Summarize Your Model?

What approach did you take this year? Which stats did you use? Massey Ordinals? Geographical? The detailed “team box score” stats? Like usual, I didn’t have as much time as I would have liked to spend on the predicting side. I’m sure we all feel that way! I got pretty far on a simplified possession-based model using some of the team box score stats, but it wasn’t helping the predictions so I abandoned it. I focused mostly on using the Massey Ordinals in an ensemble. I looked at which systems were available for the past nine years, and which of those had done best in late regular season games (daynum=100 or later) including top-80 teams, plus tourney games. As others have also mentioned, there isn’t enough tourney evidence so I…


Link to Full Article: Summarize Your Model?