Model Madness 2020: Update, Conclusion

This post serves as a quick update to the Model Madness series as it is the last day of March and I wanted to give some closure to the series for the year. As many of those following sports know that all contact sports have essentially been suspended or cancelled. That included the remainder of the college basketball season. Since basketball is cancelled indefinitely, I'll keep this post short and summarize some of the data I was able to collect.

I was able to project the conference tournament games prior to the cancellation of the remainder of the season. The table below describes the accumulated results across all conference tournament games that were able to be played for each of the models.

WRI came out on top, nearly predicting 70% of games correctly. EDM and MASPM tied for second with 68.5%. SPM was in fourth with 66.7% and EDM Contra was unsurprisingly the worst model coming in at 63%.

Although WRI had a smaller sample size, it appeared to slightly outperform the other models in the tournaments that they both covered which perhaps means that there is some information gain that we are achieving through the optimization process that WRI uses.

The other new model EDM Contra performed the worse, but was able outperform other models when upsets occured, which means that it was able to achieve what we wanted it to achieve. Overall it performed poorly, but was able to be successful in some individual tournaments.

Below are the results divided into conference. In Bold are conference champions. In Italics are potential champions (since the tournaments were cancelled prior to completion).

I plan on eventually releasing some of the model scripts that I used, but there hasn't really been a lot of incentive to work on this project given the news. But I'll link those models to this post once I get around to cleaning and commenting those up. Thanks to everyone who took the time to read and support these posts.