Here is a quick recap of how our models performed in the 2018 NCAA Tournament. At the bottom of the article, you can find links to all of our content and predictions from this year.
Model Accuracy in 2018
All of our models performed slightly worse in 2018 than they did in 2017, but they were still beating Vegas Spreads and Totals at a good rate, and our Win Probabilities still hit 70% of games correct. Our bracket was busted with Purdue and North Carolina having early exits, but we did have Kansas in the Final Four, our simulation pegged Villanova and Michigan as the 3rd and 6th most likely teams to make the championship game, and Villanova as the 2nd most likely to win it all (14%). Our Cover Probability model was also very strong on Villanova, and predicted them to cover every game they played (which they did).
Below is how each individual model performed on all 2018 tournament games (spread/cover and total/over models were measured against each game’s closing spread and total):
- Win Probabilities: 44-19 (70%)
- 284 Spread: 32-28-3 (53%)
- Cover Probability: 35-27-1 (56%)
- 284 Total: 38-25 (60%)
- Over Probability: 36-27 (58%)
Model Accuracy Last Two Years
|Year||Win Prob||284 Spread||Cover Prob|
|Year||284 Total||Over Prob|