Flo and Intermat have updated their rankings and things have shaken up a bit.
Ranking Methodology
For newcomers I thought I would include a little something on methodology as what I do differs from Intermat and Flo. I use the last 11 tournaments worth of results (2014 - 2025 - the 16 and 33 seed era) to build a model for expected points. I include placement, advancement and bonus points in my model. I fit the data in two dimensions to come up with expected points per seed, and the probability of each possible placement 1-33.
In this way my predictions are probabilistic. Intermat and Flo use binary predictions.
They assume their #1 ranked will finish first, their #2 ranked will finish second, etc. and they do not include bonus points. There are few consequences of our differences.
If a team with a lot of high seeds performs to seed my model will underestimate their results (not that I am thinking of any one particular team).
The Intermat and Flo models will underestimate the contributions of anyone ranked #9 - #33 as it will assume they never earn AA while overestimating the contributions of their top 8 ranked as it will assume they always AA.
- Intermat really loves Iowa State. They have them flipping spots with Nebraska to take over expected second by a fraction of a point.
- In Flo's opinion the top four remain unchanged.
- But the big mover is Oklahoma State which now slots in fifth in both rankings, two positions higher than last week.
Ranking Methodology
For newcomers I thought I would include a little something on methodology as what I do differs from Intermat and Flo. I use the last 11 tournaments worth of results (2014 - 2025 - the 16 and 33 seed era) to build a model for expected points. I include placement, advancement and bonus points in my model. I fit the data in two dimensions to come up with expected points per seed, and the probability of each possible placement 1-33.
In this way my predictions are probabilistic. Intermat and Flo use binary predictions.
They assume their #1 ranked will finish first, their #2 ranked will finish second, etc. and they do not include bonus points. There are few consequences of our differences.
If a team with a lot of high seeds performs to seed my model will underestimate their results (not that I am thinking of any one particular team).
The Intermat and Flo models will underestimate the contributions of anyone ranked #9 - #33 as it will assume they never earn AA while overestimating the contributions of their top 8 ranked as it will assume they always AA.
