
| 30 | TEN | -21.8% | 2-9 | -19.8% | 3.2 | 29 | 1.2% | 13 | -13.7% | 31 | 18.2% | 22 |
| 31 | JAC | -30.4% | 1-10 | -29.8% | 1.9 | 32 | 2.0% | 10 | -6.9% | 25 | 22.3% | 29 |
| 32 | TB | -31.3% | 2-9 | -23.2% | 2.8 | 31 | -5.7% | 26 | 2.8% | 13 | 26.7% | 32 |
Comments
181 comments, Last at 04 Dec 2014, 6:22am
#4 by shoutingloudly // Nov 25, 2014 - 5:49pm
Holy cow, folks. There is a very real chance that (a) New Orleans will win their division with a losing record, (b) Seattle will get the #5 seed, and (c) we'll therefore have a reverse repeat of the first round in 2011.
Anybody on FO staff want to sketch up an estimate of the odds that Seattle ends its title defense in exactly this way? The oversimplified multiplication method (57% that NO is #4, multiplied by the 16% that Seattle is #5) tell us that NO hosting Seattle is roughly 9%, and the Saints would have at least a decent shot to win that game. At least some of those 9% of simulations, though, will come when NO wins at least 4 more games...
Is it perverse to actively root for this specific outcome? Just desserts and all...
#140 by intel_chris // Nov 26, 2014 - 6:45pm
Aside from the division winner part, didn't Oakland do that a couple years back? I remember a year where the stats for the AFC West were just bizarre. One team won all its in division games and nothing else. Two other teams, SD and DEN, (or perhaps it was all three) had reasonable but not outstanding results other than their losses to OAK. I suspect the division winner ended up 10-6, Oakland at 6-10, with the rest of the division at 9-7.
#5 by Perfundle // Nov 25, 2014 - 5:53pm
Hauschka wasn't 4-4 on field goals, although I suppose he goes down in FO's stats as such. He had one blocked, and it was at least partly his fault for kicking it so low. Also, the initial trajectory looked like it might've missed even if it hadn't been blocked.
#8 by Steve in WI // Nov 25, 2014 - 6:24pm
Ah, the Bears are officially in FO projection hell - they no longer appear on either the playoff scenarios chart or the draft projection chart. Seems like a pick in the 11-15 range is coming. Oh well, Emery wouldn't know what to do with a great pick anyway.
On a more fun note, I believe I read something over the weekend that indicated it is theoretically possible for the winner of the NFC South to have a 5-11 record. God, I hope that happens. What a disaster. (Alternatively, I would love for Atlanta to win the division by finishing 6-0 in the division and 0-10 outside of the division. That's probably the greatest argument I can think of for not guaranteeing a playoff berth to every division winner).
#14 by young curmudgeon // Nov 25, 2014 - 6:44pm
If you're not going to make the playoffs if you win your division, what purpose do divisions serve? I supposed you could say "scheduling," but recently has anyone really look forward to those two annual NE-Buffalo contests or (until reality set in this year for the Steelers) the two automatic wins that Pittsburgh got from playing Cleveland each year for the past decade? (And yes, I do remember some interesting NE-Buffalo and PGH-Cleveland games, but are there enough to warrant two a year?)
If you are a good team in a strong division and you don't make the playoffs, or if you're the best of a sorry lot and do, well, that's just how it works out. It's an arbitrary system anyway (Dallas is in the "East"?); you can devise another arbitrary system, but chances are it will eventually reveal some flaws that commenters in FO threads will lament just as heartily.
The playoffs and super bowl are not designed to determine the best team; they're designed to determine the winner of the super bowl. If a team with a losing record slithers into the playoffs this year, the reaction that is best for your mental health is not outrage, it's "Yeah, well..."
If that's too passive or quiescent for you, I'm OK with that, too!
#19 by Perfundle // Nov 25, 2014 - 6:58pm
If divisions didn't exist, New England and Pittsburgh would still have made the playoffs in most years, so I don't see that as a good argument. Creating rivalries is a major purpose of divisions, and just because some teams suffer down years doesn't make it invalid.
If you are a good team in a strong division and you don't make the playoffs, or if you're the best of a sorry lot and do, well, that's just how it works out. It's an arbitrary system anyway (Dallas is in the "East"?); you can devise another arbitrary system, but chances are it will eventually reveal some flaws that commenters in FO threads will lament just as heartily.
This is a poor argument too. Not all flaws are created equal, and if changing the system produces a significantly less flawed outcome, that's the correct thing to do. For instance, the fix that the NBA made, where division winners still make the playoffs but don't get homecourt advantage if they have a worse record than their opponent can easily be applied to the NFL as well.
#22 by Cythammer // Nov 25, 2014 - 7:02pm
I don't think there's any way the NFL is going to remove the automatic qualification for division winners. I'd hope that they would at least consider seeding the playoffs based on record so we don't have so many mediocre or even bad teams hosting playoff games.
#74 by Bright Blue Shorts // Nov 26, 2014 - 8:56am
The best solution I can come up with for redoing the playoffs is to:
- get rid of the conferences
- Have a seeded tournament based on records
- Have the 8 division winners and the four teams with the best records in it.
That would also allow the Super Bowl to potentially be between the best two teams even when they're in the 'same conference'.
It's not ideal because I always liked the AFC/NFC system but actually it's got little relevance these days, they've already got rid of it from the Pro Bowl and teams like Pittsburgh, Colts, Browns, Seahawks, Bucs have played in both conferences/leagues.
#21 by JoeyHarringtonsPiano // Nov 25, 2014 - 7:02pm
I think the NFCS this year will be a historically bad outlier. If this keeps happening every 4-5 years, then maybe it would be time to consider cutting back the number of divisions. If it only happens every 15 years, then I think I'm okay with that.
#24 by Anon Ymous // Nov 25, 2014 - 7:06pm
It doesn't add much to whether they *deserve* to be there, but I find it interesting that virtually every team that inspires this kind of discussion wins their first playoff game. Seattle in 2011, SD in 2008 and Arizona in 2008 jump immediately to mind.
It's a good precedent for the NFCS champion.
#25 by Cythammer // Nov 25, 2014 - 7:12pm
That looks like a miniscule and basically meaningless sample size to me. I know sometimes people will argue that teams like the Seahawks winning over the Saints that year legitimizes their place in the playoffs, but I don't see how that's the case. It's one game. One game doesn't mean much. The Raiders just beat the Chiefs. If anything, the fact that these bad teams keep on managing to eliminate better teams make it's all that much worse that they're allowed in.
#30 by Perfundle // Nov 25, 2014 - 7:28pm
That's mainly because they have homefield advantage. Underdogs of at least three points in the playoffs have won 50% at home but only 29.5% on the road, and that's another reason why bad division winners shouldn't get rewarded this way.
#34 by Tim F. // Nov 25, 2014 - 8:03pm
Rather than change the playoffs or divisions, why not change the scheduling? It may not primarily be a result of scheduling but when you have a weak division and a strong division and they play each other — certainly that contributes to results that make us question divisions or playoffs.
Rather than:
6 division games | 4 same conference, different division | 4 other conference, different division | 2 by ranking algorithm
Why not:
6 division games | 4 different division (7-year rotation of both conferences) | 6 games ranked by algorithm (including 4 against "non-rotation conference"; that is, each team plays at least 10 intra-conference games and 4 inter-conference games)
This may not entirely resolve the division leaders with poor total ranks but it seems like it would help. The issues would be: 1. feasibility/difficulty, 2. teams are now playing every other team every 7 years instead of 4, 3. It actually leads to odder results v. ranking? Other?
Feasibility: part of what makes the current formula "easy" is the division rotation and balance of home/away (it still has "tweaks" for coasts/ travel, London, and other factors). I think this preserves the home/away balance. I'm unsure if this change "breaks" current algorithm in other ways, but it seems possible to craft one that works (only going from 2 to 6 algo-ranked games).
Not playing all 32 teams every 4 years: seems worth giving up for a longer period and promotes "rivalries" if teams are able to maintain rank year-to-year.
Imbalance: More games based on rank rather than divisions may lead to closer records — which may lead to imbalanced division leaders if teams break away from or cement previous year's ranks and/or more seemingly "worthy" teams missing playoffs? I'm unsure on this.
I guess some would argue the value of previous year's rank, but it seems a better compromise to me than tweaking divisions and playoffs (I'm open to reseeding though).
#35 by Kamab // Nov 25, 2014 - 8:11pm
I like your idea, and think it would be easier to rank teams if there were more inter-conference games. I think it's difficult to gauge SOS at the time schedules are created, otherwise that would be a more important part of the formula as well. Sports News makes a big deal about schedules pretty early on but most of the projections of what teams will be strong/not minus a few historically good/bad franchises end up being a wash. Incidentally the guys over here at FO seem to do better than most. The schedule creating algorithm is already a mess of algorithm-generated/human-cherry-picked data, that the excitement of division/conference rivalries and its ability to be marketed outweighs whatever benefit coming up with fairer(and more evenly distributed over time) schedules would be.
I REALLY like the idea of reseeding, in particular ideas I have heard about just reseeding the bottom 3-4 teams in each conference (aka top 1-2(3) are not reseeded based on record). I have a feeling that similarly to the above issue, the NFL doesn't see any benefit in changing the format.
#47 by Sixknots // Nov 25, 2014 - 10:52pm
"ability to be marketed outweighs whatever benefit coming up with fairer schedules would be"
You hit it right on the head there. -it's a business- Marketing will always outweigh whatever we fans perceive as fair. It's what we will pay for (and watch on TV) that counts.
#45 by RickD // Nov 25, 2014 - 10:33pm
The current system is designed to make sure each team plays each other team relatively regularly. Somebody noticed that under the prior system, in spite of having long careers that overlapped for most of their durations, Dan Marino and Joe Montana only went head-to-head twice.
Under the current system, every pair of intra-conference teams meets at least once every three years, and visits every intra-conference foe at least once every six years. And inter-conference foes are seen at least once (well, exactly once in the regular season) every four years, and visited once every eight years.
