Writers of Pro Football Prospectus 2008

19 Nov 2012

FO Mailbag: The Effect of Rest on Game DVOA

Tyler P: In my opinion, football commentators and analysts tend to make a big deal out of the duration of time between a team's games. Teams with a bye or a long off week after a Thursday night game have an advantage, while teams coming off of Monday night games or playing in Thursday night games have a disadvantage. I can certainly see how more or less rest and preparation can affect a team, but I'm curious if you've done any DVOA analysis for teams on long/short breaks. How significant is the advantage/disadvantage and is it relatively consistent across teams and years? If there is a difference, is it more prominent for either defense or offense?

Let's tackle the easiest part first. Since divisional realignment (2002-2011), NFL teams have played 1,331 games on atypical rest (i.e., not seven days), which translates to 4.2 games per team per season. Therefore, looking at team trends amounts to small sample size theater. Also, looking at yearly trends devolves into small sample size theater once we realize that, for instance, there are an average of only 12.2 total games per season played on four (11.2) or five days (1.0) rest.

So, with that in mind, let's eschew the specifics in favor of the more general question: Does the number of days between games have an effect on performance; and, if so, is the effect more prominent on offense or defense?

The analysis required three pieces of information:

  • The date of each game over the past 10 years.
  • The total, offense, and defense DVOA of each team as of the week prior to the game, starting after Week 4 so as to limit the wild DVOA swings early in the season. (You can find these in our Premium database.)
  • The total, offense, and defense DVOA of each team during the post-Week-4 game in question. (You can also find these in our Premium database.)

From there, evaluating the questions at hand simply involved math. I calculated the number of days between games for each team each season (after Week 4), and calculated how much each team's game DVOA fell short of or exceeded their DVOA as of the week prior to the game. If more days between games translates to better-than-typical DVOA, we can conclude that there's a positive rest effect; and vice versa. As you'll see, it's not quite that simple.

Our first table shows the leaguewide trend (Note: positive means improvement in all cases, including defense):

Rest Days N TOT AGL Chg TOT DVOA Diff OFF AGL Chg OFF DVOA Diff DEF AGL Chg DEF DVOA Diff
4-5 122 -0.38 +0.2% -0.19 +2.2% -0.18 -3.0%
6 339 -0.32 +0.7% -0.17 -0.8% -0.15 -0.1%
7 2,739 -0.21 +0.3% -0.11 -0.9% -0.10 +0.8%
8 280 -0.11 -1.0% -0.05 -1.3% -0.06 -0.6%
9-11 122 -0.01 +2.4% -0.02 +0.7% +0.01 -0.8%
13-15 298 +0.44 +5.6% +0.22 +3.0% +0.22 +2.5%

Compared to the baseline of a typical seven-day rest between games, it's pretty clear that the only large effect comes on 13-15 days of rest. And of course, the only way a team can have that many days between games is if they have a bye. So really, this is just a bye effect, and the "AGL Chg" columns, which show the average change in Adjusted Games Lost from one game to the next, suggest that one reason for said effect is that teams get healthier during the bye week.

In terms of offensive and defensive splits, the story is basically the same. Teams coming off a bye play better on both sides of the ball, and getting healthier seems to be -- at least in part -- the mechanism for that improvement.

As with any sport, though, you can't ignore home field advantage. So let's see what happens after splitting the table into home teams and road teams (neutral-site games excluded this time; and again, positive means improvement in all cases):

AWAY
Rest Days N TOT AGL Chg TOT DVOA Diff OFF AGL Chg OFF DVOA Diff DEF AGL Chg DEF DVOA Diff
4-5 61 -0.51 -10.9% -0.29 -3.1% -0.22 -7.9%
6 144 -0.35 -8.6% -0.20 -5.4% -0.15 -3.2%
7 1,378 -0.25 -8.2% -0.12 -5.0% -0.13 -2.8%
8 145 -0.14 -9.0% -0.05 -4.0% -0.10 -4.1%
9-11 77 -0.13 -1.0% -0.09 -2.6% -0.04 -0.8%
13-15 140 +0.40 -5.5% +0.23 -2.6% +0.17 -2.0%
HOME
Rest Days N TOT AGL Chg TOT DVOA Diff OFF AGL Chg OFF DVOA Diff DEF AGL Chg DEF DVOA Diff
4-5 61 -0.24 +11.3% -0.09 +7.6% -0.15 +1.9%
6 195 -0.30 +7.6% -0.15 +2.6% -0.15 +2.1%
7 1,351 -0.18 +8.9% -0.10 +3.4% -0.08 +4.3%
8 135 -0.08 +7.5% -0.06 +1.6% -0.01 +3.2%
9-11 45 +0.18 +8.2% +0.09 +6.5% +0.09 -0.8%
13-15 158 +0.48 +15.5% +0.22 +8.0% +0.27 +6.6%

This one requires a bit of unpacking. In terms of a baseline, we see that, given the typical seven days of rest, home teams play about nine percentage points of DVOA better than usual, whereas away teams undershoot their average performance by about eight percentage points. No surprise there.

