clock menu more-arrow no yes mobile

Filed under:

NCAA Lacrosse Tournament: How Does a Computer Pick the First Round Games?

log5 probabilities . . . ahoy!

Mitchell Layton

Look at you. You downloaded your bracket and you figured out which games you have to watch because they'll be the most fun. You're doing pretty good today, Internet friend. There's just one problem: You don't know which games in the first round are the most competitively balanced. Luckily I'm here to help you out with that.

There's a principle that Bill James invented called log5. It's basically a matchup probability, estimating the probability of one team beating another. Applying it to college lacrosse can create some uneven results -- teams are playing less than 20 games in their regular season -- but it does a decent job of illustrating how competitively balanced matchups are. Here's how a log5 calculation sees the NCAA Tournament first round:

(8) Penn State // Yale Penn State (51.76%) Yale (48.24%) (6) Maryland // Cornell Cornell (58.60%) Maryland (41.40%)
(5) North Carolina // Lehigh North Carolina (51.88%) Lehigh (48.12%) (4) Denver // Albany Denver (58.65%) Albany (41.35%)
(3) Ohio State // Towson Ohio State (58.13%) Towson (41.87%) (1) Syracuse // Bryant Syracuse (79.71%) Bryant (20.29%)
(7) Duke // Loyola Loyola (58.28%) Duke (41.72%) (2) Notre Dame // Detroit Notre Dame (81.39%) Detroit (18.61%)

Some brief thoughts:

  • In a season where things have gone sideways in a hurry and expected outcomes are relegated to nothingness, the probabilities for some of these games shouldn't be treated as dogma. There is some truth to the expectations -- Lehigh-North Carolina is very evenly matched (this is a style-makes-the-fight kind of game); Yale-Penn State is going to be a war with all kinds of collateral damage; and Bryant-Syracuse and Detroit-Notre Dame are expected to see the unseeded team take shots to the face (especially the latter matchup). There is, however, also some noise -- Albany's style and offensive capability combined with how Denver plays likely draws the success expectations between the two programs a little closer together; the Red as a favorite over Maryland shouldn't make brains leak out of your ears, but I’m not sure if Cornell is close to a 60-40 favorite; and Bryant's ability to dominate possession is not factored into the log5 analysis, thus making Syracuse's overall expectation probably a little smaller, etc. The log5 gets you close, but make the melon stuck in your skull do some work as well, college boy.
  • The log5 calculation projects only two unseeded teams -- Cornell and Loyola -- to trump their seeded opponents. Interestingly, those two games feature the fourth and fifth largest probability margins on the board and not, according to the log5 probabilities, functional "push" games based on where the seeded teams were slotted. Weird.
  • This tournament is all about matchups; who you draw is arguably more important than where you're seeded (or unseeded). With respect to matchups and seedings, I ran a few other permutations:
    • If North Carolina had drawn Towson as the field's three-seed: The Tar Heels would have a 70.86% probability for victory. Instead, Carolina holds a 3.76% probability margin over Lehigh.
    • If Syracuse was bracketed with Detroit (the weakest team in the field): The Orange would have an 82.25% probability for victory. Instead, Syracuse holds just a 59.43% probability margin over Bryant.
    • If Cornell earned Penn State's seeded position: The Red would have a 67.97% probability for victory over Yale. Instead, Cornell holds a 17.21% probability margin over Maryland.
    • If Ohio State had drawn Lehigh as the field's five-seed: The Buckeyes would only have a 38.10% probability for victory. Instead, Ohio State holds a 16.28% probability margin over Towson.