This post summarizes the performance of R2 Sports Metrics performance in forecasting NFL and NCAA CFB game results. Overall, we had a very successful season, with our NCAA football model, CBCR2 Sam, leading the way.
CBCR2 Sam provides tremendous value in betting markets due to its outstanding performance in choosing sides against the Vegas point spread, finishing 326-249 (56.7%) against the closing line (ATS). Because Sam's ATS performance improves the more its projection differs with the point spread, I was able to identify and publish 108 bets over the course of the season (@DawgQuant), with 66 wins and 42 losses (61.1%). If instead you bet $100 on every CBCR2 Sam projection, you would have earned a weekly ROI of 8.3%. Sam's ATS performance is the best season-long performance of any public model I can find since 2013, and Sam is the reason I finished 1st of 40+ in the 2022 reddit computer Pick'em challenge ATS category (minimum picks = 300).
While Sam's value is primarily derived from its ATS picks it was also very efficient. It picked winners at a 72.8% rate and finished with an outstanding 11.99 mean absolute error (MAE).
R2 Sports Metrics NFL system, R2 Roy, also performed very well in 2022 and provided significant value in betting markets, going 147-129 ATS (53.2%). I was able to use the difference between Roy's projection and the point spread to identify 74 NFL bets, with 46 winning bets and 28 losses (62.2% ATS). Only two publicly available models outperformed Roy in ATS picks last year (FF Winners and Cleanup Hitter) per thepredictiontracker.com.
R2 Sports Metrics plans to continue to publish R2 Roy and CBCR2 Sam for the 2023 season. We also hope to expand to NCAA and NBA basketball to unlock more markets and opportunities to find value.