Otherwise, though, we see that the rest effect seems to depend on whether a team is at home or on the road. In comparison to baseline, home teams wildly exceed expectations with a bye week's worth of rest, but also do so with the shortest rest possible. In contrast, road teams woefully underperform on short rest and don't benefit nearly as much from the bye week. Furthermore, the difference in post-bye performance exists despite both home teams and away teams enjoying similar improvements in team health.

Regarding team units, the story is once again basically the same. Home teams coming off a bye play better on both sides of the ball, and getting healthier seems to be -- at least in part -- the mechanism for that improvement. That's not the case for away teams.

So, to open up the discussion, here's a question for readers: What the heck happens to away teams on long rest? They get healthier during the bye week, but they underperform similar to baseline. What gives?

UPDATE: Since this was designed to be a short XP, I left the following on the cutting room floor. However, the comments seem to be picking up on it, so I feel obligated to post. Here are the stats translated into the rest advantage ("Adv") for a team over their opponent rather than that team's absolute amount of rest.

The main result still holds. The only way a team can have a six-to-eight-day rest advantage is if they are coming off a bye, but their opponent is not. In such cases, home teams benefit from the largest rest advantage (via an AGL change advantage), whereas away teams don't.

Carry on.

AWAY
Rest Adv N TOT AGL Chg Adv TOT DVOA Diff
0 1,301 -0.05 -8.1%
+1 & +2 166 +0.15 -5.5%
+3 & +4 66 -0.14 +1.5%
+6 to +8 121 +0.42 -3.6%
HOME
Rest Adv N TOT AGL Chg Adv TOT DVOA Diff
0 1,301 +0.05 +9.7%
+1 & +2 114 +0.30 +9.7%
+3 & +4 38 +0.63 +9.6%
+6 to +8 139 +0.69 +14.2%

Posted by: Danny Tuccitto on 19 Nov 2012

24 comments, Last at 21 Nov 2012, 3:13pm by Mr Shush

Comments

1
by RickD :: Mon, 11/19/2012 - 3:03pm

" In comparison to baseline, home teams wildly exceed expectations with a bye week's worth of rest, but also do so with the shortest rest possible."

Well, that's because a team coming off a bye is far more likely to be playing a team not coming off a bye than a team playing with four or five days rest (i.e., playing on a Thursday) is going to be playing a team playing with more than four days rest. So the team playing at home with four days rest is facing a road team with four days rest.

And just a minor quibble about how you count "days rest". If you play on two consecutive Sundays, you don't have seven days rest.

2
by Danny Tuccitto :: Mon, 11/19/2012 - 3:07pm

Sympathetic to the minor quibble, but just tried to keep things simple: Seven is a nice number that people intuitively associate with a normal week's modicum of days.

3
by JSap :: Mon, 11/19/2012 - 3:23pm

This is why I was so ticked off at the Eagles schedule this year - they had/have 4 games against teams coming off their bye weeks, and 2 against teams with 10 days rest. As it turns out, they were awful this year anyway, but that kind of scheduling should not happen in the NFL.

7
by rrsquid :: Mon, 11/19/2012 - 3:57pm

And here I thought only Seattle was being screwed by this. It would be interesting to see a count of what teams had opponents on (8-)11 & (12-)14 day rests. I just don't have the time or motivation to count this.

10
by Danny Tuccitto :: Mon, 11/19/2012 - 4:58pm

Yeah, per the table I just added to the end of the piece, home teams coming off a bye and playing opponents not coming off a bye have a HUGE advantage. It's not in the table, but they won 63.9% of post-Week-4 games from 2002-2011, which is the only home-team group that exceeds the 56.8% baseline win rate for no rest advantage.

23
by Kiwiland Ian (not verified) :: Tue, 11/20/2012 - 6:18pm

SpokaneScott recently put up a fanpost on FieldGulls that looks at the teams who have more or less rest than their opponents this season. And yep, Philly comes off worst by far, at 25 days less rest than their opponents. (Seahawks are third at -10 days).

http://www.fieldgulls.com/2012/11/17/3660104/seahawks-schedule-analysis-...

4
by Glen (not verified) :: Mon, 11/19/2012 - 3:37pm

It would be interesting to see if distance traveled for away teams on long rest correlates to the steep decline of DVOA. For example, do away teams with long travel game have a long rest before that game?

This season I know 3 or 4 teams have come into Seattle on long rest and Seattle will be in Miami after the BYE. All long travels which have shown to have an effect by themselves.

5
by tuluse :: Mon, 11/19/2012 - 3:45pm

I think there is a lot of noise in this data.

Almost every team with 4 days of rest is facing an opponent who also had 4 days of rest. While this is not true for teams with bye weeks.

16
by JasonDrake (not verified) :: Mon, 11/19/2012 - 5:17pm

Exactly my thought. And in those cases, I have to think the home team has an exaggerated advantage despite days-of-rest being technically equal. In a compressed work week, the loss of preparation time owing to travel is proportionally much larger than in a normal work week.

6
by rageon :: Mon, 11/19/2012 - 3:55pm

Aside from bye weeks, I would assume the the other odd length of rest teams are not random. Teams playing on Monday night tend to be better than teams that do not, right? Not sure how that would affect things, but it's probably a valid consideration.

