Note: For those that participated in the NFL Opener Contest, scroll to the bottom for the results.
I don't think any of us would think the Patriots or Ravens or Bears were in for a down year if they lost their first games. Likewise, none of us will be guaranteeing a post-season berth out of the Raiders/Lions winner. But winning in week one means something, right? And how much does winning in weeks 1 and 2 mean? If the Ravens (13-3 last year) are 3-5 after eight weeks while the Raiders (2-14) are 5-3 after eight weeks, which team would we suspect will do better in the second half of the season?
Certainly, no one rule can be applied to every team. If the Lions get really lucky in winning their first three games, we might not be that high on them. If Tom Brady gets hurt, we're going to downgrade the Patriots a lot more than if they just look bad. So while this may be a pointless exercise -- it's Friday, so why not.
First we need to come up with a variable for pre-season predictions. Remember the Simpe Rating System? I used each team's SRS score from the previous year as their projection. Obviously this doesn't take into account any off-season moves, so our faith in the predictions shouldn't be too high. While human projections aren't great, they're usually a bit more accurate than merely using last year's data. Using the SRS tends to be a bit more predictive than wins and losses, so that's why I chose to use that.
The 2006 Chargers and 2006 Patriots both had SRS scores of +10.2, highest of any team last year. I performed a regression analysis of teams from the past decade to convert a SRS score into a projected win total. The resulting formula to project next season's wins was 8 +/- 0.45 wins for every SRS point. So the Chargers and Patriots will both be projected to win 12.59 games this year, while the Raiders will be projected at 3.68 wins. That more or less feels right to me. Here's how many wins every team is projected for in 2007:
team SRS projWins
nwe 10.2 12.6
sdg 10.2 12.6
bal 9.3 12.2
chi 7.9 11.6
jax 7.5 11.4
ind 5.9 10.7
cin 4.1 9.8
nor 4.0 9.8
dal 3.7 9.7
pit 3.4 9.5
phi 3.4 9.5
buf 2.2 9.0
nyj 2.0 8.9
den 1.3 8.6
kan 1.0 8.5
mia 0.7 8.3
nyg 0.1 8.0
ten -1.3 7.4
car -2.7 6.8
atl -3.0 6.7
sea -3.6 6.4
stl -4.0 6.2
was -4.0 6.2
min -4.1 6.2
gnb -4.4 6.0
hou -4.5 6.0
cle -5.8 5.4
det -6.4 5.1
ari -6.9 4.9
tam -7.9 4.4
sfo -8.7 4.1
oak -9.6 3.7
I then looked at the ten seasons from 1997-2006, focusing on three variables: each team's projected number of wins (based on their SRS score from the prior year), each team's winning percentage through game X of the season, and each team's winning percentage for the remainder of the season. It's important to use the remainder and not the final season winning percentage to avoid biasing the results.
What's the verdict? How much does each week matter? The table below shows how much weight should be placed on both last season's results (N-1 weight) and the current season's results to date (N-T-D Weight) in figuring out winning percentage for the remainder of the year:
After Game N-1 Weight N-T-D Weight
1 6.06 1.00
2 2.57 1.00
3 1.44 1.00
4 1.01 1.00
5 0.70 1.00
6 0.64 1.00
Does this information jive with what you would have guessed? Remember, if you have really strong predictions (and I think most of us could come up with better predictions than the Year N-1 SRS grade), the weight placed on them should increase. If that's the case, your pre-season projections might be more important than even five or six games worth of data.
Here are a couple of examples.
If the Chargers start off 2-4, they'll be projected to have a 0.510 winning percentage the rest of the season, which means about five wins, and a 7-9 season. Does that seem right to you? If the Raiders go 4-2 through six games, Oakland will have a projected 0.496 winning percentage for the remainder of the year, which would lead to a 9-7 record.
If you asked me who would win more games the rest of the year -- the 2-4 Chargers or the 4-2 Raiders -- I'm not sure which team I'd like. On one hand, that tells me this system is pretty good, because the results are reasonable. On the other, I'd want to ask how in the world did the Chargers go 2-4? Is Tomlinson hurt? Is Gates hurt? If so, that would be a lot more valuable than any formula I could derive. But at least this gives me something relatively concrete to rely on, and reinforces the idea that what happens in week one shouldn't change your thoughts very much. It's still six times less valuable than what went on in 2006.
And for Saints fans, that's the only good piece of news they'll hear all day.
NFL Opener Contest Results
Congrats to Dr. I Don't Know, who was the closest among our 31 contestants in predicting the final score of last night's game. Please send an e-mail to email@example.com to claim your prize.
Total Correct winner
Dr. I Don’t Know Indianapolis 30 New Orleans 16 1217 YES
Yuval Indianapolis 27 New Orleans 13 1313 YES
Vince Indianapolis 38 New Orleans 24 1313 YES
Jason W Indianapolis 31 New Orleans 20 1789 YES
Jacob Indianapolis 24 New Orleans 13 1985 YES
Doug W Indianapolis 31 New Orleans 21 2022 YES
JWM Indianapolis 26 New Orleans 17 2334 YES
Doug Indianapolis 23 New Orleans 14 2466 YES
Hedgehog Indianapolis 30 New Orleans 23 2826 YES
Nick Indianapolis 38 New Orleans 30 2826 YES
MattyP Indianapolis 24 New Orleans 17 2922 YES
Dave Indianapolis 38 New Orleans 31 3146 YES
Mr Shush Indianapolis 34 New Orleans 28 3246 YES
Autumn Indianapolis 41 New Orleans 34 3398 YES
Luke Indianapolis 20 New Orleans 14 3414 YES
sepporepi Indianapolis 31 New Orleans 27 3837 YES
Joseph Indianapolis 35 New Orleans 31 4013 YES
Otis Indianapolis 27 New Orleans 24 4149 YES
Richie Indianapolis 21.5 New Orleans 18.5 4270 YES
Joshua P. Indianapolis 20 New Orleans 17 4345 YES
Scott Indianapolis 34 New Orleans 31 4345 YES
Arkuss Indianapolis 30 New Orleans 28 4578 YES
cjm Indianapolis 38 New Orleans 35 4633 YES
Sky New Orleans 38 Indianapolis 17 332 NO
monkeytime New Orleans 17 Indianapolis 7 2750 NO
Dan New Orleans 27 Indianapolis 21 3134 NO
JKL New Orleans 23 Indianapolis 19 3869 NO
Mats New Orleans 35 Indianapolis 31 4013 NO
Dan Miller New Orleans 31 Indianapolis 28 4213 NO
Don New Orleans 28 Indianapolis 27 4946 NO
MadMolecule New Orleans 31 Indianapolis 30 5030 NO
Final Score Indianapolis 41 New Orleans 10
This entry was posted on Friday, September 7th, 2007 at 1:10 am and is filed under General, History, Statgeekery. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.