Which tournament within the chosen group ("March Madness" is the D1 tournament).
Graph
Seed differential & fraction won: Chance of winning a game in a tournament
Individual team data
Individual tournament data: Number of games & upset avoidance rate within the tournament
Number of games (logarithmically scaled)
Upset avoidance rate β (logarithmically scaled)
Log odds against an evenly seeded opponent (logarithmically scaled)
Expected reseeding μ (logarithmically scaled)
Within the data, the probability of a given team winning their game is best modeled by
1/(1+exp(-β(s-μ))),
where s is the difference in seeding of the teams and -βμ is the log odds against an evenly seeded opponent.
Within a tournament, μ=0 because of symmetry.