I just don't think the problem being addressed here is that serious. Sure, once every five years or so you might get a divisional winner that's weaker than the third best wild card from the other three divisions. But so what? No matter what you do, you're going to have an unbalanced schedule anyway (with only 16 games and 32 teams, that's pretty much inevitable). And with an unbalanced schedule we cannot trust that records are going to match team strength neatly anyway. Sure in some extreme cases it'll seem "obvious" but where do you want to draw the line?
I say we can live with the fact that the set of six playoff teams isn't exactly the same as the six best teams. But it will almost certainly contain the best team (as well as the second-best team). And the purpose of the playoffs is to crown a champion, not to sort the entire league.
#48 by PaddyPat // Nov 25, 2014 - 10:56pm
It really seems to me that the bigger issue than the non-deserving home playoff team is the deserving team that gets left out. There is a very high likelihood, this year, for example, that an 11-5 San Francisco team misses the playoffs while a 6-10 New Orleans team hosts a game. That was essentially the case in 2008, when an 11-5 Patriots team sat home. 10-6 missing the playoffs is sort of stomachable, but 11-5? That's a very good season by an reasonable standard. Before recent ridiculousness, 11 wins always seemed to me to be the gold standard baseline for number of wins necessary to win the Super Bowl...
#63 by rfh1001 // Nov 26, 2014 - 4:02am
Yes. Add me to the list (there's a list, right?) of those who think that trying to calculate a 'fair' way to allocate 16 games to 32 teams whose quality varies wildly year on year is impossible, and the present system is not broken. Yes, sometimes it throws up screwy results, but it's not a system for producing definitely the best team - it's a system for producing a Superbowl champion, and sometimes that champion will be the dominant Seahawks, and about every second year it will be the Giants lucking out in some absurd way.
(I'm not saying: 'It's popular, don't fix it,' a la CFB traditionalists. It's different from CFB in that it's not arbitrary. Everyone knows the rules when they start, and the rules are clear. It's just that sometimes the best teams don't win, like they don't win football matches. That's what FO is all about, in a way. Plucking an example from the air, for instance, New Orleans current win-loss record isn't necessarily an accurate guide to their quality. They're not a top-12 team, but they're better than their record.)
#73 by Travis // Nov 26, 2014 - 8:37am
Somebody noticed that under the prior system, in spite of having long careers that overlapped for most of their durations, Dan Marino and Joe Montana only went head-to-head twice.
Only because of bad injury timing - Montana was out for the 1986, 1992, and 1994 matchups, and Marino was out for 1993's. (Montana and Marino also faced each other twice in the playoffs, but the scheduling formula has nothing to do with that.)
#101 by oaktoon // Nov 26, 2014 - 11:32am
Yep-- happening again now. Rodgers v Brady 1st time as starters (Rodgers was a backup in 2006, he was concussed and didn't play in 2010) this Sunday;
Rodgers v Manning 2nd time as starters next year when GB plays Denver
Part of it is the age difference, as was true for Montana and Marino-- but bigger is the rotation between conferences.
Brady v Manning, OTOH-- 16 times... Bit of overkill, I'd say
#56 by Pat // Nov 26, 2014 - 12:25am
Well, you could understand a rule saying that division leaders get an automatic berth into the playoffs only if their records are 7-9 or better, or something similar.
It's not too nuts to put a minimum threshold to say 'yes, winning your division gets you in, but only so long as you don't suck horribly.' I mean, at 6-10 you could get all of those 6 wins from division play. Requiring at least 7 wins doesn't seem like a crazy threshold, and a 7-9 team in the playoffs doesn't bother me that much.
Honestly I would've said before this year that that rule would be 'on paper only', because no division winner could ever be that bad. Shows what I know!
#23 by JoeyHarringtonsPiano // Nov 25, 2014 - 7:05pm
Although I think a 5 win team hosting a playoff game is wrong on many ethical, competitive, and asthetic levels, for some reason I wouldn't mind seeing it happen this year. Sometimes I just want to watch the world burn.
#43 by shoutingloudly // Nov 25, 2014 - 9:36pm
Wow, it's entirely possible that we'll see the 5-10-1 Panthers starting The Ghost of Cam Newton in a home playoff game. Gads. I can't decide if this has me giddy or terrified.
http://espn.go.com/nfl/playoffs/machine/_/factor/powerranking/results/400554371~2~400554388~2~400554390~2~400554406~2~400554404~2
#49 by Wikitorix // Nov 25, 2014 - 11:02pm
Ask and you shall recieve: http://tinyurl.com/qh6mb4g
A playoff machine scenario where the Saints win the division with 4 wins. As a bonus, this scenario has the Seahawks and 49ers both sitting at 11-5 and failing to make the playoffs.
#10 by Bright Blue Shorts // Nov 25, 2014 - 6:33pm
This is one weird season.
Pretty much every one of the big contenders has blown out another team but also been blown out themselves or just lost/looked terrible against one of the poor teams.
New England and Green Bay look the most consistent to me.
#12 by JoeyHarringtonsPiano // Nov 25, 2014 - 6:40pm
You do have to consider the fact that Green Bay was blown out by a below average New Orleans team less than a month ago (If Rodgers' hamstring wasn't tweaked, the game would have been competitive, but a Packers win still wouldn't have been a guarantee.
#16 by Perfundle // Nov 25, 2014 - 6:48pm
Pretty much every one of the big contenders has blown out another team but also been blown out themselves or just lost/looked terrible against one of the poor teams.
That's the case every season. Last year Seattle struggled mightily with Houston and Tampa Bay, San Francisco lost two games by at least 20 points, New England needed a miracle to eke out a victory at home against Cleveland and Carolina got blown out by New Orleans and Arizona.
#17 by Cythammer // Nov 25, 2014 - 6:50pm
DVOA's variance disagrees. New England has been consistently excellent in recent times, but not for the entire season. Green Bay hasn't even been consistent recently. In the last four weeks they've had two incredible blowout win along with an ugly loss to a mediocre team and narrow win over a bad team. That's pretty up and down.
#13 by oaktoon // Nov 25, 2014 - 6:43pm
This comment is not a shot at DVOA, but, as Aaron and others have observed, reflects the increasingly random nature of the postseason in the NFL for the past 6-8 years.
Look at Teams 10-17 and I wonder where the surprise SB winner might lie:
Dallas-- could, with a win this Thursday, not only take a huge step toward winning the East but also has a realistic shot at a #1 or #2 seed;
San Francisco-- could, with a win this Thursday, not only take a huge step toward a WC (and with easier schedule than SEA) but also could still catch the Cardinals and win the West. And they have reached the NFC championship or SB three years running;
Indianapolis-- headed for a home game with a QB that has produced chills and spills the past two years in the postseason. And then presumably off to Denver or NE and who knows?
Pittsburgh-- in the AFC North scrum. has been alternately very good or pretty bad the past month or so. QB has a postseason pedigree, though;
Detroit-- defense is still first-rate. Offense really struggling, but schedule still sets up well for a WC or even a division title with a win at Lambeau in the final. (They haven't won in Lambeau since pre-Favre, though...)
Cincinnati-- they still lead the AFC North. Another schizo group this year and, of course, a miserable recent playoff history;
San Diego-- a lot of smart folks had them pegged as serious SB contenders early in season-- have looked flat lately. After all the disappointments of the Rivers era, could one magical run still be in store?
Arizona-- still has best record in NFC, though the schedule is pretty brutal.
Other than the Lions, I wouldn't be shocked to see any of these teams in Glendale... GB, Sea and Philly may be the top NFC teams but none are unbeatable-- and none may get home field. NE and Denver always seem to be the class in the AFC-- but if Baltimore broke through two years ago, who's to say they, or someone, couldn't do it again? And frankly, other than the Ravens the AFC teams from 10-17 seem scarier to me than those above them-- MIA, BUFF and KC.
#15 by JoeyHarringtonsPiano // Nov 25, 2014 - 6:47pm
What's so frustrating about Detroit this year, is that the offense has to improve to merely competent (not even good) for them to be a scary team in the playoffs. As it stands now, the offense can't get out of its own way, and the only way I see them winning a playoff game would be against the NFC South winner.
#31 by Otis Taylor89 // Nov 25, 2014 - 7:29pm
Personally, as strange as this may sound, but I would rather be DET than the other two strong defensive teams (SEA,SF) with 50-55% chance of making the playoffs. DET have had so many key injuries skill position players the 1st half of the season that you have to figure that they will improve once everyone is back for more than a game. SEA and SF? I think their offenses are about as good as it will get this season, which isn't good enough to go very far.
#32 by Perfundle // Nov 25, 2014 - 7:41pm
Uh, Seattle's offensive DVOA this year is better than their offensive DVOA last year, and the same is true with weighted DVOA. Did you forget how bad they looked on offense for the last few games in the regular season? For that matter, their offensive DVOA is better than Baltimore's in 2012, NY's in 2011 and Green Bay's in 2010. History has shown that all a team needs is a hot streak in the playoffs, and both teams are fully capable of doing that.
Also, Seattle and San Francisco are getting a lot of key players back from injury/suspension too.
#33 by Otis Taylor89 // Nov 25, 2014 - 8:01pm
SEA offensive DVOA is better vs. last year solely based on their line play compared to the injuries of last year. Tell me exactly which skill position player is coming back for SEA? Uh, that would nobody. Golden Tate is not walking through that door. Percy Harvin is not walking through that door. Matter of fact there is a better chance Lynch will never be 100% again. And what SF skill position player is coming back? They are getting players back on D, but I see both offenses staying the same or getting worse - especially after they beat the crap out of each other 2 out of 3 games.
#36 by Perfundle // Nov 25, 2014 - 8:33pm
SEA offensive DVOA is better vs. last year solely based on their line play compared to the injuries of last year.
Because Seattle's line has been so healthy this year?
Percy Harvin is not walking through that door.
Good thing Seattle got that inferior offensive DVOA last year almost completely without him.
And you didn't seem to get the point that Seattle has a better offensive DVOA than last year despite the lack of those offensive skill players. If Seattle's offense last year was enough to get the job done, and Baltimore's and New York's were as well, then their offense this year is definitely capable of winning a Super Bowl, as are the offenses of about half the league.
I'm also unsure what you see in Detroit that makes you think they'll be that much better even with their skill players back. They didn't exactly light it up against the bad defenses of Green Bay and Carolina with their players healthy, did they? When has trusting a Stafford-led team to improve on offense paid off, exactly?
