SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.
Also, our existing PFR blog rss feed will be redirected to the new site's feed.
Pro-Football-Reference.com ยป Sports Reference
For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.
ShareThisNew York Times post: Week 1
Once again, Pro-Football-Reference.com will be teaming up with the New York Times and the New York Times' Fifth Down Blog. Every Tuesday on the Fifth Down blog (link available here) and every Wednesday in print (page B14 of your newspaper today) we'll be running a weekly update (for you baseball fans out there, Sean has been doing essentially the same thing through Baseball-Reference.com this season).
Today's article references one of the discussions Jason, Doug and I got into on a recent podcast: when do good teams lose games? As it turns out, "week 1" is a pretty common answer.
This entry was posted on Wednesday, September 15th, 2010 at 9:23 am and is filed under Announcements. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
Week 1 of 2010 had low margins of victory, as expected, but it was also unusually low-scoring: half of the winning teams scored fewer than 20 points, the first time that's happened since Week 5 of 1999. Is it typical for Week One to be low-scoring, as well as close?
So, we're just ignoring the fact that Week 10 has the exact same win % as week 1 for the sake of the narrative? I expect that form most comentators, but not from you guys.
Allow me to make you lose even more faith in us, Alvaro.
I think week 1 is a legitimate outlier. I'm not surprised to see the best teams (relatively) struggle in week 1. Meanwhile, week 10 is almost certainly just noise. There's nothing special about week 10 relative to weeks 6, 7, 8, 9, 11, 12, 13 or 14. But if you look at all of those weeks together, it's not surprise that over a sample of just 82 teams, one week will be lower than the rest. In week 10, the teams went 62-20; the teams went 71-11 and 68-14 in the weeks before and after. we're only talking about six or seven extra losses over 82 games. If week 6 or 9 or 10 or 13 was the week with a lot of losses, you'd likely have the same reaction as if they occurred in week 10. It's not surprising to see one week in the middle of the season be a little low; on the other side, I think there's something legitimately there in the beginning of the season.
But maybe I'm way off base here. What do you think?
I think that it makes a lot of sense to think that week 1 would be a week more likely for teams who went on to finish strong to lose due to the team "shaking the rust off."
I do however also think that there is absolutely nothing in the data to support this. I see no reason to believe (other than the pre-concieved notions that we bring to it) that week 10 is noise, bit week 1 isn't.
Of course it makes sense that one week would be lower than the rest, but the fact that it is exactly as low as week 10 makes it seem just as likely that week 1 is noise as well.
Week 16 and 17 not only make sense due to us knowing that many of those games were without starters playing a significant ammount, but the difference between them and the rest of the season is significantly more than that of weeks 1 and 10.
Maybe it is just a coincidence that random noise produces a result that is equal to one produced by an actual effect, however we can't see that from this data. Hopefull as the data set grows we'll be able to either confir or deny that week 1 is just noise as well, but with what we have I don't see the results as anything but inconclusive.
Chase, can you expand your data set back to 1970, to see if that settles down any of the "noise"?
There were 63 teams from 1970 to 1989 (excluding '82 and '87) to win 75% of their games; on average they won 80% of their games, winning 12, losing 2.9 and tying 0.1 times.
79.8% Gm 1
83.1% Gm 2
75.0% Gm 3
88.7% Gm 4
77.4% Gm 5
85.5% Gm 6
78.2% Gm 7
79.0% Gm 8
79.8% Gm 9
80.6% Gm 10
75.8% Gm 11
85.5% Gm 12
79.0% Gm 13
80.6% Gm 14
74.2% Gm 15
77.4% Gm 16
Interesting... while this does indeed plant the week 1 data firmly in the "noise" category, it also does the same for weels 15 and 16. And those had a pretty wide margin before. Is this due to era? Did teams rest their starters just as much? Was the era before free-agency responsible for the benches of power-houses still being good enough to compete with other teams? Was the diference in play-off spot responsible? ANyone know the answers to these questions?
I don't think Week 1 is much more significant than everyone else seems to believe.
In 2009, 13 of the 16 Week One winners had a better record than the team that lost. Of the rest, one had the same record (Jets), one lost on a ridiculous fluke (Bengals), and one finished only two games worse.
In 2008, 12 of the 16 had better records. In 2007, 12 of 16, with two being equal. So if you lose early, it's probably a big deal.
Just poking around...
Week 1 winners in 2009 went 9.00-6.00 in the next 15 games.
Week 1 winners in 2008 went 8.78-6.22 in the next 15 games.
Week 1 winners in 2007 went 8.63-6.38 in the next 15 games.
Obviously, the averages for week 1 losers are reversed.
About half of the week 1 winners go 10-6.
About 15% of the week 1 losers go 10-6.