This is the outcome of blindly betting this in the NFL since 1990 based on the repole numbers. The criterion was where spread >= (1/4)(total). This only counts weeks 1-9 of 2007 and the regular season only for all other years. The 3 numbers are wins-losses-halfwins. A half win is where you win half of the parlay and tie the other half.
<font class="small">Code:</font><hr /><pre>Season Over/Fav Under/Dog
2007 6-7-0 2-10-1
2006 10-27-1 9-28-1
2005 5-19-0 8-16-0
2004 3-13-1 4-13-0
2003 4-15-0 6-13-0
2002 6-16-0 4-18-0
2001 5-16-0 2-17-2
2000 8-20-1 6-22-1
1999 6-20-0 8-17-1
1998 8-18-0 4-22-0
1997 3-17-1 4-16-1
1996 6-16-1 4-19-0
1995 4-21-0 9-16-0
1994 9-15-0 9-15-0
1993 12-31-0 15-28-0
1992 8-44-1 15-38-0
1991 8-23-0 7-24-0
1990 3-20-1 7-17-0
TOTAL 114-358-7 123-349-7
</pre><hr />
Here are the games this year that qualified (weeks 1-9).
2007 Weeks 1-9:
Week 2: Buf(3) +10 @ Pit(26) o/u39 [loss]
Week 2: NJY(13) +10 @ Bal(20) o/u 33 [half]
Week 2: Oak(20) +10 @ Den(23) o/u 37 [win]
Week 2: KC(10) +12 @ Chi(20) o/u 34.5 [win]
Week 3: SF(16) +9.5 @ Pit(37) o/u 38 [win]
Week 3: Buf(7) +16.5 @ NE(38) o/u 41 [win]
Week 4: Stl(7) +13 @ Dal(35) o/u 47 [loss]
Week 4: KC(30) +12 @ Sdg(16) o/u 38.5 [loss]
Week 5: Cle(17) +15.5 @ NE(34) o/u 48 [win]
Week 6: Stl(3) +9 @ Bal(22) o/u 36.5 [loss]
Week 7: NE(49) -15.5 @ Mia(28) o/u 49.5 [win]
Week 8: Was(7) +14.5 @ NE(52) o/u 46.5 [win]
Week 9: Bal(7) +9.5 @ Pit(38) o/u 36 [win]
With standard payouts of +2.6 for a win, -1 for a loss and +0.91 for a half win, it looks like blindly doing this for the NFL is -EV. This year with the New England anomaly it appears to be doing well.
Edit: I was improperly counting half wins where the spread was tied. If under/dog tied, I was not counting this as a loss for over/win.