#40 by Otis Taylor89 // Nov 25, 2014 - 9:16pm
So you are saying that their best OLineman, Russell Okung being out over half the year last year had no affect on their DOVA?
Seriously, SEA won last year with historically great defense (which is also not walking through that door) and an offensive that was able to get the job done when Okung came back. They could win on the road last year, but didn't need to. This year? They are going to have to win on the road multiple times.
You know what the big difference in the offense this year vs. last year? Russell Wilson had as many carries already this year as he did all of last year and he's much more productive running it this year. Good luck with that on the road in SF and ARZ and hone against SF.
#58 by Perfundle // Nov 26, 2014 - 12:53am
So you are saying that their best OLineman, Russell Okung being out over half the year last year had no affect on their DOVA?
If you actually watched Seattle last year, you'll know that Okung wasn't that good even when healthy, and also that Seattle had some great offensive games even when he was out (Jacksonville, Indy, at Arizona, Tampa Bay, Atlanta), so basically, I don't think he had that big an effect, no.
They could win on the road last year, but didn't need to. This year? They are going to have to win on the road multiple times.
Well what do you know, that's a perfect description of the 2010 Packers, who went 3-5 on the road, barely squeaked into the playoffs and rattled off three straight road victories. You seem to be quite ignorant of NFL trends and the concept of Any Given Sunday. Oh, and Detroit, with their 3-3 road record, including two blowout losses, has a better chance of winning three road games?
Good luck with that on the road in SF and ARZ and hone against SF.
So let me get this straight. Things that Seattle aren't doing well clearly have no chance of improving, things that Seattle does do well are obviously going to decline, and things that Detroit is horrible at can only be on the upswing? How can anyone argue with that? Seattle's offense doesn't need to play well to beat Arizona and San Francisco; they just need to play better against their defenses than their mediocre offenses against Seattle's defense.
#68 by Otis Taylor89 // Nov 26, 2014 - 6:59am
Nice job cherry picking my comments.
I can't believe you brought up the 2010 Packers as a possible comparable. You do realize that team had Greg Jennings at the top of his game, Jory Nelson and James Jones in his prime? Hell, they even had a 35 year old Donald Driver, who I may take over any receiver SEA currently has. Seriously, SEA has Lynch and Wilson as above average talent (and Lynch is 28 with almost 2,000 regular season carries), who else on that offense do you expect to do anything on the road?
#102 by oaktoon // Nov 26, 2014 - 11:36am
Yes but James Starks as the RB? I think the truth, on this one, is much more like baseball has become.... Meaning that you can pretty much throw out the regular season when the playoffs begin.
2007-- Giants
2008-- Cardinals (yeah, they lost the SB. Barely)
2010-- Packers
2011-- Giants
2012-- Ravens
None of those teams were anywhere near the "best" in the league going into the postseason.
#134 by Perfundle // Nov 26, 2014 - 4:46pm
Nice job cherry picking my comments.
Nice job ignoring most of my responses to your amazingly flawed comments. What comments of yours did I cherry pick in that post? You made four points - Okung was hurt, Seattle's defense was historically good, Seattle has to win on the road this year and Seattle's offense is dependent on Wilson's running which won't be there in their other divisional matchups - and I responded to three of them. As for Seattle's defense, I'm pretty sure San Francisco didn't have a historically good defense, and their offense was about the same as Seattle's last year, and they were one catch away from doing what you claim Seattle can't do, which is win three road games in the playoffs. But let's get back to your laughable evasion. So far, you've failed to reply to:
Detroit's unimpressive performance even when all their players are healthy
Detroit's own unspectacular road record, including two blowout losses to Seattle's zero
Detroit's history of underperforming their offensive talent under Stafford
Is every single response of yours going to blatantly ignore your initial point, which is that Detroit has a better chance of going further in the playoffs than Seattle? Instead you take every chance you can of talking about how bad Seattle's offensive skill players are, despite the fact that they're 8th in DVOA with them? DVOA is not VOA, so your ramblings about how they'd do against Arizona and San Francisco, presumably harder competition, are nonsensical. Assuming no change in performance, they would still average their normal offensive DVOA. Oops, I just mentioned something relating only to Seattle, so I guess you're going to respond to that and nothing about Detroit again.
You do realize that team had Greg Jennings at the top of his game, Jory Nelson and James Jones in his prime?
You do realize that despite all that they still had a lower offensive DVOA than Seattle this year? Yes, Rodgers was out for 1.5 games, but Flynn did very well in the second game, so it didn't make much of a difference. It's like, maybe there are multiple ways of being good on offense.
#136 by Otis Taylor89 // Nov 26, 2014 - 4:49pm
Didn't realize I was exchanging responses with Doug Baldwin.
I've answered all your responses, you haven't answered what skilled position player on SEA is capable of stepping up on the road, in the playoffs that is not named "Wilson" or Lynch" - probably because you can't name them. None of those position players have been hurt, but they all fall in the bottom half of DVOA. Yeah, their QB is having a great year running the ball and Lynch is #2 in DVOA, but I'm not sure that's going to work when you are playing SF twice. Their next opponents defensive DVOAs are #2,#8,#2,#5, with 3 on the road. How many games have they played against top 10 defensive DVOA teams? Two, ARZ and DEN, both at home.
The Lions offense has been pretty bad this season, however Megatron was out for weeks, Bush has been out for weeks, the TE's have been out for weeks (although they have been a big disappointment. Yeah, Johnson has probably peaked, but he still requires doubling, Bush isn't what once was, but he is still a threat coming out of the backfield. Stafford finished ahead of Andrew Luck last year in DVOA and now has Tate. They have 3 straight home games vs. mediocre to bad teams (those teams defensive DVOA's are #23,#21,#20,#23). That's a lot easier that going against the #5, #6 and NE's defense (which will be in the top 10 soon), with 2 of those games on the road. I say they turn it around.
That is why I like DET chances better than SEA. SF is a different story...
#147 by Perfundle // Nov 27, 2014 - 12:03am
I've answered all your responses
No, you clearly hadn't, considering all of your comments on Detroit are the first you've made since your first post.
You haven't answered what skilled position player on SEA is capable of stepping up on the road, in the playoffs that is not named "Wilson" or Lynch" - probably because you can't name them
With the way Wilson spreads the ball around, you don't really have one or two skill-position players that you can rely on to step up, and the result is different receivers leading in yardage from game to game. It's not a desirable situation, but they've worked around it. Also, any player could step up for multiple games, because, again, the NFL is rather random. Did you see Starks being the difference maker when Green Bay made their run in the playoffs?
Yeah, their QB is having a great year running the ball and Lynch is #2 in DVOA, but I'm not sure that's going to work when you are playing SF twice.
Lynch has gotten 103, 111, 98, 72 and 109 yards against San Francisco in his five games since Wilson's got here. Why can't it work against San Francisco, especially with rush defense not being the Niners' strong suit?
Their next opponents defensive DVOAs are #2,#8,#2,#5, with 3 on the road. How many games have they played against top 10 defensive DVOA teams? Two, ARZ and DEN, both at home.
Their next opponents' offensive DVOAs are #18, #16, #18 and #23, and Seattle's defense got a big boost from Chancellor and Wagner coming back, even though Mebane is a big loss. Again, Seattle doesn't have to play well on offense, only to play better than the opposing offense.
The Lions offense has been pretty bad this season, however Megatron was out for weeks, Bush has been out for weeks, the TE's have been out for weeks
After three weeks, during which Johnson, Bush and the TEs were all healthy as far as I know, Detroit had a 17th-ranked offensive DVOA, which is why I don't see much improvement with Stafford leading them.
That is why I like DET chances better than SEA.
That by itself is a perfectly reasonable view to have, because Detroit's defense has been excellent, and because they have a considerably easier schedule, as you pointed out. What I found objectionable was your argument that you liked Detroit more because Seattle's offense isn't good enough to get it done. It might be lacking good wide receivers, but it makes up for it with option plays and Wilson's scrambling, and these are just as hard to account for as having good receivers, even for great defenses.
#89 by JoeyHarringtonsPiano // Nov 26, 2014 - 10:15am
Just let me say for the record that the ONLY advantage DET has over SEA is comparatively cupcake schedule (26th by DVOA vs. 12th). The Seahawks offense is playing way better, and their defense isn't that much worse. SF looks a lot like DET to me, except having a marginally more competent offense and better special teams.
#20 by Cythammer // Nov 25, 2014 - 7:00pm
I think just about any team that could get in could win. A one-and-done playoff format will greatly increase and chances of the tournament having a wacky final result. Just think about some of the crazy upsets we've had in the regular season: Titans blowing out the Chiefs, Jets beating the Steelers, Bucs beating the Steelers, Redskins beating the Cowboys, Rams over Broncos, Rams over Seahawks. In the regular season we can view games like these in the larger context of the teams' seasons and rightly dismiss them as outliers. But in the playoffs a few odd results of lesser magnitude than any of these can see a mediocre team winning it all.
#29 by Bobman // Nov 25, 2014 - 7:28pm
That's an interesting and logical way to look at it. The top two seeds in each conf are the favorites for a reason, but history says things rarely pan out that way.
This Colts fan finds himself longing for a playoff rematch with Pittsburgh--sort of a karma-fixing game that reverses 2005 (Colts win pretty big in reg season, lose in playoffs), and the five playoff losses the Colts have vs the Steelers. (Kordell Stewart runs out of EZ and is not called! Kill me! 19 years of grudge-holding.)
One of my sons asked me about the Colts' chances for the SB and I explained with something like this: "Well, their offensive line isn't very good, and their running game is pretty weak, and their D is not what it needs to be...." So can Luck and TY Hilton somehow steal 2-3 playoff games? Andy Reid says yes. Maybe with some D help and the D has been up and down. Of course about this point in the 2006 season (after the Jax game) I (and everyone else) had written Indy off--then they win it all. And who really believed in the Ravens two years ago? Next time I'll tell my son, "Well, son, sometimes sh!t happens."
#37 by Anon Ymous // Nov 25, 2014 - 8:40pm
From 375 to carrying a QB with 3 TDs and 7 picks. Whoda thunk it? ;-)
I still maintain that the Colts lucked out with KC and Baltimore's coaching staffs being too dense to recognize Indy's complete change in DL roles in the playoffs. I noticed it within a few drives against KC... was even imploring them to run a few play action passes on first down which they never did. Not once.
