NFL Divisional Rematches Study
by Scott Kacsmar
The NFL's Week 15 schedule won't win any awards for originality. Eleven of the games are divisional rematches from earlier this season, including two matchups just played on Thanksgiving. Since 2010 the league has used the entirety of Week 17 for divisional games, but setting those aside, this is only the fourth time since 2002 that at least 11 divisional games have been scheduled.
So familiarity is on display more than usual this week, but how much information from a first meeting can we really use to predict expectations for a rematch? Sure, it feels like Seattle can hold San Francisco to three points again this week, but what about Dolphins-Patriots? In Week 1, the Miami pass rush got to Tom Brady in the heat and Knowshon Moreno carried the offense to a 33-20 win. Those elements won't be present Sunday in Foxboro, nor will the Patriots have to worry about a recovering Rob Gronkowski, a suspended Brandon Browner, or Branden Albert at left tackle for the Dolphins.
Teams can change so drastically during a season. I contacted Giants offensive lineman Geoff Schwartz and asked him if first matchups are a major focus in rematch preparation or if teams care more about recent performance. "I'd say probably more about how the team is playing lately rather than just focusing on the one game," Schwartz told me. "Things can change so much over just a few weeks -- injuries or maybe a change in play-calling. Divisional opponents know each other so well already that I'd say the toughest part is finding wrinkles each time you play. In the end it's just making plays against what you know will happen."
I used the results of all 1,152 divisional games played in the 2002-2013 regular seasons to study the changes from one meeting to the next. Some of the data is presented in even splits between 2002-2007 and 2008-2013 just to test for consistency.
Full disclosure: it's tricky business to use even 16 games to predict what will happen in one NFL game, let alone doing a one-for-one comparison. I did not expect to find significant correlation between each team's performance in the first round of divisional games and that same team's performance in rematches, but sometimes finding nothing is a story itself. Not finding any correlation at all would suggest NFL outcomes are completely random, which I don't think any of us wants to admit even if some weeks suggest that's the case.
|Special Teams DVOA||0.01||-0.01||0.03|
I tried regression, but the r-squared value was only 0.04. The p-value was very small for the scoring correlations, so at least the weak correlation exhibited there is not likely occurring by random chance. We don't expect a team that scored seven points to suddenly rack up 49 in a rematch, for example. I thought it was interesting that the numbers in the 2002-2007 period were a little stronger than the last six seasons, which have featured more passing (higher variance), more scoring and more big comeback wins.
So scoring at least has a non-zero correlation, but DVOA did not correlate well from game to game. The p-value was pretty close to 0.05 as well. At least we still see the common relationship in the numbers that suggest offense is more consistent than defense or special teams.
The home team has won 55.6 percent of divisional games from 2002-2013. Maybe the familiarity with each other has an impact there, because home teams have won 58.6 percent of non-divisional games and 59.9 percent of non-conference games (the least familiar opponents a team will play). When teams know each other so well, something like home field should become less important because even the experience of playing in that environment is already known to the opponent.
Something we often hear about in rematches is that the losing team had more to learn from the first game. That makes sense in theory. Schwartz notes that "it's also easier to game plan wrinkles in a rematch if you lost. The scheme advantage would go to them because the winning team isn't changing much. They will use what got them there."
Schwartz played in two very high-profile rematches in recent years. In 2012, Schwartz and the Minnesota Vikings beat Green Bay 37-34 in Week 17, then matched up again just six days later in the NFC wild card game. Schwartz says the Vikings ran the exact same game plan, but the quarterback was different. Minnesota had to play Joe Webb instead of an injured Christian Ponder and lost 24-10. Last year with Kansas City, Schwartz and the Chiefs lost 23-7 in Week 16 to Indianapolis. Just 13 days later the Chiefs raced out to a 38-10 lead in the AFC wild card game, even after losing Jamaal Charles to injury, before falling victim to a monster Colts comeback. "We played badly in the Week 16 game, but I knew we could score," Schwartz said. "We ran way more sub/nickel runs in the playoff game so they couldn't pack the box. Especially way more read option than usual. That was a wrinkle we could add."
