An Attempt at Prediction At-Large Bids in the NCAA Tournament

I have been working on a revamped NCAA Bracket Simulator that will be available to you all early next week. On Friday, with a few moments to spare, I decided to see if I could take all of the information I’m loading for that model and use it to predict which teams in the Men’s tournament are most likely to receive an at-large bid.¹ I truly have no idea how well this will work, but want to show my work, just in case there’s something to all this.

First, I want to note that my biggest hurdle is a lack of training data. I’m not using any “expertise” here, just purely running the numbers. And the numbers I have to go off consist of three seasons of data².

Second, here are the alleged criteria³ for selecting an at-large team:

• Game results

• Strength of schedule

• Location (home, away, or neutral site)

• Scoring margin — Teams receive no added credit for victory margins above 10 points. Additionally, overtime games will be assigned a scoring margin of 1 point, regardless of the actual score.

• Net offensive and defensive efficiency

• All games will be evaluated equally; there is no bonus or penalty for when a game is played within the season.

• Quality of wins and losses (this gets a little complicated)

Fortunately, In the amount of time I had to throw all of this together, I can pretty easily get this data put together with two exceptions:

1. Net efficiency ratings - I would need to track down another data source for this and I’m not sold on them actually using this criteria anyway.

2. Tracking overtime games - If a team loses by 5 in overtime, I’m just counting it as a 5 point loss, not 1.

So, what are the results?

A fun little Treemap of all 364 NCAA men’s basketball teams

The larger the rectangle, the better chance the team has for an at-large bid. The darker the blue, the higher I have a team rated. I think it looks pretty decent!

And here are teams’ actual “probabilities”⁴ in an easier-to-read table:

a list of teams sorted by the likelihood that they will receive an at-large bid on Selection Sunday. If your favorite team is not on this list, they have a worse chance than UAB. Sorry..

I don’t think this is perfect. But I am interested to see how this holds up. I’ve got a busy weekend and some of these teams are still playing games, but I’ll check back in and re-visit this after selections are made.

1. Since the tournament expanded to 68 teams, there are 32 teams that automatically qualify for the tournament by winning their conference championships. The remaining 36 teams are selected in a smokey room by mysterious characters - and are called at-large bids.

2. If you are interested, all my data is on GitHub. Take a look and tell me how I can do better :)

3. source: wikipedia (I know…) - https://en.wikipedia.org/wiki/NCAA_basketball_tournament_selection_process#Selecting_the_field.

4. I think probability is too strong of a word, but these values are in order from most likely to least likely, so they do have value regardless.