But hey, you take a championship any way they come, right?
#66 by Jerry // Nov 26, 2014 - 6:23am
Let's see. In their 1976 playoff game, the Colts reduced the Steelers to one healthy running back for the next week's AFC Championship. And I don't know that the Colts will ever be able to repay the karmic debt for the Steelers letting Johnny Unitas go.
#41 by Rick_and_Roll // Nov 25, 2014 - 9:16pm
Pittsburgh is the team on your list that could surprise. They have the running game, the WRs, QB and a savvy DC. Except for Baltimore, Pittsburgh probably has the best chance to beat the Pats in Foxboro of all of the AFC teams.
As a Bronco fan, Baltimore and Pittsburgh are the two teams I do not want Denver to face in the first round.
#50 by CaffeineMan // Nov 25, 2014 - 11:08pm
Of all the teams the Pats are most likely to face in the first round, I'd way rather see Pittsburgh than any of the other teams. Although if the Chiefs have to come to Foxboro, I don't think they'll be able to duplicate the result from earlier in the year. Funny how match ups work...
#51 by mehllageman56 // Nov 25, 2014 - 11:13pm
I would argue KC has a better chance, because they've already shown they can do it. The key to attacking the Pats is pound them on the ground, and then throw to the backs and tight ends. The Chiefs don't care about throwing to their receivers, so that fits into their plans. As far as dealing with them defensively, interior pressure works best. The Chiefs have Dontari Poe, the Steelers have...
#97 by Anon Ymous // Nov 26, 2014 - 11:16am
This got me in trouble last week, but it is most likely that NE's prior match up with KC has limited predictive ability. Sure, TEs have given NE trouble and a team that commits to the run (and has a defense to allow it to stay that way) is better equipped than the alternative, but a rematch between the teams is unlikely to unfold in a similar manner.
#55 by nat // Nov 26, 2014 - 12:09am
I'm generally a fan of DVOA. But it's very odd that Denver's rating seems to depend on the proposition that its three point nail-biter comeback over Miami was as dominating as the Patriots' twenty-two point shellacking of the Broncos.
DVOA sometimes highlights things we would otherwise miss. But this particular weirdness calls out for some 'splaining. Is it just DVOA's habit of grading on an easier curve for teams playing from behind? Or is there some great insight in DVOA, and if so, pray tell us.
#59 by Ketawa // Nov 26, 2014 - 1:23am
Honestly, I don't know all the ins and outs of DVOA. But my biggest guess that I recall from watching the game would be really bad fumble luck for Denver. Miami recovered all 4 fumbles in the game, 3 by Denver and 1 by Miami. Not sure if it's counted, but Chris Harris Jr dropped an easy pick-6. Denver seemed more successful at steady gains in the running game.
#70 by nat // Nov 26, 2014 - 7:25am
That's the one thing I can see. DVOA should value those fumbles as if Denver recovered two of them. But it doesn't do like a fan might and then imagine a whole series of what ifs. It's simply given the value of a certain amount of field position.
If I recall, the blowout loss to the Patriots featured no such fumble luck.
I can't see how two fumbles can translate into nineteen points of field position value. But it is a partial explanation.
Unadjusted VOA rates fumbles as actually recovered. Does anyone know Denver's unadjusted VOA for these two games? Better yet, for each half or quarter? It would help us figure out what VOA thinks it is seeing.
#79 by nat // Nov 26, 2014 - 9:34am
The second thing I looked for was returns on interceptions and fumble recoveries. DVOA (and VOA too, I think) treat those as completely luck, assigning the value of an average return to each.
In the Miami game, Denver benefited from a 37 yard interception return. In the Patriots game, the Patriots had 41 yards of returns on two interceptions. Average returns are not zero yards. So, as much as "return luck" favored the Patriots a few weeks ago, Denver had as much or more this past week. It certainly had a ton more "return luck" per return.
So that's not the explanation.
#85 by nat // Nov 26, 2014 - 10:01am
Another thing to look at is fourth down failures.
DVOA treats these as it treats third down failures. That's fair enough as far as DVOA's predictive value. But it means that DVOA (or VOA) misses out on the extra damage such failures do in terms of lost field position or field goal attempts left unattempted.
At the Patriots, the Broncos went 0-4 on fourth down. That would go a fair way towards explaining the blowout game not being given a blowout VOA. Foregoing a punt (by turning it over on downs) costs a couple of points worth of field position. Foregoing a chip-shot field goal costs three points, but maybe gets one back in field position by pinning your opponent inside their own ten.
So maybe VOA should be seeing a 22-(4x2) = 14 point solid win for New England rather than the 22 point blowout.
#86 by dmstorm22 // Nov 26, 2014 - 10:04am
I was going to bring that up as a reason why DVOA may not love the Patriots win over Denver as much as it should.
0/4 on third down all came at the end of to-that-point good drives. On a play-by-play basis, the Broncos played reasonably well, but failed at crucial times.
I have a harder time explaining why DVOA likes Denver's win over the Dolphins so much.
I will say that for the season-long DVOA, Denver leads the league in both yards-per-play (6.2) and yards-per-play-allowed (4.8). For a system that generally works on a per-play basis, that would seem to indicate they should have a really high DVOA, especially when they don't turn it over a lot, and have played a tough schedule.
The better question is probably why, despite their great DVOA and great yards-per-play differential, is their scoring differential far lower than we would expect.
#96 by nat // Nov 26, 2014 - 11:12am
I just used this week's numbers to redo a VOA/Pythagorean Expectation/Winning Percentage analysis I had done on the topic of Arizona's strange success and/or low DVOA. It was a longish table I won't repost unless people really insist. But the highlights are...
Denver still leads the league in under-producing for its VOA. Their expected winning percentage from their VOA is 77.4%, while their expected winning percentage from their points for and against is 64.1%. That's almost a win and a half's worth of under-production. They make up some of that by winning almost a game extra beyond what you would expect from their points scored and allowed.
Arizona still leads the league in winning more than their points would indicate. No surprise there. They also score better than their VOA would suggest, enough to lead the league there, too. The total effect is a huge outlier.
Oakland and New Orleans bring up the rear in winning less than their points suggest.
The New York Jets are Denver's only serious competition for under-producing their VOA, just a little behind (ahead?).
dmstorm22:
As your comment says, the really interesting question is WHY? What is it that Denver does, that VOA loves (or fails to hate), but that the game of football doesn't reward (or does punish) with points? Does this thing say something about the Broncos? Does it say something about DVOA?
#99 by dmstorm22 // Nov 26, 2014 - 11:21am
I think it says something about the Broncos are doing that is not translating to points, particularly on defense.
You have a team that has gained more yards per play and allowed fewer yards per play than any other. They don't have a far negative turnover differential. They play well in a lot of areas.
A couple possibilities: field position is really bad - seen in their poor ST DVOA. Also their worse-than-expected performance on 3rd/4th down on both sides of the ball (0/6 on 4th down in opponent territory in their two recent losses).
The Broncos play-by-play performance on the field is just not translating into a large edge in points for and allowed.
#103 by bmay // Nov 26, 2014 - 11:37am
"Also their worse-than-expected performance on 3rd/4th down on both sides of the ball (0/6 on 4th down in opponent territory in their two recent losses)."
It's probably something along these lines. I've learned from the world of baseball statistics that it's often just random variation or "bad luck" in high-leverage situations.
For instance, Denver's defense is top 5 in preventing 3rd down conversions overall, but maybe they've allowed a higher-than-expected portion of those conversions in their own territory or when the game is on the line. These bad "clutch" performances hurt the team's chances of winning in that moment, but they are likely not predictive.
#105 by nat // Nov 26, 2014 - 11:42am
Regarding Field position via special teams: If true, this could mean that VOA is undervaluing special teams play, especially punt coverage and returns.
Another possibility is that they are cramming their positive VOA into a relatively small number of drives. That would point to exploiting personnel mismatches that come up during games, perhaps after injuries. It could also be explained by the Mile High Effect, where opponents stop looking good after three quarters of thin air. VOA splits by quarter home and away could shed light on this.
A third possibility is that Denver's VOA is benefiting unusually from low baseline situations. This would lead to higher VOA without higher actual success. Playing from behind can have this effect, although I don't know how big it is. This wouldn't explain the yards/play stats, though.
#60 by Vandal // Nov 26, 2014 - 1:40am
Don't waste your breath.
I called out the metrics 2 weeks ago, when Denver was 22% higher than New England, despite having identical records, and identical Net points +-.
Now Denver is still 10% higher, despite having HALF the net points. New England is literally winning by a touchdown more per game that Denver, and the rationalizations continue.
The unfortunate part, is that all the other advanced stats have lapped DVOA now. SRS on pro-football-reference for example, is SIGNIFICANTLY more relevant and performs much better.
DVOA suffers from two major problems. It's linear, without diminishing returns. Which means 1 terrible game can cancel 4 very good games. Which leads into the second major issue: DVOA is a "per play" metric. Which sounds advanced, but is actually inferior to win% metrics. The correct play is always the play which maximizes your chance of winning, at that point in time. Ahead by 14 with 6 minutes left, and trailing by 21 in the 3rd quarter have vastly different optimal strategies. These concepts are being covered by some of the newcomers, and DVOA is still stuck in "first downs are god" mode, that don't account for game situation. 538's series on Quarterback win% vs Expected win% is a great example of this.
Anyway, there's nothing else to say about this, until a real Mea Culpa comes down from above, and they have an honest come-to-jesus about methodology and improvements. You can have a snarky "disclaimer" at the start of every DVOA thread, but now it just reads as desperation and ignorance. Hell, they are tied with Keyshawn & Boomer, at 62% right now. Which is good for last place among stat sites, and about 80th of 130 among talking heads... Oh, and they're 43% against the spread.
Sources:
pro-football-reference.com
nflpickwatch.com
fivethirtyeight.com
P.S. Perfundle & TomG from Week 11? I'll have a coffee and an Onion Bagel, anyone else want anything while they're out?
#61 by Joshua Northey // Nov 26, 2014 - 2:14am
This is half fair and half unfair.
