I took the preseason data going back to 2001 (the last year NFL.com has quarter-by-quarter scoring data for the preseason), and ran the four preseason ranking algorithms (listed below) on the scores for each season.
Using each power ranking as a means to pick the week one games (i.e. predict the team with the highest power ranking to be the winner), and then compared the pick to the actual game winners. Below is a table summarizing the results for the data:
| Year | Preseason Week 3 | Pre/half Week 3 | Preseason Week 4 | Pre/half Week 4 |
| 2001 | 4-11 | 6-9 | 6-9 | 6-9 |
| 2002 | 7-9 | 7-9 | 9-7 | 8-8 |
| 2003 | 9-7 | 12-4 | 9-7 | 7-9 |
| 2004 | 6-10 | 12-4 | 12-4 | 12-4 |
| 2005 | 6-10 | 6-10 | 7-9 | 7-9 |
| 2006 | 10-6 | 10-6 | 7-9 | 10-6 |
| 2007 | 10-6 | 8-8 | 10-6 | 10-6 |
| Total | 52-59 | 61-50 | 60-51 | 60-51 |
I was happier with these numbers before I ran the 2001 data set. Any way you look at it, it's a pretty sparse data set to be making strong predictions. But it does look like there's reason to think preseason power rankings are useful data.
I'll be doing a more extensive analysis later. And, of course, pick results for the 2008 data will be posted next week.