Something else -- Intuitively, if I was picking between 2 teams playing on very short rest, I would tend to go with the team that's been "playing together" longest (same coaching staff, same system, etc...), just on the assumption that they would have less time to prepare and would have to revert to prior knowledge.

I'm with the above comment that there may be too much noise here to draw a meaningful conclusion. Interesting though.

17
by RichardD (not verified) :: Mon, 11/19/2012 - 5:23pm

Teams playing on Thursday night seem to be worse on average than those playing on Sunday. I don't know how intentional it is, but NFL network seems to get the biggest dud game every week, to go along with their boring coverage. It's as if CBS and Fox get to choose which game will be played on Thursday. And think of all those Lions games on Thanksgiving.

8
by Kulko :: Mon, 11/19/2012 - 4:05pm

As long as both teams playing against have the same 4 days rest, they could both ber playing like Cleveland vs. Arizona, and it would not show up in your method.

DVOA is a zero sum game, meaning if you play HOU on a thursday night and they play like crap, your DVOA rises and offsets your own playing like crap. As long as the short week does not effect OFF and DEf differently, you will have problems picking it up.

OTOH long weeks are usually not shared by the opponents so the positive effect gets noticed.

The only thing your method is picking up is that Home team are less effected than away teams, which makes sense and provoided circumstancial evidence that such a effect really exists. But the rise in DVOA in this scenario actually proves my point, as its unlikely that teams play better on a short rest, the effect is just that the home team plays a bit worse, and the away team plays much worse and therefore DVOA rises.

As for what happens to the away teram on a long rest, they play better than expected. Being the away team docks 8% from your DVOA and they only loose 5,5% Its just that the away effect is much more important then the long week effect.

11
by jonnyblazin :: Mon, 11/19/2012 - 4:59pm

DVOA isn't a zero sum game though.

24
by Mr Shush :: Wed, 11/21/2012 - 3:13pm

Indeed. Even VOA isn't a zero-sum game (missed field goals are bad but opponents' missed field goals aren't good, for example).

9
by Danny Tuccitto :: Mon, 11/19/2012 - 4:54pm

Hey everyone, just letting you know I updated the end of the piece to reflect what you guys are picking up on in the comments.

12
by Tballgame (not verified) :: Mon, 11/19/2012 - 5:01pm

Danny,

I think the discrepancy is nothing more than small sample size. If you grouped 9-11 days and 13-15 days, the discrepancy is less prevalent.

I do find it interesting that at every game interval, teams with home games are healthier than teams with road games. It also seems that, following a Thursday game, teams are more likely to be sent on the road than head home (hence the home/road split of 45/77 for 9-11 day intervals), but teams are more likely to play at home following a Monday game (195/144 home/road 6 day interval). Is that known to be a biasing effort on the part of the NFL scheduler to try to balance inequities?

14
by Danny Tuccitto :: Mon, 11/19/2012 - 5:08pm

Actually, if you group 9-11 and 13-15, the TOT DVOA discrepancy still holds (+13.9% to -3.9%), but TOT AGL change goes from being nearly equal to home teams getting healthier on long rest than road teams (+0.42 to +0.21). So there's that.

Also, I don't know if the league office is setting the schedule up that way on purpose, but I agree it sure seems like it.

13
by tballgame (not verified) :: Mon, 11/19/2012 - 5:07pm

Home and road teams with a 6+ games of rest more than their opponent benefit by 4.5% over having the same rest as their opponent (14.2-9.7=4.5 and -3.6 minus -8.1=4.5).

Road teams seem to benefit from having 1-4 days of extra rest whereas home teams do not demonstrate added benefit beyond the standard home team benefit.

15
by Danny Tuccitto :: Mon, 11/19/2012 - 5:14pm

Yeah, you're right, +4.5% improvement compared to baseline. Not sure what I was smoking there. I think I meant to say that away teams on long rest still don't actually play better than their DVOA going into the game.

21
by MyrEn (not verified) :: Tue, 11/20/2012 - 2:12am

Was looking for this comment (and the associated author reply).

A great conclusion to this article would be that:

Home field advantage is approximately twice as strong as bye week advantage (9% vs 4.5%).

18
by Intropy :: Mon, 11/19/2012 - 8:21pm

Has any team ever played Thursday three days after playing MNF?

19
by Danny Tuccitto :: Mon, 11/19/2012 - 8:45pm

Nope. There haven't been any games on 3 days rest over the past 10 years, and all of the games on 4 days rest have been of the Sunday-then-Thursday variety.

20
by usernaim250 :: Tue, 11/20/2012 - 1:45am

I've always wondered why they don't try to even out the rest with the Thursday games. It would seem the ideal would be MNF/bye/Thursday for the home team and Sunday/bye/Thursday for the visitor. And shouldn't every team that plays MNF have a home game the following week (unless bye, of course)?

Of course I'm just assuming that short rest correlates to higher rate of injury.

22
by Intropy :: Tue, 11/20/2012 - 10:21am

Scheduling the games is a hard problem (NP-hard even). I wouldn't be surprised if something like that is already considered but not with much weight.