Redzone Metrics NFL picks

NFL picks from a real statistical model
Every game modeled.
Every line priced.
Every pick graded.

Moneyline picks where the model and the market disagree, ranked by edge. Every week is predicted by a model that has only seen the games before it, and every pick is graded on the final score.

01/ 05This week

Thepicks

Every game this week, ranked by edge against the market, with live scores while they are on and every pick graded on the final.

Loading the board…
02/ 05Record

Graded inpublic

Each week is predicted by a model that has only seen the games before it, so this is the record a reader would actually have had, at one unit flat per pick.

Loading the record…
03/ 05Props

Playerprops

DraftKings' line for every prop this week, how often each player has cleared it, and the model's projection.

04/ 05How it works

How picksare found

A team-strength model fitted only on past games, the market's price, and the gap between the two.

01
Pre-game features

Margin-of-victory Elo, weighted offense and defense EPA per play (overall, passing and rushing), point differential, turnover margin, the starting quarterback's own passing EPA, rest and venue.

02
Fitted only on the past

Each week is predicted by a logistic regression fitted only on the games before it, so a graded week is the record a reader would actually have had, not a refit.

03
Blended with the market

The number behind a pick is 70% model and 30% the market's no-vig probability. The edge is that blend minus the market.

04
Tiers by edge

A pick needs 2 points of edge: Low from 2, Medium from 4, High from 6. Anything under 2 is a pass, and every pick is graded on the final score at one unit.

05/ 05Redzone

Beyond thepicks

Team, player and red zone analytics on redzonemetrics.com.