2017 March Madness — Firing up the machine!

My friend Tanner from work updated a script I wrote last year to scrape NCAA team stats that I used to create an algorithm to predict which teams would make it to the Final 4. This made me want to look back at my machine learning performance from last year. My bracket performed OK, ending in the 79th percentile.

I’m going to track my predictions leading up to the 2017 tournament and see how consistent they remain. I am simply using Win %, Strength of schedule, Offensive efficiency, defensive efficiency, and the Associated press pole high ranking as features to determine success in the tournament.

Here are the top 10 performing teams according to a Random Forest Regression model. Remember, 1.0 = Sure bet to get to the final four or win!