Machine Learning in Fantasy College Football: Joseph Richardson

The scoreboard. Four ballots play the same ten games every week: The Machine, Vegas, the Crowd and the Writers. Hover a name for how each one picks. Pick a week from the strip: the track is the running score through that week, the dashed line is a perfect card in every week of it, and the panel beside it is what each ballot scored that week. The grid is the week's ballot, dots sized by confidence, red where a ballot takes a different side from the Machine, faded once a pick has been decided and was wrong.

The Machine is a machine learning system that fills out a College Pick'em ballot every week of the college football season. Six models read thirty inputs on each game and vote on a win probability, the ten games on the ballot are ranked by how sure it is, and the ballot is scored the way ESPN scores everyone else's: The most confident pick is worth ten points, the least confident is worth one, and a wrong pick is worth nothing. Three other ballots play the same ten games every week: Vegas takes the favorite and ranks by the size of the spread, the Crowd takes the side most ESPN entries picked, and the Writers take the higher-ranked team in the AP poll. Essay one explains the game, the four ballots, and what the numbers say it takes to win. As the weeks come in this page will grow, and analytics will be added.