"2 weeks ago, when Denver was 22% higher than New England, despite having identical records, and identical Net points +-."
Means almost nothing. That is well within the margin of any system that is not just referencing records and/or net points.
"DVOA suffers from two major problems. It's linear, without diminishing returns. "
I think they would say that they have tested this and the system performs better this way. I know you think it is a problem, that does not make it one.
"Which leads into the second major issue: DVOA is a "per play" metric. Which sounds advanced, but is actually inferior to win% metrics."
Well this does cause a lot of issues, but it also allows for a lot of additional granularity to the data and analysis. Certainly FO is not a bastion of ideal methods.
The real thing that is a bugbear is that it has got one foot in description and one in prediction. It is not set up to be solely a predictive tool and probably shouldn't be evaluated as such.
And don't even get me started on 538, they are extremely good at some things, but often are missing the forest for the trees. All the analysis in the world doesn't help if you are analyzing the wrong thing.
#62 by Malene_copenhagen // Nov 26, 2014 - 3:22am
538 seems to suffer from editorial problems. Chase Stuart's stuff on Football Perspective is always interesting, intelligent and on point, while on 538 he's sometimes just dealing with straight up irrelevant ideas and concepts.
#104 by MJK // Nov 26, 2014 - 11:37am
Well, one big issue is DVOA's inherent bias towards good offenses over good defenses that comes from the fact that it is a per-play metric averaged over all plays a team has actually run.
Consider a team with an offense that plays well one week and poorly the next week. An offense that plays well will get something like 60-70 total snaps in a game, while one that plays poorly will only get 30-40 or so. So, objectively, the team had 50% of its games in this two game sample where its offense played well, and 50% where it played poorly (grading out to average), but in DVOA's mind, it had twice as many "good" offensive snaps as "bad" ones and it's going to grade out to above average over the two game sample.
Defense is just the opposite. Take a team with lights-out defense forcing 3-and-out after 3-and-out one week, and then has a key injury and gives up long drive after long drive the following week. Again, one game good, one game bad, average performance, right? No, because the team has 2-3 times as many bad defensive snaps as good defensive snaps.
In other words, DVOA tends to discount good defensive performance and more heavily weight good offensive performances, at least on a per-game basis.
#77 by bravehoptoad // Nov 26, 2014 - 9:21am
Which leads into the second major issue: DVOA is a "per play" metric.
Isn't that one of its advantages? So that plays can be broken out by pretty much any commonality?
The correct play is always the play which maximizes your chance of winning, at that point in time.
This might be true if football players had no memories, but happily for all of us they do, and some plays can be used to set up other plays.
#81 by Otis Taylor89 // Nov 26, 2014 - 9:39am
A lot of Denver's big advance in DVOA came from 3 games (ARZ,SF,SD) playing against teams that were going through major injury problems (ARZ was paying a 3rd string rookie QB on the road for his 1st game for much of the game). There seems to be a need to factor in games where big time players, like Aldon Smith)are missing.
#87 by dmstorm22 // Nov 26, 2014 - 10:07am
I agree that SF had major injury issues (though even with the injuries, their defense was far better against other teams at similar injury levels after that), but saying ARZ was going through injury issues is very misleading.
Logan Thomas played less than half that game, and did contribute 7 points (and the longest drive - though it all came off of one play). The Cardinals were down 24-13 when Stanton left, and then lost 41-20.
#82 by Pat // Nov 26, 2014 - 9:41am
Which leads into the second major issue: DVOA is a "per play" metric. Which sounds advanced, but is actually inferior to win% metrics. The correct play is always the play which maximizes your chance of winning, at that point in time.
"Always"? Arguments are always stronger when you put hedge words in them, you know. It's a game. Nothing is ever "always" true, except aphorisms like "to win, you have to score more points." You would've been better off saying "usually always" or "almost always the play." And your first statement of "inferior to win% metrics" is actually demonstrably wrong.
1) A team's goal is not to win the game, but to win a championship. Optimizing for winning a game can sometimes reduce chances of winning a championship. This is even more true in baseball, obviously, hence the reason why players get days off, but it's true in football as well.
2) Win% metrics are examples of extreme diminishing returns, and so are always going to have limited predictivity, since they will obviously throw out virtually all data when the game is essentially won, and that data *has* predictive value, which can be *proven*. Throwing away predictive data cannot lead to a better predictive metric.
3) DVOA actually only compares like situations to like situations, so your situational comment doesn't really make sense.
4) I can't imagine a situation where not gaining a first down is preferable to gaining a first down in terms of win%, so I don't see how your 'stuck in "first downs are god"' comment makes any sense. Playing around with the win% estimator tools confirms this.
#95 by bmay // Nov 26, 2014 - 11:03am
"4) I can't imagine a situation where not gaining a first down is preferable to gaining a first down in terms of win%, so I don't see how your 'stuck in "first downs are god"' comment makes any sense. Playing around with the win% estimator tools confirms this."
In most 1st-and-10 situations, gaining 9 yards to make it a 2nd-and-1 is actually preferable to gaining 10 yards for another 1st-and-10:
http://archive.advancedfootballanalytics.com/2008/09/2nd-down-and-1.html
http://outsidethehashes.com/?p=220
#133 by Perfundle // Nov 26, 2014 - 4:32pm
P.S. Perfundle & TomG from Week 11? I'll have a coffee and an Onion Bagel, anyone else want anything while they're out?
I have no idea what this is supposed to mean, but did you miss the part where I thoroughly discredited your worship of Silver's Elo stats and your absurd love of straight-up prediction percentages? In case you missed it, I'll repeat it here. It was easy enough to work out Silver's formula since he provided most of it, and so his straight-up prediction percentages for last season could be calculated, yielding a miserable 61.9% through week 10, despite fully incorporating your precious diminishing returns. Also, using the point differentials from last season and no other inputs to predict results from this season yielded an even better success rate than Elo did, which shows how static this season has been compared to last, which is why Elo has been giving such good results.
#91 by Rick_and_Roll // Nov 26, 2014 - 10:34am
Yes, New England beat Denver by 22 points, but in reality, the difference in the game was turnovers and special teams. My only guess is that DVOA rates (or perhaps underrates) turnovers as either random or not heavily.
The offense essentially gave New England 14 points off of Manning picks that were returned in or near the Denver red zone.
The special teams gave up 10 points. 7 on the punt return, and three as the result of a 15 yd PR INT penalty that allowed NE to get a FG without making a first down. Additionally McManus missed a FG which led to Denver not trying another makable FG due to lack of confidence.
#94 by nat // Nov 26, 2014 - 10:52am
Turnovers (especially interceptions) and special teams are covered by DVOA. So they can't explain the weirdness of equating a three-point squeaker with a 22 point trouncing. You've missed the point badly here. In reality, that is.
Other factors partially explain the weird results. See my posts above. It's interesting to explore what VOA is trying to tell us about these two games.
#106 by Rick_and_Roll // Nov 26, 2014 - 11:56am
I don't think I did at all.
The salient point I made was that in the DEN-NE game, the margin of victory was significantly enhanced by a few critical plays (turnovers, special teams) and perhaps these plays are undervalued in terms of DVOA. Turnovers led to 14 points and Special Teams led to 10 points. Denver had multiple long drives that led to no points and since DVOA measures per play, they essentially got credit in terms of DVOA for those drives while not getting points.
Another factor that could affect DVOA of the NE - DEN game is the 3rd down conversions percentages for both teams were outside of each teams' respective conversion rates, with NEs being much higher and DENs being much lower. I think DVOA considers abnormal 3rd down conversion rates, whether positive or negative, as unsustainable.
#109 by Tim F. // Nov 26, 2014 - 12:14pm
"So they can't explain the weirdness of equating a three-point squeaker with a 22 point trouncing."
If you can't understand how a game with a 3-pt differential versus a game with a 22-pt differential can have similar DVOAs, it is you who is missing the point of DVOA.
#112 by intel_chris // Nov 26, 2014 - 1:02pm
It's interesting to explore what VOA is trying to tell us about these two games.
Agreed. It is hard to disentangle this from the argument about whether DVOA is a good predictive statistic, because in some sense a descriptive statistic is "only" interesting because despite all the advice to the contrary, human nature expects past results to be predictive of future performance or whatever the mutual funds prospectuses warn us against.
And, to that extent, it is interesting the the VOA for the NE blowout of DEN was much closer than the score.
A similar point was mentioned recently in the AZ loss to SEA where AZ got only 3 points. The explanation of that game I read that was intuitively somewhat satisfying ascribed it to "luck" with the following particular definition of luck: one can have good plays and bad plays in any order, the ordering of the plays (good v. bad) can be the difference in a successful drive and an unsuccessful one. In the AZ game, the luck shifted away from many successful third down plays, to more success on first and second down, but ending up short of third down. The total number of successful plays is the same, but the sequencing means the drive stalls.
I think a similar factor might be present in DEN's two recent losses (NE and STL), where the Broncos were particularly failing on third (and worse fourth) down plays. The one "feeling" I have is calling this merely "luck" is perhaps missing something. I know that previously, FO has shown that abnormally good third down VOA is not sustainable and tends to revert to the mean. However, I'm not sure that there aren't "predictive" factors in that unsustainable metric, even if they aren't statistically significant due to small sample size. For example, I'm not at all confident that the Broncos defense (or offense) is as good when playing from behind. I think PM may throw more INTs (esp. on third down) when trying to catch up. Lack of faith in the FG kicker is also not a positive. (Rick_and_Roll did a nice dissection of problematic factors in the Broncos losses.)
Thus, I think your point, analyzing what VOA is telling us--and what it isn't telling us, is a worthy topic. And, it isn't so we can make VOA more predictive. Maybe we can. Maybe we can't. But, it is so we can understand why some team is doing better (or worse) than VOA or the W/L record is suggesting, and what matchups we should be watching. Because if we write all the discrepancies off to luck without further analysis, we are missing the chance to learn something.
#115 by Pat // Nov 26, 2014 - 1:18pm
For reference, I'm pretty sure Aaron has said that third down VOA imbalance tends to stick around for a while, but regress over a longer timescale - that is, you expect it to regress by next year, but not necessarily by next week. Unlike fumble recoveries, which are completely random.
