Writers of Pro Football Prospectus 2008

02 Oct 2006

FO on BSMW: Bye Weeks

The first of two BSMW pieces this week take a look at the issue of early/late bye weeks affecting overall team performance, as brought up by the Dallas Morning News's Matt Mosley.

Posted by: Bill Barnwell on 02 Oct 2006

18 comments, Last at 03 Oct 2006, 11:08am by Mitch Cumstein

Comments

1
by BK (not verified) :: Mon, 10/02/2006 - 11:57am

I don't know if you can make this kind of statistical correlation since the schedule makers determine the bye weeks (i.e. bye weeks are not random). The NFL schedulemakers (when the league was 31 teams) used to give late bye weeks to poor teams so that the good teams would be playing late in the year to pique fan interest.

2
by Vern (not verified) :: Mon, 10/02/2006 - 12:04pm

I think you should cross compare to your "expected wins." For example, look for correlation between teams that outperformed their expected win totals and when their bye week occured.

I doubt you'll find much, and I'm not sure how far back your DVOA projections go, but if it's already there, it might be worth a try.

3
by ToxikFetus (not verified) :: Mon, 10/02/2006 - 12:14pm

Why did the 1993 season have two bye weeks for each team? Was there some event that caused one team to have two byes, and therefore every other team was given a bye? Or was it a one-year trial run?

4
by dryheat (not verified) :: Mon, 10/02/2006 - 12:26pm

I don't care what Simmons says. I'm certain that running QBs age at the same rate as the rest of us.

I believe 1993 was a strange year altogether. If I'm not mistaken, that's the year the Patriots played two games against a non-division opponent. I think either Cleveland or Denver.

5
by B (not verified) :: Mon, 10/02/2006 - 12:28pm

1: Except, of course, the schedule makers arn't that good at figuring out which teams won't be good. That's how the Patriots ended up with a bye on week 16 (Would have been week 17 but for the 9/11 stoppage of play) before thier 2001 superbowl run.

6
by dryheat (not verified) :: Mon, 10/02/2006 - 12:37pm

To amend #4, it was 1991, and it was against the Broncos. The Patriots lost on Oct. 27 and again on Dec. 1. I still haven't heard an explanation of how that happened.

7
by Spartacus (not verified) :: Mon, 10/02/2006 - 1:03pm

From a physics standpoint, the closer you travel to the speed of light, the more time slows down for you (time dialation).

Thus a running quarterback would (albeit almost imperceptibly) age slower than a stationary quarterback.

8
by andrew (not verified) :: Mon, 10/02/2006 - 1:03pm

damn forgot to change name back (ex-Spartacus)

9
by Scott de B. (not verified) :: Mon, 10/02/2006 - 1:16pm

Why did the 1993 season have two bye weeks for each team? Was there some event that caused one team to have two byes, and therefore every other team was given a bye? Or was it a one-year trial run?

It was a trial run. Pretty much everyone agreed it made the season too long.

10
by dryheat (not verified) :: Mon, 10/02/2006 - 1:21pm

almost imperceptibly?

11
by empty13 (not verified) :: Mon, 10/02/2006 - 1:49pm

9. if i rememebr right, think they did that for 2 years

12
by Dean (not verified) :: Mon, 10/02/2006 - 2:33pm

RE#6

In the 28 team era, you played your division opponents twice, and you played conference opponents based on where you finished in the standings. There were 3 divisions/conference, one of which (NFC West, AFC Central) only had 4 teams. The two fifth place teams in each conference would play each other twice the next season.

13
by ammek (not verified) :: Mon, 10/02/2006 - 3:21pm

Might it be worth adding the number of wins teams had in both the previous and the following seasons before proclaiming the benefit of the early bye. After all, this year the Steelers, Bucs, Broncos and Giants were off in week 4: that's three division winners and a Superbowl champion. If this lot averages 8.4 wins this season, it won't only be a result of the early bye.

14
by Richie (not verified) :: Mon, 10/02/2006 - 3:42pm

9. if i rememebr right, think they did that for 2 years

Nope. Just one year.

15
by RobinFiveWords (not verified) :: Mon, 10/02/2006 - 3:44pm

Using a linear trendline in Excel, the Week 3 through Week 10 data suggest that each week is worth about 0.1 wins (the slope of the trendline is -0.1024).

16
by kibbles (not verified) :: Mon, 10/02/2006 - 5:23pm

Re #4: I believe 1993 was a strange year altogether. If I’m not mistaken, that’s the year the Patriots played two games against a non-division opponent. I think either Cleveland or Denver.

In 1993, it was the Seahawks.

17
by Pat (not verified) :: Mon, 10/02/2006 - 8:28pm

#15: It's a little weaker than that - about -0.08 (the points aren't equally statistically valid - adding quantization error weakens it a bit, although that's not quite the right way to do it). Note that the points on the edge basically drive the trend, and statistically, they've got the weakest significance.

It's basically statistically insignificant then, too: only a p-value of about 0.2 or so.

18
by Mitch Cumstein (not verified) :: Tue, 10/03/2006 - 11:08am

Rather than total wins, you might want to look at how teams performed in the last month (4 games) of the season, relative to their won/loss record in the first 12, depending on whether they had an early bye or late bye.

For example, last year's week 10 byes were Cincinnati, San Diego, New Orleans, and Tennessee. I don't think any of them finished particularly strong down the stretch, relative to how they played early. Perhaps this is why there are no longer week 10 byes this year?