The Importance of Week One Results
- Thread starter bowiac
- Start date
With all the salacious happenings in the NFL today, I don't want this to get lost.
This is one of the finest things I've ever read on SoSH, and that includes some of the best baseball content we've featured over the years. Everyone, please take a few minutes and read - and comment - on this article written by a fellow SoSHer.
This is one of the finest things I've ever read on SoSH, and that includes some of the best baseball content we've featured over the years. Everyone, please take a few minutes and read - and comment - on this article written by a fellow SoSHer.
Great article.
I was eager to exploit this knowledge for gambling purposes, but it appears the market may be efficient here. The four teams to bet against based on bowiac's insights seem to be the Rams, Bucs, Chiefs and Patriots. The Rams and Bucs play each other. The other two lines have already moved 2-3 points compared to early lines that were published before yesterday's games. In the case of Vikings-Pats, that movement could be attributed to the Vikings' unexpectedly strong performance yesterday, but in the case of Broncos-Chiefs the movement has to be ascribed to KC's struggles, as Denver's performance almost perfectly matched pre-game expectations.
I was eager to exploit this knowledge for gambling purposes, but it appears the market may be efficient here. The four teams to bet against based on bowiac's insights seem to be the Rams, Bucs, Chiefs and Patriots. The Rams and Bucs play each other. The other two lines have already moved 2-3 points compared to early lines that were published before yesterday's games. In the case of Vikings-Pats, that movement could be attributed to the Vikings' unexpectedly strong performance yesterday, but in the case of Broncos-Chiefs the movement has to be ascribed to KC's struggles, as Denver's performance almost perfectly matched pre-game expectations.
I don't know if I'd call them inefficiencies - this is based on information not yet available (Week 1 results), so it's not really exploitable in the way we'd like an inefficiency to be. As maufman points out, Vegas is moving the line on the teams that this analysis flagged. If I have time before this weekend, I'll try and look at how these teams did in Week 2 though.SMU_Sox said:
Interesting topic. I have a lot of this stuff easily accessible, so thought I'd run some numbers.
Wins are the total for the season, including playoffs. So I don't use winning percentages. To keep the numbers consistent, I used all the seasons from 1978-2013, eliminating the short seasons of 1982 and 1987. There were two fewer playoff games per season before 1990. That change consistently affects all weeks and is tiny. The differences between the league size of 28 (through 1994) and larger afterward, related to the number of playoff teams is similarly tiny. I didn't separate week 18 because there was only one year that had a week 18.
The average number of wins per team is 8.51 per season (this is far from the center of the week averages shown below because playoff weeks only contain teams with a lot of wins).
Week / Average total wins for team that won / Average total wins for team that lost
1, 9.64 - 7.05
2, 9.56 - 7.07
3, 9.74 - 6.89
4, 9.57 - 7.00
5, 9.65 - 7.10
6, 9.65 - 7.09
7, 9.58 - 6.94
8, 9.51 - 7.12
9, 9.68 - 7.08
10, 9.68 - 7.00
11, 9.67 - 7.01
12, 9.72 - 6.97
13, 9.58 - 7.08
14, 9.65 - 7.06
15, 9.50 - 7.19
16, 9.56 - 7.09
17, 9.61 - 7.13
Total, weeks 1-18, 9.62 - 7.05
So, week three stands out a tiny bit, maybe. Seems well within random fluctuation. Week one is pretty much exactly the average. And maybe a bit of a slope in weeks 15-17 as the top teams can start to prepare for the playoffs.
I don't see anything to indicate that a week one win or loss is any more important than any other week. Week one performance seems very representative of overall team quality.
Wins are the total for the season, including playoffs. So I don't use winning percentages. To keep the numbers consistent, I used all the seasons from 1978-2013, eliminating the short seasons of 1982 and 1987. There were two fewer playoff games per season before 1990. That change consistently affects all weeks and is tiny. The differences between the league size of 28 (through 1994) and larger afterward, related to the number of playoff teams is similarly tiny. I didn't separate week 18 because there was only one year that had a week 18.
The average number of wins per team is 8.51 per season (this is far from the center of the week averages shown below because playoff weeks only contain teams with a lot of wins).
Week / Average total wins for team that won / Average total wins for team that lost
1, 9.64 - 7.05
2, 9.56 - 7.07
3, 9.74 - 6.89
4, 9.57 - 7.00
5, 9.65 - 7.10
6, 9.65 - 7.09
7, 9.58 - 6.94
8, 9.51 - 7.12
9, 9.68 - 7.08
10, 9.68 - 7.00
11, 9.67 - 7.01
12, 9.72 - 6.97
13, 9.58 - 7.08
14, 9.65 - 7.06
15, 9.50 - 7.19
16, 9.56 - 7.09
17, 9.61 - 7.13
Total, weeks 1-18, 9.62 - 7.05
So, week three stands out a tiny bit, maybe. Seems well within random fluctuation. Week one is pretty much exactly the average. And maybe a bit of a slope in weeks 15-17 as the top teams can start to prepare for the playoffs.