I really think that there's a ton to investigate in terms of third down performance. It's such a high leverage situation that any small advantage there a team can eke out will have a big impact on the results, but it's also makes total sense that teams can't sustain an imbalanced advantage on third down for very long.
Lack of faith in the FG kicker is also not a positive. (Rick_and_Roll did a nice dissection of problematic factors in the Broncos losses.)
You know, that's really interesting, and I have no idea how you'd compensate for it. That's a great example of how bad special teams won't show up in a play by play (since they don't even attempt the kick), and will in fact lead to fewer points scored.
I dunno... maybe something like for a 4th down play, a failed offensive play is assigned partly as negative value to the field goal unit, based on how often other teams in similar field position and similar score differential kick a field goal (A successful offensive play still goes to the offense, of course).
As field goal kickers get better and better at long range, this is going to end up being a bigger effect. Already, for instance, "short range" kickers like Billy Cundiff can be a big detriment to the team.
#125 by nat // Nov 26, 2014 - 3:06pm
I thought I'd be clever and look at drive stats to see what Denver was doing on offense that New England wasn't doing to earn its higher VOA (26.2% vs. 15.0%) over the 2014 season.
Tm. Drs [Yds/Dr ] [Pts/Dr] [TOs/Dr] [INT/Dr ] [FUM/Dr] [LOS/Dr .] [ DSR ]
DEN 127 35.69 (7) 2.53 (3) .079 (4) .071 (13) .008 (1) 26.73 (19) .753 (3)
NE. 122 34.61 (8) 2.70 (2) .074 (3) .049 (6 ) .025 (6) 30.22 (5 ) .752 (4)
That's stunningly close to identical, certainly so in the most important Yds/Drive, Pts/Drive, TOs/Dr, and Drive Success Rate.
Whatever Denver's offense is doing or not doing that New England's offense is not doing or doing, it isn't having any effect on success at the drive level. None.
It really looks like VOA is valuing something - we don't know what - that isn't of any practical value. At least in the case of these two teams.
My guess is that this is an artifact of some grading curve-like adjustment. Perhaps there's a stadium adjustment to the VOA baselines in Denver. Perhaps Denver's comeback against Miami or fall-far-behind-then-keep-pace against New England put them in a bucket filled with weaker teams and thus easier baselines.
It is interesting. Maybe Denver isn't all that VOA claims it to be. Or maybe they have mastered a way to get better stats without getting more success. Or maybe they are just unlucky in some invisible way.
I'd love to hear ideas. Because this is getting weirder the more I look into it.
The last time I saw something this weird, it turned out that FO had given the NFC home field advantage in the Super Bowl for the playoff odds simulations.
#128 by nat // Nov 26, 2014 - 3:44pm
It's got to be something like that. Hmmmm....
Denver averages 5.81 plays per drive. New England averages 6.38. Without any other advantages.
That could explain the 11.2% VOA difference. We'll see what it means as the season progresses.
#137 by Rick_and_Roll // Nov 26, 2014 - 5:09pm
In the Denver New England game, there was a big difference in starting field position due to Denver's turnovers and special teams woes. I also remember the Pats punter being awesome in that game.
Trying to simplify it, the difference in DVOA and score could come down to three things that may be underweighted in DVOA calculations.
1. The impact of Special teams could be under-weighted in some games, because the Pats dominated Denver in special teams.
2. Average starting field position
3. The difference +/- in 3rd down conversion and stop rate compared to the teams season average.
#145 by deus01 // Nov 26, 2014 - 8:51pm
Well starting field position is really a function of offense and special teams strength plus some randomness; you can't really predict when a receiver is going to tip a ball high in the air. The 3rd down conversion and stop rate could be interesting. Are there teams that are more successful on 3rd or 4th down on average then they are on first or second? Maybe that's an indication of bad play calling on early downs, or maybe teams like NE under utilize players like Gronk on early downs (maybe to try and keep them healthy) but rely on them more when they need a big third down conversion. It would be interesting to see each teams success on each down and determine if there is a significant difference.
#139 by intel_chris // Nov 26, 2014 - 6:31pm
Looking at your numbers, I was going to suggest INT/Dr, but then noted INT+FUM/Dr, and said that isn't likely it. And, you are talking VOA not DVOA, so it isn't opponent adjustments.
My guess it might be 1st/2nd down "success rate" v. 3rd down rate. If Denver's numbers were somehow coming from more success on earlier downs with fewer big plays, that might positively effect VOA, while the lack of long plays would prevent any plays from being capped. In contrast, if New England had a few back-breaking long third downs making up for "unsuccessful" early down plays, those might have gotten capped by VOA thus not raising New England's VOA while still making the yardage to come out the same.
As to baselines, I doubt that the FO stats are that well tuned. I'm sure they do splits for down-and-distance, but I think that is rolled into their "successful play" metric. I suspect their play metrics are roughly linear with distance, at least piecewise linear with only a few distinct line segments, e.g. < 0 yards, 0 to successful yards, successful to 20 yards, > 20 yards. I would be surprised if it were some more complex curve with multiple parameters, no f(yards, down, yards to goal, points for, points against, temperature, domed stadium, natural turf, home field, ...).
#108 by Tim F. // Nov 26, 2014 - 12:06pm
"But it's very odd that Denver's rating seems to depend on the proposition that its three point nail-biter comeback over Miami was as dominating as the Patriots' twenty-two point shellacking of the Broncos."
You find it odd because your own proposition is wrong. DVOA doesn't claim to measure "dominance" — it's measuring efficiency in an idealized scenario. If anything, it attempts to subtract out "dominance."
#67 by mitch // Nov 26, 2014 - 6:41am
Remember Denver being the 6th best in DVOA history ? They were coming off what proved to be the best run of games this season.
Now it will be NE coming off their best run of games. Looking back in a few weeks NE won't look as good as they do now, these web sites can't figure this out. That's why many times you'll see a new no.1 team in their rankings fall just shortly after becoming no. 1, this happens all the time if you pay attention to it.
And remember the research done some years back comparing weighted DVOA to DVOA in the playoffs, it concluded that DVOA covering the entire season, not the weighted DVOA was more accurate than the weighted DVOA, and you can google that and find it, it's still there last I checked.
And now they make NE the super bowl favorite based on a meterics that is less accurate in the playoffs.
Got me on that one, they don't even use their own research properly, and don't believe in their own research.
What happens is, because NE appears to be playing so well that sways their opinions to not beleive in their own research and that'd be a big mistake on their part. Stay with what the research has shown, Denver is the team to beat and will almost certainly beat NE in the playoffs if NE gets that far which they may not.
NE is not and should not be the favorite to win the super bowl at this time. At the end of the day, bet I'll be proven correct.
#71 by Otis Taylor89 // Nov 26, 2014 - 7:41am
Really? Seriously?
NE looks like a train to me. I think when you look back in a feww weeks they will look a lot better than they do now, as their early season malaise games are given less weight. They've made some very good teams look like crap many different ways and they will probably play everyone at home, a place that they are very tough to beat, in the playoffs. DEN is very tough at home, but will probably need to beat some pretty good teams on the road to get a chance to play a game at home during the playoffs, plus BUF at home who has been giving up an average of 147yd/game passing since the NE game.
#93 by Rick_and_Roll // Nov 26, 2014 - 10:49am
While I agree with some of your points, don't forget that Cincinnati, Seattle and Denver have also looked like trains at some point this year, as the season is cyclical.
Denver would have to completely fall apart to not earn the second bye. They have the tie breaker over the Colts, and the AFC North still has a lot of games against each other, and I can't see one team in that division running away with it, its just too competitive.
I agree that New England should be the favorite because they will likely have home field advantage, thanks in large part to their Denver game being played in bad weather in Foxboro AGAIN.
#76 by BJR // Nov 26, 2014 - 9:09am
There are some fairly ridiculous statements here.
"Denver.....will almost certainly beat NE in the playoffs". Vegas had the last Pats/Broncos game as ~45/55, and that was before the Pats won handily and went on their current tear.
"NE is not and should not be the favorite to win the super bowl at this time. At the end of the day, bet I'll be proven correct." Of course the likelihood is that NE will not win the Super Bowl and you will be "proven correct". Even if we agree they are decisively the best team they would still only be around a 25% chance to win it all.
You have valid points about weighted DVOA/DVOA and the dangers of judging teams at the very peak of their performance, but go easy on the hyperbole.
#129 by intel_chris // Nov 26, 2014 - 3:45pm
Your point about streaky NFL teams is fair. It is not clear that the current NE streak will last all the way to the SB. However, that said, I'm more in line with those that believe that it will last long enough to ensure the #1 seed. I think 15-20% odds of a NE SB victory are not *too far* exaggerated.
What I wish I had was something that would predict who was likely to have a sustained hot streak starting in a week or so. I can think of two teams, BAL on 2012 and NYG in 2007 that rode such streaks all the way to SB victories. It would be nice to be rooting for a SB winner, which I haven't done recently. Ok, yes I rooted for BAL in their SB win, but only because they beat my team in the playoffs, and thus I could say my team only lost 1 playoff game and that was to the eventual SB winner, so it was arguable that my team might have been 2nd best that year.
Fandom is a cruel mistress.
#69 by eagle97a // Nov 26, 2014 - 7:22am
I won't pretend that I know the inner workings of DVOA or even a general idea on its methodology but my 2 cents is people are getting worked up on DVOA's divergence from the supposed "general" consensus and/or eye test when in fact other football analytics sites are also puzzled. One of them was Sterling Xie of AFAs' team efficiency rankings 2 weeks ago when they had NE ranked as average. Not to sound repetitive but as I said before matchups matter, team development,coaching, roster construction etc. are just some of the intangibles that can't be factored in and some like injuries are imperfectly captured. But more importantly the way these "intangibles" interact with the other more measurable traditional metrics and how luck factors in the equation together with sample size considerations makes this exercise fiendishly difficult. Nevertheless I still enjoy all of these and am really looking forward to the rest of the season.
#72 by BDC // Nov 26, 2014 - 8:36am
Eagle97a, it isn't that simple. This comes every now and again and is always dismissed for various reasons, but as Vandal pointed out DVOA is around 62%. Every single year I run the numbers (since 2003 when I started here) and if memory serves, DVOA's best year was like 66%. To put that in perspective, you can pull any half way competent fan out of a sports bar and they can pick winners at a 66% clip.
