|
Q: Are there any stats from the previous season that have value in predicting who will cover against the spread in week one this season?
| Statistic |
Won |
Lost |
Win % |
Comments |
| Winning % |
55 |
64 |
46% |
23-37 in games w/5+ point spread |
| Spread Record |
61 |
57 |
52% |
37-24 in games with small lines (<5 pts) |
| Points Scored |
50 |
69 |
42% |
20-34 w/5+ point favorites |
| Points Allowed |
56 |
61 |
48% |
(playing the team that allows fewer points) |
| Net Points |
50 |
73 |
41% |
7-15 with underdogs |
| Total Points |
52 |
68 |
43% |
(playing the higher "total" team) |
Net Points,
compared to line |
51 |
72 |
41% |
only big favorite (10+) games positive,
23-30 w/dogs, 28-42 w/faves |
|
All records against the spread covering 1992-2000 NFL week one games. All statistics encompass a team's entire previous season (regular season and any playoff games)
|
ANALYSIS:
Obviously for the most part, playing teams based on a better previous season stat is a bad idea. However the sharp observer notices that the results were so bad, there is perhaps value in going contrarian and betting the team with the worse previous season numbers (after all 41% can be turned into 59% by reversing the bet). Indeed the results based around the "Net Points" (points scored minus points allowed) statistics do make for an interesting possibility -- the lines may be getting set based too much on reputation, and do not allow for enough movement for impovement of lesser teams and decline of stronger teams.
If we were trying to create a method for playing the week one "bettor's madness" games using stats, the above would make us inclined to focus on net points, but rather than simply playing the team with the worse stat, we'd suggest using the net points to make a line with some additional home field advantage factored in. Doing this produces results which look very similar at face value to just playing the worse "net points" team period. However, by requiring a team to have both the worse overall stat AND also to be going contrarian to the "value based play" you would see a nice 55-33 record over the past ten seasons.
The next question though becomes whether there has been some shift over time. Here are the year by year breakouts for playing against the better "net points" team and then only playing against teams which also had perceived value against the spread based on making a "net points" line:
| Year |
Play against
better
"Net Pts" team |
Win% |
Play against
better net and
perceived value |
Win% |
| 1992 |
6-6 |
50% |
4-6 |
40% |
| 1993 |
4-8 |
33% |
3-3 |
50% |
| 1994 |
6-8 |
43% |
5-6 |
45% |
| 1995 |
10-4 |
71% |
9-3 |
75% |
| 1996 |
10-5 |
67% |
7-4 |
64% |
| 1997 |
10-3 |
77% |
7-3 |
70% |
| 1998 |
6-7 |
46% |
4-3 |
57% |
| 1999 |
13-2 |
87% |
9-2 |
82% |
| 2000 |
8-7 |
53% |
7-3 |
70% |
| Total |
73-50 |
59% |
55-33 |
63% |
It might be there has been a shift, but if so, it has been in our favor! From 1995-2000 you would have experienced six straight winning years with the aforementioned strategy. Of course this method has come about through backfitting and there is no guarantee it will continue. On the optimistic side though it seems to be logical that people would tend to play the teams based on last year's numbers, and consequently there is not enough adjustment being made to account for the rapid rise and fall of NFL teams in this era.
2001/2002 Update:
Two more winning seasons chalked up to make it eight straight years of success!
| Year |
Play against
better
"Net Pts" team |
Win% |
Play against
better net and
perceived value |
Win% |
| 2001 |
7-4 |
64% |
4-1 |
80% |
| 2002 |
7-4 |
64% |
6-4 |
60% |
We did a little more investigation of the "net points week one" angle and found the following possible improvements to the approach: require at least a 2 point difference between the teams' net points the prior season, and at least a 2 point "perceived false value" when comparing the actual line to a line constructed by using last season's net points (with a fixed 3pt home field adjustment).
Copyright © 2002 by TwoMinuteWarning.com, All Right Reserved
|