In theory it should be easier for you to correct your own mistakes than for the opponent to repeat their performance or tinker with a new way of finding success. However, what we often see is that one team is simply better than the other, and that gap is too much to make up regardless of what the first game taught the teams.
The first 576 divisional series in the eight-division era have featured 328 sweeps (56.9 percent). So more often than not, the team which wins the first game wins the second game as well. Here's a further breakdown.
|Home Team in Divisional Games, 2002-2013|
|Record in Divisional Rematch, 2002-2013|
|Won first game (all)||328-246-1||0.571|
|Won first game (home)||162-155||0.511|
|Won first game (road)||166-91-1||0.645|
Something interesting here is that when the home team wins the first meeting, that team is only seven games above .500 in the road rematch. However, when the road team is able to win the first game, they win 64.5 percent of the time in the rematch, which is considerably higher than the normal home-field advantage demonstrated in the upper half of the table. A good sign of a superior team is to come out and win the first meeting regardless of being on the road, which is what we saw happen on Thanksgiving with Philadelphia and Seattle winning in Dallas and San Francisco.
Here is a summary by division of sweeps and splits since 2002.
|Divisional Sweep and Split Summary, 2002-2013|
|NFC North||44||8 (GB vs. DET, MIN vs. DET)||8 (CHI vs. MIN)|
|NFC East||44||7 (PHI vs. WAS)||6 (NYG vs. PHI)|
|NFC West||43||7 (SEA vs. STL, SF vs. ARI)||7 (ARI vs. SEA)|
|NFC South||41||6 (CAR vs. TB, NO vs. ATL)||6 (ATL vs. CAR, ATL vs. TB, NO vs. CAR, NO vs. TB)|
|AFC South||40||8 (IND vs. HOU)||8 (JAC vs. TEN)|
|AFC North||39||9 (PIT vs. CLE)||8 (BAL vs. PIT)|
|AFC East||39||10 (NE vs. BUF)||7 (BUF vs. NYJ, MIA vs. NE, MIA vs. NYJ)|
|AFC West||38||7 (SD vs. OAK)||8 (DEN vs. KC)|
|Note: Splits are 1-1 only and do not include ties (2012 SF-STL, 2013 GB-MIN)|
Not too many surprises here, though I am not sure why the NFC has more sweeps than the AFC when the division winners have been more consistent in the AFC.
[ad placeholder 3]
I am not the biggest fan of using a game's final score for analysis, but it's an option we always have readily available.
- The average final score in a first divisional game is 27.1-15.6 (42.7 total points)
- The average final score in a divisional rematch is 27.2-15.3 (42.4 total points)
Total scoring in the NFL is pretty consistent, but what's the average turnaround in these games? The answer is 4.2 points. Given the estimated value of home-field advantage to be around three points, this number makes sense.
When the first game is decided by a one-score margin, 49.0 percent of the rematches are also decided by 1 to 8 points. First games decided by at least 17 points (three scores) result in a three-score rematch 28.8 percent of the time.
Do teams that win in blowout fashion tend to win the rematch more often and by bigger margins? They sure do, though the bigger margins tend to regress to a closer game.
|Point Differential (PD): First Game vs. Rematch|
|30+ pts||26||33.2||16-10 (.615)||8.8||30+ pts||2||7.7%|
|20-29 pts||92||23.9||60-32 (.652)||4.0||20+ pts||13||14.1%|
|9-19 pts||168||13.1||98-69-1 (.586)||3.1||9+ pts||54||32.1%|
|4-8 pts||160||6.3||89-71 (.556)||2.5||4-8 pts||24||15.0%|
|1-3 pts||129||2.5||65-64 (.504)||0.9||1-3 pts||14||10.9%|
When a team wins the first game by 1 to 3 points, then the rematch is basically a coin flip. Only two teams were able to sweep an opponent with a pair of 30-plus-point wins: the 2007 Patriots over the Bills, and the 2002 Falcons over the Panthers.
