Prediction Primer

Before I start putting up the picks, I'll write a quick little blurb about what I'll be posting and how to interpret all of the info. A game's prediction will look like the following:

AWAY at HOME, TOTAL
Winner: PICK, %
Spread: PICK, %
Total: PICK, %

NYG -8.5 at STL, 41.5
Winner: NYG, 76%
Spread: NYG -8.5, 50%
Total: OVER, 66%

So that means: The Giants are an 8.5 pt favorite playing in St. Louis. The over/under line is 41.5. NYG wins outright in 76% of the simulations (380 of 500 simulations). NYG covers the 8.5 point spread in half the simulations. The total score exceeds 41.5 points in 66% of simulations. Simple enough, yes?

Last season, it turned out that the confidence when picking winners actually matched the real life probability of that team winning. (One of the first blog entries has more info on that.) If I simulated 100 different match-ups over the course of the season and they all had a confidence of 65%, in reality, 65 of those 100 picks would be correct. That's extremely useful for gambling purposes, so hopefully that correlation will continue, though there are no guarantees.

I'll also be posting weekly offensive, defensive, and overall power rankings and ratings based on my simulations.

I combined the Kelly criterion with some of my own calculations to determine the optimal bets based on the payout of the bet and the confidence of the pick. Check out en.wikipedia.org/wiki/Kelly_criterion for more info. I'll post the results of the calculations as betting suggestions.

I got a request to post some stats from the simulation, so I'll put those up as well.

Any questions/comments, feel free to shoot me an email.