Hey All,
In the offseason I worked on a new football model for spread betting and I'll be posting my picks here throughout the year for tracking purposes. I have backtested 5 seasons and the results are below. This season is purely a test run, but I will be betting a on all the picks at $10 per unit to make it interesting. There are 3 unit sizes depending on the confidence of the model. Feel free to tail the picks but do so at your own risk, there are no guarantees!
Results:
2008:
1* - 37-38
2* - 11-10
3* - 4-3
-4.2 units
2009:
1* - 37-43
2* - 19-15
3* - 9-5
-4.5 units
2010:
1* - 49-25
2* - 17-5
3* - 6-4
+30.8 units
2011:
1* - 44-29
2* - 16-10
3* - 5-2
+17.5 units
2012:
1* - 42-34
2* - 16-6
3* - 5-5
+11.7 units
Model Basics:
The model looks at each game as a combination of both team's "profiles" (essentially a collection of stats) and compares the current game to a database of past seasons. Through a set of parameters, it identifies similar matchups and calculates the average score difference in those games. If it is significantly different from the line for the current game, then it is selected to be bet. Depending on the confidence of the model is it either a 1*, 2* or 3* game. What I believe is the novel idea in the model is the method used for comparison to the historical games, but for obvious reasons I won't be disclosing what it is.
In order to establish team profiles, I won't be posting any picks until week 3. Picks will generally be posted earlier on in the week.
Good luck in 2013-2014!