Rushing yards projections for the running backs on this week's slate, ranked. Each number is the model's median: the back should finish above it about half the time and below it about half the time.
Last week's graded record — wins and losses — and the plays for the week ahead. One email, no daily noise, unsubscribe in a click.
Every player on this week's slate with the model's median projection, the book line once one is posted and the side it falls on, rebuilt daily with inactives removed. See plans, or read how the projections are built below.
See plans →An XGBoost model is trained on past NFL running back games and asked for the median rushing yards in the next one. It reads the back's carries and rushing yards over his last three, six and ten games, using prior games only, and the rushing yards his opponent has allowed over its recent games. It is retrained on every run, so each week's results feed the next projection.
No sportsbook line is an input. The model never sees the market, which is what makes the comparison with a posted line meaningful: when the two disagree, they disagree for independent reasons.
Rushing yards are one of the noisiest markets in football. A back's recent average on its own predicts his next game worse than simply guessing the league average does, because game script decides so much: a team that falls behind early stops running, and a single 50-yard carry can make a quiet afternoon look like a big one.
That is why this page publishes a median rather than a pick. A back projected for 60 yards finishing anywhere from 30 to 100 is an ordinary week. The projection is most useful where it sits well away from the posted line, not where it is a yard or two either side.
Where a sportsbook line has been logged for a player, the board shows it with the side the projection falls on. Books post player lines late in the week, so most rows show a dash until close to kickoff. These are reference projections, not graded plays: in the NFL only the moneyline is graded. How the sides have fared against real lines still shows on the performance page, wins and losses alike.
Players ruled out or placed on injured reserve disappear from the board when you load the page. A backup's projection still reflects his old role, though, so when a starter is ruled out, adjust his replacement upward yourself. The NFL injury report lists who is out, game by game.
An XGBoost model trained on past running back games projects the median rushing yards for each back. It uses his carries and rushing yards over his last three, six and ten games, prior games only, and the rushing yards his opponent has allowed recently. Sportsbook lines are never an input.
It is the number the back should beat about half the time. A 65-yard projection is not a promise of 65 yards; it says the over and the under at 65 are roughly a coin flip. The useful signal is a book line several yards away from the projection.
Players ruled out or on injured reserve are removed when the page loads, and a back needs a few games of recent usage before the model projects him. Quarterback rushing yards are not covered. Backups promoted by an injury keep their old projection, so adjust those yourself.
No. The board is for members, updated daily through the week. Members see every back, along with the graded NFL moneyline plays and the MLB, NBA and NHL boards. These are reference projections, not graded plays.
NFL moneyline picks every week, plus MLB, NBA and NHL picks, player props and live trackers. Every play is graded in public, wins and losses alike. See plans to get started.