To put another way, imagine if I devise a system for predicting the outcome of rolling an ordinary 6 sided die. All these calculations go into it, tons of work, years of adjustments....and at the end of all that I correctly predict dice rolls 1 out of 6 times.
There comes a point in time where you just have to accept that whatever merits the system may have, it is simply not at all useful as a predictive tool.
#75 by eagle97a // Nov 26, 2014 - 8:57am
Therein lies the problem, using DVOA as a predictive tool. For me the beauty of DVOA lies in its descriptive and explanatory power. I find their explanations regarding past performances very enlightening. I'm less sanguine about its predictive powers because of reasons I've enumerated before plus the changing nature of football itself with re to rules changes/emphasis, practice regulations changing etc. makes me circumspect about its season projections and predictions. For me I'd rather enjoy the unpredictable nature of the NFL and let them play to win the game instead of constantly comparing real world results with spreadsheet projections. But that itself is also rewarding for me since any divergence between the two leads to the question why which in turn drives better analytics for better explanatory and predictive power for the future.
#78 by bravehoptoad // Nov 26, 2014 - 9:26am
Yeah, I really think you're missing the point of DVOA, which is its granularity. Something like Pythagorean Expectation has none, for instance. All DVOA has to be is in the same ball park for basic win prediction for that granularity to shine.
#83 by Pat // Nov 26, 2014 - 9:45am
To put that in perspective, you can pull any half way competent fan out of a sports bar and they can pick winners at a 66% clip.
What makes you think it's possible to pick winners better than that?
What makes you think it's a good idea to judge a metric's predictivity on win%? It isn't. Statistics don't care about betting. The best way to judge a metric's predictivity is how well it predicts how well it's doing.
#100 by BDC // Nov 26, 2014 - 11:24am
Then what's the point of all the numbers and all the computations if it is no better than the average fan can do? This has nothing to do with betting, just correctly predicting the winner of a football game. If it is no better than any random fan can do whats the point?
"The best way to judge a metric's predictivity is how well it predicts how well it's doing."
Thats....just no. If I came up with a metric at the beginning of the year that predicted Jax was the best team in the league, and then I ran the numbers this week and sure enough it still says JAX is the best team in the league, you're saying that system is a successful one?
If so I take back everything I said, I see why you love DVOA,
#110 by Pat // Nov 26, 2014 - 12:39pm
Then what's the point of all the numbers and all the computations if it is no better than the average fan can do?
One thing that DVOA is really good at is separating offense, defense, and special teams performance. Very, very often it doesn't correlate with 'traditional' thinking, and usually it's because 'traditional' thinking is wrong. (Example: people often talk about how strong the Eagles special teams are, but I don't think people realize that it's actually the strongest unit on the team).
But, more in general, what's the point of paying attention to the yards that a team gains? Yards/reception of a player? Fumbles? Interceptions? What's the point of QBR, or any other statistic that doesn't predict wins better than an average fan can do?
Because it helps describe what actually happened in the game and helps to inform what to look for in future games. It is a game. Why do you want to predict the result? What the heck fun would that be?
(Oh, wait! That's right. Betting again.)
Thats....just no. If I came up with a metric at the beginning of the year that predicted Jax was the best team in the league, and then I ran the numbers this week and sure enough it still says JAX is the best team in the league, you're saying that system is a successful one?
You keep saying "best". What does "best" mean? It's an abstract concept. I can't test 'best'. Best at scoring the fewest points relative to the other teams? Then I'd say good job!
DVOA isn't an arbitrary magic number that means nothing, like an Elo rating. It means something. VOA is "value over average": the 'value' there correlates with points, and "value per play/average value per play" gives you a measure of points/play. DVOA correlates to average point margin, per play - an offensive DVOA of +20% means that the team scores 20% more points per play than the average team. Per game, you'd expect a +20% offensive DVOA team to score around 26.4 points/game (assuming average game length: games have variable length).
So testing DVOA would result in looking at how well it predicts the point/play (points scaled by total number of plays) differential between two teams, not winning percentage.
#124 by intel_chris // Nov 26, 2014 - 2:57pm
Pat I think you are over-stressing your point. We do want a prediction of how a game will go and how good a team is. We may not want a guarantee, in some kind of deterministic sense, which isn't even possible given the results of Quantum Mechanics. However, we would like a prediction, i.e. we would like to know something.
Your point of knowing the reliability of our metric (i.e. how confident we are that the prediction will be reflected in the actual result) is relevant, but that is because we are interested in the actual result and having a prediction of what that result may be. A metric that predicts with 100% accuracy that Brady's jersey number will be 12 and PM's will be 18, is not interesting because we aren't interested in that fact--that isn't the result we care about. We are interested in who will (or did) win the game. We want a metric that tells us that fact. Thus, VOA, DVOA, Elo, Win%, Power Rankings, Beat Graphs, et al, are interesting only in as much as they are relevant to who won the game (and who would be likely to win a rematch).
And, within that, I will agree that a metric like VOA, which stresses some things not on the scoreboard and downplays some things that affect the score, may give a different version of who "won" than the actual score and that's ok, particularly when it impacts who would be likely to win a rematch (or some other future game against a different opponent). However, in the long run, we care about winning. There is no point in playing the game, if there is no winner. VOA is ultimately useful because it informs us about winning. If it didn't, if it informed us about Jersey numbers, or some other irrelevant fact, we wouldn't care.
Ok, now I'm over-stressing my point....
#153 by Pat // Nov 27, 2014 - 12:43pm
Statistics need two things: they need explanatory power (they have to explain the observed variance in some measurement) and predictive power (future situations need to continue to have variance explained).
"Silly" statistics, like 'predicting Brady's jersey number' have no explanatory power (because there is no observed variance, so any statistic which attempts to correlate with Brady's jersey number would explain none of the variance).
DVOA has explanatory power (explaining point differential) and predictive power (predicting future point differential). In fact, it's known that it doesn't correlate as well with wins (hence winning percentage) - as other, more situational statistics, hence the existence of estimated wins. But estimated wins isn't as predictive as DVOA, hence the existence of both.
I've always found that fact really interesting: but what does it mean that DVOA doesn't correlate as well as estimated wins? It means that it's a game, and the things that are sustainable (every-down performance, performance all over the field) aren't always what determines the game.
This is why I have no idea why people are so obsessed with predicting who wins. The fact that the "best team" - the one with the most sustainable performance - doesn't always win is what makes it exciting.
#155 by intel_chris // Nov 27, 2014 - 1:27pm
I heartily agree with most of what you are saying. We definitely have different stats, because the things which are sustainable and lead to winning in the long run aren't the only, nor even necessarily the most important factors in winning a given game (i.e. in the short run).
The fact that its a game and one plays a game to win. That's why we want to know who will win. Bettors bet because they think they know something about the game. That's exciting too. It's a game about the game. (I know this isn't your favorite model, but stick with me.)
And you are right, the "best" team doesn't always win "and that's why we play the game". That makes it exciting, because there is a challenge to do better than you are expected to (and to root for a team you are hoping will do better than expected).
That goes for "the game about the game" also. The more you know, the better you can predict (guess) who will win the game. It will never be exact. There will always be risk. The "better" team (no matter how you define better) will not always win. However, you can learn from your mis-predictions, your losses in "the game about the game", and get better at it too.
So, while I'm not a gambler, I can understand wanting to know more, and I think it is fair to argue that knowing more will make you a better predictor of who will win a given game. I agree that we are not likely to ever know enough to be perfect predictors and that's a good thing because it means playing the actual games is meaningful. However, at the same time, I think we are far from our limits of prediction and the desire to know who will win is not a bad thing.
So, again over-stressing my point, don't be surprised that people, even non-gamblers, want predictions, because predictions are the result of knowledge, and knowledge is a good thing. Don't worry that the excitement of the game will ever be taken away. We won't ever get that good, but wanting to get that good is a good thing.
#156 by BDC // Nov 27, 2014 - 3:15pm
Pat, I think the point though is that we are starting with a baseline rate of around 65ish percent to pick winners. Any fan can do this. If DVOA was at ALL successful, we would expect it to be better than this. No, we don't expect it to be perfect. But if it is no more successful than your average fan then it is not successful.
I think the flaw here is you are making two assumptions. 1. DVOA successfully explains what has happened in the past. 2. The things that DVOA successfully explains in the past correlate well with future wins. The problem though is that if both of those things are correct, we would expect DVOA to be at least somewhat more successful at predicting winners than polling drunks in a sports bar. The fact that is is most certainly NOT more successful than that tells us that one of these two things is wrong. Either it isn't nearly as accurate at telling us who played well in the past, or it is measuring things that don't have nearly as much to do with winning future games as we think they do.
This comes up from time to time, see the article discussing broadcasters using advanced stats. And the answer typically given to why people don't get is some combination of condescension, people being stupid, not understanding stats, etc. And that would be all well and good if the system actually worked, but the average person looking at all of this sees a baseline "success" rate at picking winners of 65%, and says well if I can do that myself already, why spend all this time on advanced stats that don't actually tell me anything I don't already know how to do myself. See where the issue is?
#168 by Vandal // Nov 28, 2014 - 2:26pm
http://nflpickwatch.com/
64-65% is basically the mid point of all talking heads. Vegas favourites are 67%, and the best stat sites and experts are around 70%. FO is around 62%, sourced from their premium section. Which puts them last of all the systems, behind:
4. Fivethirtyeight
5. Accuscore
15. Vegas Favourites
22. ESPN Numberfire
30. ESPN Team Rankings
38. Whatifsports
46. Prediction Machine
T-81. Football Outsiders
#169 by Tim F. // Nov 28, 2014 - 5:45pm
Didn't Aaron explain that the Premium Picks don't use DVOA and are dependent on several proprietary algorithms that he knows need to be tweaked? So — why are people still blaming DVOA as bad at predicting game results?
#174 by Vandal // Nov 29, 2014 - 3:53pm
No.
Cut and Paste from the Premium Picks page. Implying that it is more accurate, than ONLY DVOA. Which implies that DVOA would actually perform Worse, without the modifications this year...
