We’re back for the 2022 season!

Get ready for a new and improved model to pick NFL games in the upcoming season! These picks are the result of 4 different machine-learned models that have been optimized to win bets. In testing, these models won ~65% of the bets they placed, picking ~6 games per week. If this translates to real world results, then we are going to make some money.

Picks will be created starting in week 5 because we need 4 weeks of games to feed into the algorithms. Once we get there, bets will be posted on Wednesday afternoons. If I can make it work, this site will update during games to show how the picks are doing. After MNF, a new post will be created to show the win rate for the week and the whole season so far.

And now the long wait until week 5 :-/

Continue reading


NCAAM Picks

Can I predict the results of NCAAM Games?

The short answer is no.

The long answer is that I can try. I created a big data model to pull in some data points about two teams playing against each other, then spit out a predicted score.  The idea is simple: to predict the score of an upcoming game, find a really similar game in the past and assume that the result of that game will be the result of the upcoming game. This is an example of a KNN Prediction Model, but the description is on the Projects Page. This page will be focused on the daily picks and running totals.

The current model is very similar to my old bracket creating model.  I have changed my data set since then, but the methodology is the same. 

Continue reading