Before football season started last year, I came up with an "unders" system. Basically, I discovered a way to detect certain situations where scoring is unlikely to occur, which means one should bet on the under for the game total. I backtested it using all of the data I had available (I only have reliable NFL data going back to 1993, and only for the regular season).
+------+---------+
| year | record |
+------+---------+
| 1993 | 10-5-0 |
| 1994 | 17-10-0 |
| 1995 | 10-6-0 |
| 1996 | 12-8-0 |
| 1997 | 12-7-0 |
| 1998 | 11-18-1 |
| 1999 | 17-9-0 |
| 2000 | 22-14-0 |
| 2001 | 14-17-0 |
| 2002 | 29-16-0 |
| 2003 | 25-17-0 |
| 2004 | 20-14-0 |
| 2005 | 22-11-0 |
| 2006 | 23-9-0 |
| 2007 | 23-13-0 |
+------+---------+
In the 2008 season I forward-tested it and ended up 2-1-0 in the preseason, 22-18-2 in the regular season, and 1-1-0 in the post season (combined 25-20-2). So, the system has an all-time record of 292-194-3 (just over 60%) going back to 1993.
I should mention that although the system can find low-scoring games, it can't be "reversed" to reliably find high-scoring games... that is, it never picks the "over". So I found it to be rather boring and frustrating to actually sit down and watch the games, as I was sitting there basically hoping that nothing would happen. It's not much fun rooting for the clock.
Anyway, in this 16-year timeframe, a flat bettor risking $110 to win $100 on each game would be up $7,860, although this would clearly be sub-optimal. The best betting strategy, of course, would be to use the Kelly criterion. If we assume a 60% chance of winning each wager, then we should risk 16% of our bankroll per bet. So, a $1,000 starting bankroll in 1993 would have turned into about $340,186 by Super Bowl 2009 (although, again, I only ran this system on the regular season games for 1993-2007 since that's all the data I have).
Even with such a strong record, the system has had 2 losing years (1998 and 2001), so it's not exactly perfect. I'm not planning to offer the system or its picks for sale until I can do a couple more years of forward-testing, but if anybody's interested I can post the picks on the forum once the season begins.