This is going to be short, because we still have some testing left—and potentially a few updates in addition to the flip back to simulations—but we always want to give those viewing any of our models’ outputs an explanation of why the model is saying what it says, and the current model is significantly different from where it was as of our last explanation, so an update is in order. As always, feel free to ask questions if you’d like to know more details. As always, no, we’re not FiveThirtyEight in terms of our quality quite yet. That’s a high bar, and we’ll hopefully one day get there.
Predictive, Not Reflective; Current Shortcuts
Our model continues to be forward-looking, a projection of where things will land on Selection Sunday rather than a reflection of where they stand today. It looks at each team’s median projected résumé, although in a temporary break from how our models usually work, it’s doing that with a few shortcuts rather than a robust sample of simulations. Those shortcuts right now consist of using rating systems to project where the “raw score” portion of the eventual seeding formula (explained below) will land. For other variables within the formula, median results are still being used, making them identical to what we’d be seeing with that robust sample of simulations, with the lone exception that we aren’t correlating NET changes to results, as would happen within the simulations, which may be leading to a few marginal errors.
Automatic Bids
For the NCAA Tournament, the automatic bids go to the current conference tournament favorites. For the NIT, the automatic bids go to the current regular season favorites most likely to lose their conference tournament, should they retain the number one seed. To determine the number of NIT automatic bids awarded, we look at the expected number of NIT automatic bids, summing the probabilities of an NIT automatic bid from each conference.
Bid Thieves
To determine where the NIT’s upper cut line falls, we sum the probability of a bid thief emerging from each conference and raise the cut line from its natural spot by the number of thieves expected. As of today, March 1st, we expect two bid thieves. This can and may well change, but we don’t know in which direction.
Seeding Formula
Our seeding formula is based off of a raw score and a broad set of exceptions.
The raw score is a combination of NET, SOR, KPI, and KenPom, comprised of thirteen parts NET, thirteen parts SOR, twelve parts KPI, and one part KenPom, with each variable adjusted so as to plot it on a normal distribution, making the gap between 1st and 2nd significantly wider than the gap between 21st and 22nd, and so on. We’re using a projection for this, rather than just where things stand right now, but the projection is in large part based on where things stand right now. This is the backbone of our seeding formula.
From the raw score, the model moves teams up and down based on an array of possible exceptions—variables we’ve seen in recent years (2018-present) to affect the committees’ evaluation of teams. Current exceptions consist of:
- Is the team projected to finish at or above .500 overall? Is the team projected to finish more than one or two games above .500 overall?
- Is the team projected to finish with a Q1 win percentage at or below 25%? At or below 15%?
- Is the team projected to finish with five or more Q1 wins?
- Is the team projected to finish with ten or more combined Q1 and Q2 wins?
- Is the team projected to finish with a winning record in Q1 and Q2 games?
- Is the team projected to finish with a nonconference strength of schedule, by the metric on the NCAA’s team sheets, of 340th or worse?
- Did the team go undefeated in nonconference play against a full nonconference schedule (eleven games or more)?
- Is the team’s worst rating-that-matters (NET, SOR, KPI, KenPom) projected to finish 50th or better?
- Is the team projected to lose a Q4 game in its conference tournament?
These exceptions may change, and we’re still working on the proper weights to assign each. We may add or remove exceptions. When the formula is locked for the time being and we’re running the full simulations, we’ll post another update. For now, though, this is what we use to line the teams up, 1 to 358.
Bracketing
Once we have the automatic bids determined, the NIT’s upper cut line determined, and the teams lined up in order of seeding, it’s as simple as following our impression of the NCAA’s bracketing principles for each tournament, with one exception: We don’t move NIT teams across seed lines for geographic convenience, as the committee is allowed to do. We found our previous attempts at doing this in the model introduced unnecessary confusion and chaos within the projections.
***
We’ll try to have another update for you all soon. In the meantime, thanks for reading The Barking Crow, and thanks for making our bracketology part of your college basketball fanhood.