There have been six turnarounds of at least 50 points, and the Patriots were involved in half of them.
|50+ Point Turnarounds, 2002-2013|
|2003||Patriots||Bills||Bills 31-0||1||Patriots 31-0||17||62|
|2012||Seahawks||Cardinals||Cardinals 20-16||1||Seahawks 58-0||14||62|
|2010||Patriots||Jets||Jets 28-14||2||Patriots 45-3||13||56|
|2005||Redskins||Giants||Giants 36-0||8||Redskins 35-20||16||51|
|2009||Chiefs||Broncos||Broncos 44-13||13||Chiefs 44-24||17||51|
|2002||Patriots||Jets||Patriots 44-7||2||Jets 30-17||16||50|
The 2002 Jets-Patriots can be summed up by Chad Pennington rescuing New York's season after a bad start. I guess 2003 is known as the Lawyer Milloy game in Week 1 when Tom Brady threw four interceptions in a 31-0 loss. The Patriots returned the favor in Week 17 when Brady threw four touchdowns in a 31-0 win. The most shocking turnaround for New England was really against the Jets in 2010, which doesn't even feature the 28-21 playoff loss at home just weeks after that 45-3 drubbing.
For 2012 Seahawks-Cardinals, that was the difference in Russell Wilson making his first career start on the road versus Seattle developing into a juggernaut later that season.
(Ed. Note: Also, Arizona was stuck using their third-string quarterback, some kid named Ryan Lindley. Sound familiar? Remember, that says 58-0. -- Aaron Schatz)
So what can we take from Dallas losing at home to Philadelphia by a 33-10 final this year? Teams losing by at least 20 points in a home meeting are just 17-36 (.321) in the road rematch.
[ad placeholder 4]
The average change in single-game DVOA is 1.9 percentage points, but that's looking at both teams for every game, so we would expect a number close to zero.
Teams with at least 50.0% total DVOA for the game are 106-1 in the first matchup. The only loss was the 2008 Ravens (50.1%) in overtime against Pittsburgh. In the rematch, those teams are 48-59 (.449) with an average DVOA of -0.2%.
Teams with worse than -50.0% total DVOA for the game are 2-139 in the first matchup. The only wins belong to the 2007 Cardinals (-50.6%) over the Rams and the 2009 Bengals, who somehow beat Cleveland in overtime despite a -75.0% DVOA for the game. In the rematch, those teams are 65-76 (.461) with an average DVOA of -1.0%.
I am not sure we can draw anything conclusive here, but a really good or really bad DVOA seems to regress back to around zero, understandably producing a record just below .500 in the process.
The biggest increase in team DVOA actually does not belong to the 2003 Patriots against Buffalo, but that is second place (+162.7%), with Buffalo suffering the steepest decline (-161.2%). The biggest increase is 2009 Tennessee against Jacksonville (+180.1%). The Titans used their bye week to replace Kerry Collins with Vince Young at quarterback, and for some reason Tennessee usually played better defense in those games with Young.
As always, offense sustains itself better than defense. Teams with at least 50.0% offensive DVOA in first meetings went 21-13 (.618) in rematches with 4.7% offensive DVOA. Teams with at least -50.0% defensive DVOA in first meetings went 19-22 (.463) in rematches with -4.7% defensive DVOA.
Timing Is Everything?
The only true battle for first place this week is Cowboys (9-4) at Eagles (9-4). Philadelphia won decisively, 33-10, just 17 days ago in Dallas. Does the timing of the matchups have any impact? The longer the gap between the games, the more likely the teams will be different due to injury accumulation. I plotted the average scoring turnaround based on the number of weeks in between games.
The biggest scoring turnaround (12.4) is the maximum time between games of 16 weeks and the next biggest (10.4) is the smallest time of two weeks. That's really interesting, though there are serious sample-size issues with only 10 and eight cases respectively, including our biggest outlier of 2003 Bills-Patriots. But even the third-biggest turnaround (8.2) is neatly placed at the midway point with eight weeks in between games.
It's a good thing for the NFL that rematches are fairly unpredictable, because the games otherwise wouldn't be as interesting to follow. Sure, a series like Bills-Patriots can resemble Friday the 13th in its bland repetitiveness, but those NFC South rivalries can offer as much variety as a season of The Twilight Zone. That's why we keep watching and continue guessing what will happen. We know the characters and the plots, but the NFL always finds a way to keep things fresh anyway.