I received the following email yesterday :
“Can you help me understand your logic behind the Wright State Pick. I am a fairly serious winning small stakes CBB player and I am interested in your methodology if you would be willing to share.”
I really wont divulge my exact methods but can make the following comments:
This is the first year I have been playing college baskets using this method. There are so far then only 96 plays and we are hitting 63.5%. Statisically it means very little. Once we get up to 1000 plays then maybe the numbers start meaning something and at 5000 you will get a pretty good idea how valid the method is.
So with that caveat this is the first season I have completely ignored what the public is doing in ncaa baskets and also have eliminated short dogs from consideration no matter how good they look to me. A short dog as I define it is +1.5 to +6.5. With the nature of the scoring in the end game a pick that is a short dog to win has to end up with things going perfect for you at the end of the game.
A game that is a good example of that is the texas tech Oklahoma game Saturday night. The line was TT -4.5. Oklahoma lead by around 8 at the half. And it wasnt until the last 30 seconds that the score ballooned up to 10. A bettor that had Oklahoma would be telling everyone today what a close game that was and really looked like Oklahoma should of covered. But most short dog games look that way going into the last 30 seconds are so…then the fouling and 3 point shots determine the final score.
Now I have had a handicapper I know mention I should consider +1.5 and +2 as not falling into the short dog category. My cutoff of 1 is arbitrary to a degree…but I have found at some point you have to give some conseration that the books know what they are doing…in otherwords respect the line! So I start respecting it at 1.5 to 6.5.
As to Wright State……….I use 2 neural net models to come up with a predicted winner. Both are run independent of each other and both use different approaches to determine the outcome of the game. I combine that with 2 other models that look for strong situational tendencies that might exist in the matchup between the two teams. For me to pick a side all for of my models has to be in agreement. If any one of the models is even neutral on the game I pass.
I hope that explains a little about my methods. The software packages I use for my models was developed in principle by me and then put into action using a programmer to get everything workable for me.
Thanks for the email