"My rule was I wouldn't recruit a kid if he had grass in front of his house.
That's not my world. My world was a cracked sidewalk." —Al McGuire

Friday, February 05, 2021

Patriot Games

Earlier this year, some talking heads around college basketball were calling for a season without conference games. To anyone that studies brackets, this would be a disaster. Why? Because systems like the NET need the cross-pollenation of data we get from non-conference games. If teams don't play other leagues, or only do so in a limited fashion, it's more difficult for a computer system to separate them those leagues. Especially a computer system that uses only season based data like the NET (as opposed to predictive systems that incorporate past data). After all, how do you know whether the Big East or MEAC is the better league if those teams never play each other or share common opponents?

Thankfully, league agreed to play non-conference schedules and we were saved from a seeding nightmare. Almost.

Are we ready for Jordan Burns and...**checks notes**...4-seed Colgate?

Photo courtesy of Colgate Athletics

Enter the Patriot League. Three Man Weave touched on this in their podcast, but it's worth going deeper into this topic. Only two teams in the Patriot League elected to play non-conference games: Army and Navy. They did pretty well for themselves in those games, going a combined 6-2. The only losses came to high major programs projected to make the NCAA field. Their wins included victories over teams from the Big East, A-10, and the MAC, leagues typically much better than the Patriot. Because of this small sample size, the Patriot's value is massively overinflated. In the first two years of the NET, the Patriot ranked 19th and 21st among conferences in the NET and 21st and 23rd in kenpom. This year, the Patriot is ranked 14th among conferences in the NET but just 22nd in kenpom. So predictive metrics indicate the league is no stronger than usual, but the NET rates the league far stronger.

The conference rating isn't the only thing out of whack. Colgate has been the league's best team three years running, with NET rankings of 132 in 2019, 121 in 2020, and 15 in 2021. According to kenpom, they have been ranked 126 in 2019, 118 in 2020, and 99 in 2021. Clearly something is wrong, but it isn't just with Colgate. Take a look at the disparity between Patriot League team rankings in the 2021 NET and kenpom:

Team NET Rank kenpom Rank NET Overvalue
Colgate 15 99 84
Army 87 177 90
Navy 95 165 70
American 175 251 76
Lafayette 177 214 37
Bucknell 184 245 61
Boston University 217 258 41
Holy Cross 240 339 99
Lehigh 275 328 53
Loyola-Maryland 277 260 -17

Average 174.2 233.6 59.4

Trust me when I say that while NET and kenpom disparities are common, it's not common for teams to have a difference of more than 30. Every now and then you see it with an overachiever like Drake who is killing everyone they play, but that's because they are fighting their way up against the preseason projections in kenpom that drag them down. In the case of the Patriot, it's a lack of data population for the NET, and it's a problem with almost every team.

Colgate is the easiest example to look at. They have outscored their Patriot opponents by an average of 21.4 ppg. That includes an average margin of 21.0 in two games over an Army team that holds wins over Buffalo and La Salle, even though one of those two contests was a LOSS for Colgate. As a result Colgate is rated higher by the NET than any team in the MAC or A-10. However, according to kenpom, Colgate would be the 3rd best team in the MAC and 9th best team in the A-10. Perhaps even more impressive (and likely due to their margin over the Army team that beat the Navy team that beat Georgetown) is that Colgate would be the 2nd best team in the Big East (behind only Villanova) despite kenpom indicating they would be 10th best in the league, ahead of only DePaul by a narrow margin.

While Colgate is the easy one because of their absurdly high NET, the consistent overvaluation of these teams indicates that unless Santi Aldama and Loyola Maryland represent the Patriot in the NCAA Tournament (an anomaly due to narrow losses against teams that didn't play non-con games and a blowout against Navy, who did), any team that earns the Patriot bid will be overseeded compared to their actual quality. What does all this mean? Non-conference games matter. They are the only thing that allows a metric like the NET to operate with some level of sanity. Here's hoping the NCAA Selection Committee heavily considers this when it comes to Selection Sunday, because if the Raiders win the auto-bid, they probably belong around the 14-line (the rest of the league are 15-seeds at best), but when the NET is considered, it wouldn't be crazy to have them as a single digit seed and the rest of the league a couple lines higher than they deserve.

Here's the current S-Curve:


2-Seeds: 8-Iowa 7-Ohio State 6-VILLANOVA 5-ALABAMA

3-Seeds: 9-Texas 10-Illinois 11-VIRGINIA 12-Tennessee

4-Seeds: 16-Oklahoma 15-Missouri 14-Texas Tech 13-Wisconsin

5-Seeds: 17-West Virginia 18-Kansas 19-UCLA 20-Usc

6-Seeds: 24-Colorado 23-Purdue 22-Virginia Tech 21-Florida State

7-Seeds: 25-Creighton 26-Connecticut 27-Clemson 28-Oklahoma State

8-Seeds: 32-Minnesota 31-Byu 30-Xavier 29-Florida

9-Seeds: 33-Louisville 34-DRAKE 35-Rutgers 36-Arkansas

10-Seeds: 41-San Diego State 40-Loyola Chicago 38-ST BONAVENTURE 37-BOISE STATE

11-Seeds: 41-Oregon 42-Colorado State 43-Richmond 44-Lsu

12-Seeds: 50-WINTHROP 49-TOLEDO 48-Stanford 47-Utah State 46-Seton Hall 45-Maryland






Last Four Byes: Oregon, Colorado State, Richmond, LSU

Last Four In: Maryland, Seton Hall, Utah State, Stanford

First Four Out: VCU, Indiana, North Carolina, St. John's

Next Four Out: Pittsburgh, Syracuse, SMU, Georgia Tech

Multi-bid Leagues

Big 10: 9

Big 12: 7

SEC: 6

ACC: 5

Big East: 5

Pac-12: 5

Mountain West: 4

Atlantic-10: 2

Missouri Valley: 2

WCC: 2

No comments: