2023 NCAA Tournament -- Formula for success? Apply math to filling out your men's March Madness bracket

A little bit of an introduction. I'm a researcher in ESPN's Fantasy department and I've developed a formula, abbreviated here as SOP, as a way to help make filling out the bracket a little less stressful. This formula is inherently simple: Measure and weigh appropriately the handful of stats over time that have proven predictive in March. What is spit out is both an outright winner and a projected point difference, with my custom spread for every game.

I've run this formula for a few years, back-tested it even further back and it has proven to be accurate. As accurate as you can hope to be for an event labeled as «madness.» The tournament last year, as a whole, was wildly encouraging despite my strongest teams coming up short.

If we exclude St. Peter's three wins last year — understanding that no one in their right mind will adjust anything based on a historic outlier — the formula was «too heavy» on favorites by 0.615 points per game. By «too heavy,» we mean cumulative spreads versus actual results. That is, I made Kansas a 7.5-point favorite in the national championship. It won by 3 points, so I was too heavy by 4.5 points. In the other direction, in the Final Four, I made Duke a 6.4-point favorite over North Carolina and it lost by 4. So, in hindsight, I was too light by 10.4 points. Using just those two games, I was cumulatively too light by 5.9 points (too light by 10.4 points and too heavy by 4.5, thus the cumulative result is being too light by 5.9 points).

Complete your bracket by selecting the winner for each game of the 2023 men's NCAA tournament. Play Tournament Challenge

Again, without St. Peter's wins, my custom formula was 0.615 points per game too heavy on the favorites. Vegas? Their opening lines were