- David J. Phillip/AP
Filling out a NCAA Tournament bracket and predicting winners, upsets, and the ultimate champion has become a passion for college basketball fans and non-fans alike.
While nobody has ever picked a perfect bracket before, ESPN’s college basketball analyst Jay Bilas has an interesting strategy for picking a winner that could help dictate your picks.
In a video for ESPN, Bilas broke down how he chooses the ultimate tournament winner.
First, though it’s fun to pick upsets or a Cinderella team, according to Bilas, no team with a seed line of nine or higher has ever won the NCAA Tournament. You can cross off 36 teams immediately.
Next, Bilas says, the last 13 winners of the tournament have had 25 wins or more when entering the tournament. This year, you can cross off 12 more teams to get down to 20.
From here, it gets a bit more stat-heavy. Bilas likes to look at Rating Percentage Index (RPI), a stat that measures teams’ strength of schedule and how they do against that schedule. While it’s not a perfect stat, according to Bilas, since 1994, 22 of the last 23 national champions have had an RPI of 17 or better. We’re down to 14 teams.
Basketball is obviously a two-way game – teams have to play good offense and defense to win games. Bilas looks at teams who have ranked in the top 20 in adjusted offensive efficiency and top 15 in adjusted defensive efficiency – points scored and allowed per 100 possessions, adjusted for strength of opponent. By filtering out those teams, this year, we’re left with just three contenders: Villanova, Kentucky, and Gonzaga.
Bilas then uses one more criterion to pick a winner: offensive rebounding. As Bilas puts it, “It helps to have a safety net in case shots aren’t falling.” Bilas looks to see which teams have an offensive rebounding rate above 30%. This year, that leaves just one team:
It’s an imperfect strategy, but it’s methodical and finely tuned and could bear results if you’re looking to win a pool amongst friends, family, or co-workers.
Watch Bilas’ breakdown below: