For the fourth year in a row, the TeamBrunnhilde linear model was the most accurate ranker in predicting the outcome of state tournament games. The linear model for win-loss had an impressive debut, tying for second place with MaxPreps. Fourth was a composite ranking. WIAA RPI was fifth. Everybody did a decent job of picking champions as their #1 ranked team. Three rankers tied with eight #1’s that were champions: Evans Ranking, WIAA RPI, and TeamBrunnhilde linear model. Lynden the top seed in Boys 2A won that championship, but was #2 in WIAA RPI. Sorry, WIAA. No mulligans.
I’ve read complaints that WIAA RPI doesn’t count Win-Loss enough. Win-Loss as a ranking algorithm finished dead last. I think there are better rankings than RPI, but RPI is an attempt to improve on Win-Loss with an understandable method for the less-mathematically inclined. This year demonstrates that a properly done RPI does that. Last year WIAA RPI had fatal flaws: not counting playoff games and using fake data for out of state teams. With these corrected WIAA RPI finished 5th, well ahead of Won-Loss (which it trailed last year).
The TeamBrunnhilde linear model for win-loss, similar to TeamBrunnhilde linear model for points but with a different Y vector, did pretty good considering that the mathematics are not best suited for binary data (win-loss). I’ll have to look at that before next year.
Everybody had successes. Everybody had misses. Not just at the top of the rankings. We should all go back and work on slaying the variance monster next year. But as Jesus might say, “The Random you will always have with you.”