The suffix -ology is generally reserved for disciplines that tackle the most complex and integral problems in our society.
Virology, psychology, theology — essential ologies that help us make sense of our baffling world.
With fans becoming increasingly confused over the nuanced and ever changing process of selecting the NCAA basketball tournament field, the term Bracketolgy has aptly navigated its way into the public lexicon.
Bracketology is the process of using analytics and machine learning to more accurately predict NCAA tournament seeding and results. Generally, these predictions are generated on a frequent basis, starting as early as December.
Florida’s men’s basketball team has generally avoided the dreaded NCAA bubble — UF has qualified for 18 of the last 22 tournaments, earning a seed higher than seven only twice. But the Gators’ current odds teeter on a knife’s edge, with BartTorvik giving them a 57% chance of making the tournament fewer than two months from Selection Sunday.
alligatorSports: What is Bracketology at its core?
Cook-Shugart: It’s really a growing area of interest and it’s still certainly being refined. Data science and machine learning are kind of the buzzwords people use. Where machine learning comes in, is from a statistical standpoint, you sort of set up an algorithm that you want the machine to execute, and then the machine learns.
That's where the name machine learning comes in, it learns from data that you feed it, and iterates over that data and makes predictions about what's going to happen based on what happened in the past, essentially.
You take everything that the committee would look at and put that in a tabular format. Then the computer, based on what I put in will say, ‘Okay, I can recognize patterns that humans literally aren't capable of recognizing, and then spit out these projected seeds for these teams,’ and then I take that as a baseline and then use sort of an eye test to blend with it.”
alligatorSports: What data goes into the projections, and how are these points weighted?
Cook-Shugart: So that's the cool thing about machine learning, I don't have to set any weights or anything. I feed the program data past data, and it determines how every input should be weighted based on how it's been weighted in the past.
You can look at what’s called feature importance. In machine learning each of these things that determine each of the independent variables that determine the dependent variable output, are features. There's, based on my model, five features that really stand out as really driving how a team will be seeded:
One, it's unsurprising, it's the reason that they [the NCAA selection committee] made it, the committee really leans into the NET Ranking. The whole point of developing it was that the RPI was kind of outdated, and wasn't perfectly accurate. [The committee] got to put in mostly team performance, but then have some elements of how a team should have performed based on advanced analytics, like your Kenpom, that feeds into the net.
Second, is actually the amount of points the team is receiving in the AP Poll. That is essentially the best way to sort of fledge or quantify an eye test because that’s where you get the opinions of AP voters.
Those are your top two, then quadrant one wins. The committee talks about it every time they interview the tournament selection committee chair, he says, ‘Well, they had a lot of quad one wins and a lot of quality wins.’
Number four is quadrant four losses. Unfortunately for Florida, because when you're on the bubble, that's about the worst thing that you can have. Essentially you need more good wins than bad losses. The last thing is strength of schedule. Strength of schedule as a whole is more indicative of how a team's going to be seated than any quad two or quad three wins or losses.”
*NOTE Florida currently ranks 39 in NET with one quadrant four loss to Texas Southern and one quadrant one win over Ohio State.*
alligatorSports: Where does Florida stand in these data points?
Cook-Shugart: Florida will have a lot of opportunities to play the way in or out of the tournament. The SEC is strong, at least the tournament caliber teams, but the bottom of the SEC is not great. you need to kind of avoid losing to the you know, like the Ole Miss-Georgia-Missouri contingent, and you'll be fine there.
Other than that, road wins, because you like that you can beat less good teams on the road and get a quad one win just because of how the quadrate system works. If it's in the top 75 in the net on the road, that's a quad one win, but it has to be in the top 30 for a home win to be a quad one win. Compared to the rest of the country, for a bubble team only having one quad one win, they're a little bit light there. They only really have one bad loss, so they're in a pretty good spot there.
They’re (Florida) literally my last team in, but I wouldn’t want Florida fans being mad at me if they ended up being the first or second team out if Selection Sunday was today. They’re right there.
alligatorSports: What does Florida have to do to cement themselves as a tournament team?
Cook-Shugart: They kind of control their own destiny, because the SEC is pretty strong at the top and in the middle.
I think 19 wins in this conference. You know, like that's not hard and fast. 18 might do it, it might take 20, but about 19 wins just based on the remaining schedule that they have and the strength of the SEC. And, regular season wins are more important than conference tourney wins.
So a lot of bubble teams love the idea of playing themselves in, in the conference tournament if they're kind of close. You usually have to make it to the conference tournament final to really have an impact to offset your regular season performance. Otherwise, regular season performance tends to be weighted a little bit heavier. You have to take care of bad teams. After that, other games are just opportunities to sneak a couple more good wins in. I think a good finish would be 21 wins.
I think anything less than 18 wins is kind of a disaster. Florida fans should be mad at the team and not the committee if they leave them out with 17 wins or something. I’d say best case would be [the nine seed], worst case is out of the tournament and my best guess is they’ll be one of the last four in, right where they’re at right now.