"are based on an equation that measures current DVOA (using certain specific splits, not just each team's total) as well as weather, injuries..."
#180 by intel_chris // Dec 03, 2014 - 6:25am
However, extending that from Vegas favorites (or ESPN power polls) to any random Sports bar denizen is a stretch. Most people only have an opinion on one or at most a few key teams they follow, so the numbers are biased. Sure I suspect I could pick reasonably accurately on DEN and NE, but ask me about TB v. HOU and I doubt that I would guess that right that often. However, if you asked in HOU or TB, I bet you would get a reasonable guess. Now, compare that to DVOA metrics, it doesn't get to pick which teams it focuses on, it covers them all. Thus, it has to be right across the entire spectrum. So, therefore DVOA is designed to be consistent across all teams and all years, by being an objective measurement of one aspect of the teams performance--that is the ability to move the ball consistently and prevent the opposition from doing so. It measures that and just that. When that metric corresponds well to winning games, DVOA does well. When it doesn't, it does poorly. However, it always measures the same thing.
Now, I know your issue is that the DVOA metric doesn't appear that accurate compared to other "universal" ratings systems, e.g. Elo, or ESPN power rankings. I want to separate such systems into three groups, and deal with each distinctly.
1) objective systems, like Elo and DVOA. These are directly comparable because they take some objective measurements and run them through a formula to get an answer. Each formula has a model behind it and one of those models is most accurately predicting the games, but just like mutual funds, the games changes over years and last years top-performing mutual fund is not likely to be this years top performer, but some funds perform well across may years. Likewise, this years top-performing metric will not likely be the top-performing metric next year, because the teams will have adjusted their playing styles, players will get older, new players will be drafted, etc. However, some models will do well over many years. If you don't think some model is working well, don't use it, just like if some fund doesn't perform well, don't invest in it. (Too make this part of the argument more precise, we need to restrict our mutual funds to index funds, which have specific required investment styles that are followed rigorously.)
2) Expert opinions like Peter King's "fine fifteen". Those are based on whatever subjective judgement the expert is applying each week. They are not comparable to objective systems at all. They are totally the whims of the particular expert. As much as the expert knows his topic. (Most mutual funds are actually like this, as there is some particular manager running them and they set the style, within certain parameters.)
3) Polls like the ESPN Power Rankings, or Vega odds. Those try to combine numerous opinions to form a consensus. These spread the risk around, because you don't have just one expert, but many opinions. However, such polls tend to be influenced by what is "hot".
Given this, lets look back to those two games where Rothlisberger got hot and then ran into the Jets. After those two games, the Steelers ratings probably soared in most subjective polls. They may have soared in certain objective polls. However, at least one objective poll probably discounted that "abnormality" and didn't over-rate the Steelers going into the Jets. That's what we are looking for with metrics like DVOA. Moreover, we know there isn't just one metric that captures that subtlety. You need many metrics emphasizing different aspects of the game. Then, you can form you own "expert" opinion by looking at the different metrics and see what they are telling you. DVOA not so accurate this year, ok what it emphasizes hasn't been important in deciding games. Is it useless because of that, no. It's just emphasizing something that hasn't been the key factor.
The correct perspective is to see DVOA as one of your tools for evaluating teams and their performance, and letting it be one of many. To look at it from a gambling perspective, pick you favorite rating system, I don't care which one. Now allow me five different rating systems, one of them being DVOA, and none of them being yours, and bet me that your system will consistently beat all five of mine, that is you place a bet the way your system indicates is right and I will accept that bet only if the majority of mine predict the reverse. Personally, I want DVOA in that set of five, because even if it is a dissenting opinion, it is measuring an important aspect of teams performance, one that I believe is relevant to how many games a team wins. If DVOA tells me that team A is playing better than team B, I would be cautious of betting in favor of team B. Would I base my bets solely on the DVOA metric? No. Do I value what it is telling me? Yes.
Note, that one is recommended to do the same things with one's retirement money, diversify one's portfolio. Don't invest it all in the one fund that did best last year. Put some in US stock funds, some in bongs, some in foreign funds, some in large caps, some in small. Each of those funds takes a different metric on where the market might go. If one of them does poorly one year, the others probably won't. Equally importantly, if you have one fund that takes off, you haven't missed that rise, even if for the past ten years the fund has had only middling performance.
So, don't discount a metric just because it only gets 60% of the answer right, that's still right more than half-the-time. It is still winning better than one could by tossing a coin. Some weeks that statistic may be the one which makes you look like a genius, knowing that the AZ Cardinals were not the top 5 team their record was suggesting. Knowing the dissenting opinion is valuable, especially when the consensus has missed something.
#170 by RickD // Nov 28, 2014 - 10:17pm
The best way to judge a metric's predictivity is how well it predicts how well it's doing.
I use the coin flip method. I predict that, in the long term, my rate of being correct will converge to 50%.
I doubt any other method can "predict how well it is doing" as well as I can.
#178 by Pat // Nov 30, 2014 - 10:35pm
Absolutely! Of course, I said that's how you judge a metric's predictivity. Not its usefulness.
Another way to say that is that a metric needs to be both *correlative* (high R-value) *and* causative (predictive).
Obviously, a coin flip prediction method isn't correlative.
#80 by Mugsy // Nov 26, 2014 - 9:36am
Maybe we can start a new thread somewhere where people can bicker about the intricacies of dvoa-- kind of like the old brady vs manning thread.
It gets kind of old listening to someone kick and slap around a dead horse.
#84 by Pat // Nov 26, 2014 - 9:53am
It's the new football stat nerd metagame. Instead of actually rooting for teams, they start rooting for stats. You get all sorts of new statistics myths, like "expected wins is everything" and "at the end of the day, win% is all that matters" instead of actually understanding what things are trying to do and what things mean. I would love to keep my cynicism out of it, but you almost always hear "against the spread" or "spread picks" in threads like this, which just shows where the silliness is coming from. Some of us actually enjoy understanding the game, you know, and don't gamble.
Seriously, it's hilarious. Baseball has a bajillion statistics. They all measure things in different ways. You don't have people saying "WAR is useless!" or "VORP is just fundamentally flawed!"
Apparently we'll need to have a Statistics Outsiders soon.
#88 by jacobk // Nov 26, 2014 - 10:13am
Baseball's statistics are massively superior to anything anybody has come up with for football. The reason nobody is saying "WAR is useless" is because WAR works.
The whole point of DVOA is to tell us which teams are good. If DVOA's claims that Team A is better than Team B are not consistently borne out by the results then DVOA is not doing its job.
#90 by Pat // Nov 26, 2014 - 10:29am
See? Definitely need the Statistics Outsiders. There's another statistics myth: "the whole point of a metric is to tell us which teams are good."
What makes you think it's possible for one number to uniquely describe a team? What does "good" even mean? Which team wins most? Football results are about 50% non-predictive. Why would it make sense to focus only on half of the game?
The reason nobody is saying "WAR is useless" is because WAR works.
Then why does any other metric exist in baseball? Why even talk about xFIP for a pitcher, or VORP? WAR works. Just use that!
They exist because they all measure different things. It's a game, made up of a bunch of people. It doesn't boil down to one number. Baseball stat fans understands that. Football stat fans don't.
#149 by jacobk // Nov 27, 2014 - 9:07am
Ok, let me ask you this: why do you care about DVOA? Will it make you happy if your team is awarded the DVOA Lombardi? How does studying DVOA help improve your knowledge of football?
Every week this site makes a big deal out of putting up its DVOA rankings. In the accompanying text it takes notice whenever a team is doing historically well or poorly per DVOA. It's almost as if the designers of DVOA think it can be used to rank teams, and that ranking is meaningful.
DVOA is a team statistic. xFIP, VORP, and WAR are methods of measuring individual performance that are, yes, used to evaluate which players are good and which players are likely to be good in the future.
Baseball fans get pretty upset if their GM signs a player to a contract that pays significantly more per WAR than the average free agent deal. Even though a player might be a great guy and is more than just a number. Why is that? Because they think that WAR roughly measures player quality and they don't want their team to overpay.
The only possible point of DVOA is to tell us which teams are good. There's the sub-point of telling us what they are good at. However, if we can't trust DVOA to tell us which team is good, it's hard for me to see how you could trust the decomposition of DVOA to tell us which offenses and defenses are good.
Now, we could do what you have implicitly done and define "good" as "good per DVOA." Then as long as DVOA stays self consistent over the course of a year it looks like a useful metric. The problem with that is that there is no inherent reason to care who DVOA says is good. The reason we care is if DVOA's sense of who is "good" matches up with who wins actual games.
#152 by Pat // Nov 27, 2014 - 12:26pm
How does studying DVOA help improve your knowledge of football?
DVOA splits teams into offense/defense/special teams, and situational behavior, and removes high leverage non-predictive events (fumble recoveries, non-predictive penalties, etc.), and helps to identify teams that are getting 'clutch' lucky (better than their average performance in high-leverage situations).
The only possible point of DVOA is to tell us which teams are good.
See above. I could not disagree more with this statement.
#154 by intel_chris // Nov 27, 2014 - 12:50pm
Why do we care about DVOA?
Simple, because DVOA tells us which teams are good at consistently moving the ball (on offense) and keeping the other team from moving the ball (on defense). That's what DVOA measures.
Those are not the only factors in winning football games. However, do you really want to argue that teams shouldn't try to do that, and they shouldn't be measured on how well they do that?
Note, that I did not say scoring. DVOA does not directly measure scoring. To that extent DVOA will always be a "flawed" metric for those who want to predict games and their scores. I believe it has been adjusted slightly to better correlate to scoring, but there are limits to how much you can adjust it without diminishing its capacity to measure what it is intended to measure.
Different metrics measure different things. If you want to know who is ahead in the playoff race, use W/L records. If you want to know who has outscored their opponents best, use points-for (PF) and points-against (PA). If you want to know who has accumulated the most yards, use a yards metric. If you want to know who has been most consistent in slowly piling up advantages in yards play-after-play, use VOA. If you want to account for the strength of the opponents, use DVOA. So on and so forth.
One size does not fit all. Use the stat that tells you what you want to know. Getting frustrated with the wrong stat isn't useful. It is simply measuring something different than what you want. Instead find the right stat, knowing there is no "perfect" stat.