NCAA Tournament Round of 64 Predictions

Why do we love March Madness so much?

Because we enjoy the variability a one-game series can bring — anything can happen in March. But, this makes for major issues in predicting winners in the NCAA Tournament, which is why we love filling out our brackets every March.

However, it wouldn’t be fun if we didn’t cling to the information we do have and try to predict tourney games. So, I have used kenpom’s pre-tournament rankings and the selection committee’s 1–68 rankings to create a team composite score for all teams entering the tournament from 2015–2017. Using this composite score and several measures of team efficiency we can model both the expected point spread and the win probability of each of the Round of 64 matchups.

There are a number of limitations — namely, that something as variable as the NCAA tournament can be very difficult to predict with much confidence. In the future, I would hope to find other predictive variables. For instance, I did not find the distance teams had to travel to be a significant predictor, but surely it has some effect on the strength of teams entering the contest. Further, would be simulating every round of the tournament-but with the tournament starting today, here are my round one predictions.


First, the left side of the bracket: the East and West regionals. The two biggest games to watch, in terms of close contests are #7 South Carolina vs. #10 Marquette and #8 Northwestern vs. #9 Vanderbilt. Marquette posses an upset threat in the East regional. #11 Xavier will also likely give #6 Maryland a handful in the first round. It is intriguing to compare both how the probability model holds against the weekend’s results and how the MOV model compares to Vegas lines:


The right side of your bracket, the Midwest and South regionals, pose many more close contests and upset threats. The #8 vs. #9 contest, as expected, between Miami and Michigan St. should be a close race. Last year’s upset darling, Middle Tennessee, is back with a 37.1% chance to knock off #5 Minnesota, but the line produced by the model is relatively even. Lastly, Wichita St., though the #10 seed facing the #7 Dayton, is a favorite from my model with a 58% chance to win.

The models appear very similar to other forecasts of the tournament, which provides validation. Yet, it will be intriguing to see, in places of difference, how successful certain predictions are.

As the tournament advances I will post the successes, and quite possibly, the failures of the model, as well as how the model performs against the spread. Secondly, I will display predictions once the Round of 32 match ups are established. With such a turbulent time of year in college hoops, let’s all be prepared to have our brackets busted — it’s just a matter of time.