Thanks for asking. I probably wasn't going to throw it on the CS post, but I will put it here now. The original prediction was a four point victory and 69% chance of a win, so all of the recommendations baseline from that. Each recommendation is cumulative. Yes, I know that the prediction is plus 13 when we won by 22. The recommendations are the "top five", and not all aspects of the model, and it's only a model so there are limitations.
Recommendations vs. Results
Specific Statistics Recommendations (In order of importance) with results
1. Achieve an eFG% of 55% or higher, which translates to 27 made field goals (including 7 threes), which is right at season average and gives 70% chance of Marquette win
Result – eFG% was 45.5%. Advantage West Virginia. Changed margin of victory prediction from plus four to minus four
2. Limit WVU to an eFG% of 47.5% or less, which translates to 26 made field goals (including 7 threes), which is right at their season average
Result – WVU eFG% was 39.2%. Advantage Marquette. Changed margin of victory from minus four back to plus four
3. Limit WVU to an offensive rebounding percentage of 40% or less (16 offensive rebounds, which is one less than their average)
Result – WVU had an OR% of 35.9%. Advantage Marquette. Changed margin of victory prediction from plus four to plus six
4. Commit turnovers at a rate of 21% or less (~14 turnovers)
Result – Turnover Rate was 12.3%. Advantage Marquette. Changed margin of victory prediction from plus six to plus 13
5. Force WVU into a turnover rate higher than 20.5% (also about 14 turnovers)
Result – WVU Turnover Rate was 20.4% - Push.