I don't see anything to indicate that a week one win or loss is any more important than any other week. Week one performance seems very representative of overall team quality.
Based on some stuff I've looked at before, and based on what you show here, I agree (although it's interesting even Week 17 looks the same as ever).Chemistry Schmemistry said:
I wasn't trying to suggest there's anything more important about Week 1 than any other week - I was mostly trying to see how much we should adjust season expectations based on Week 1 results.
Yes, I was just extending the thought a little. It's hard not to get into tiny sample size issues when trying to factor in expectations. Maybe the best way to do that is simply use last season's win totals. So, looking at that piece, I have around 900 data points for week one. Of those, 58 had 13 wins (like New England, including playoffs), in the preceding season.
The average win total for the following season was only 9.64. That's quite a drop. I think we can agree that what New England has accomplished in the Brady/Belichick era is remarkable.
Those 58 teams went 29-29 in week 1 the following year. The 29 that won averaged 10.79 wins, the 29 that lost averaged 8.48 wins.
That's really interesting. Maybe there's something there. Not in the final results, wins versus losses. That's consistent with the above chart. But the 29-29 record itself is quite interesting, as these teams wind up being a couple of wins better than average during the season.
So, a deeper look shows that only 23 of the 58 were at home in week one. That really is the problem when you start isolating things - you get unexpected biases here and there.
Going further, let's see what happens with 12 wins the previous season...
Record: 29-14 (25 of 43 at home). 10.97 wins for winning teams in week one, 7.79 for losing teams.
And 14 wins...
Record: 15-11 (11 of 26 at home). 13.50 wins for winning teams, 8.45 for losing teams.
And 11 wins...
Record: 47-27 (38 of 74 at home). 10.51 wins for winning teams, 8.74 for losing teams.
I don't have any really solid conclusions. The average team that won 13 games, including playoffs, wins close to 10 games the following season. So, at this point, with very small sample sizes to base this on, you could set an expected win total of 9 on the Patriots right now based on their loss. But anyone with even a tiny bit of understanding of sample size and how statistics work wouldn't call that a confident prediction. After all, Miami is probably a decent team this year, and this was a road game in 90-degree heat and a heat index of 99.
The average win total for the following season was only 9.64. That's quite a drop. I think we can agree that what New England has accomplished in the Brady/Belichick era is remarkable.
Those 58 teams went 29-29 in week 1 the following year. The 29 that won averaged 10.79 wins, the 29 that lost averaged 8.48 wins.
That's really interesting. Maybe there's something there. Not in the final results, wins versus losses. That's consistent with the above chart. But the 29-29 record itself is quite interesting, as these teams wind up being a couple of wins better than average during the season.
So, a deeper look shows that only 23 of the 58 were at home in week one. That really is the problem when you start isolating things - you get unexpected biases here and there.
Going further, let's see what happens with 12 wins the previous season...
Record: 29-14 (25 of 43 at home). 10.97 wins for winning teams in week one, 7.79 for losing teams.
And 14 wins...
Record: 15-11 (11 of 26 at home). 13.50 wins for winning teams, 8.45 for losing teams.
And 11 wins...
Record: 47-27 (38 of 74 at home). 10.51 wins for winning teams, 8.74 for losing teams.
I don't have any really solid conclusions. The average team that won 13 games, including playoffs, wins close to 10 games the following season. So, at this point, with very small sample sizes to base this on, you could set an expected win total of 9 on the Patriots right now based on their loss. But anyone with even a tiny bit of understanding of sample size and how statistics work wouldn't call that a confident prediction. After all, Miami is probably a decent team this year, and this was a road game in 90-degree heat and a heat index of 99.
I took a quick look at your heat idea, and there does seem to be something to it. Favorites from 1978-2012 (haven't added 2013 yet) win at around a 63-64% clip overall. For games payed at least 85 degrees (I don't have humidity info yet for heat index stuff, but am building out that database as well), that number drops to around 58%. This is quick and dirty - I'm not doing a comparison of spreads here, and I'm looking at all games, to avoid cutting my sample size too much (games that hot are rare regardless).Chemistry Schmemistry said:
Push comes to shove, I do think this loss should lower expectations a bit, although as you note, the Belichick-Patriots have defied every other statistical rule, including regression to the mean (as shown in your analysis of 13 win teams above), so it's hard to go to far with this. I'm a bit worried however.
I'd love to see whether games with temperature extremes show more of a homefield advantage since the NFL went to the more limited practice schedule under the new CBA. Belichick had the team practicing in sweats last week to simulate the hot temperatures, but there's a limit to how much of that sort of preparation they can do now. There's probably not enough of a sample of such games to draw a conclusion, but it bears watching.bowiac said:
The struggles of the Vikings and the Titans so far today support bowiac's hypothesis that unexpectedly bad Week 1 performances tell us more than unexpectedly good ones.
Edit: And the Dolphins.
Edit: And the Dolphins.