Watchlist

How the score is built

Projected closeness — 35 points

Decays exponentially in the model’s projected margin: a pick’em is the full 35, a touchdown favorite about half, three scores nothing. The largest term, because a blowout is a blowout no matter who is playing.

Team quality — 30 points

Elo percentile of both sides, with two thirds of the weight on the weaker team. A top-five side hosting a bad one is not a good game, and an average of the two would claim it was.

Stakes — 15 points

Poll position going into the week, plus a share for a conference game. Two ranked teams take most of it; the higher the pair, the more.

Rivalry history — 10 points

From the all-time series: a named trophy game, the length of the series, and whether the all-time ledger has stayed close.

Points on the board — 10 points

The projected total, from the drive-sim. A 31-28 is better television than a 10-7 and that is allowed to matter a little — but only a little, or the whole list becomes bad defenses.

There is no term for brand, television window or network. A watchability score that rewards blue bloods for existing is a popularity ranking wearing a lab coat.