Weekly Rankings — 2026, week 6

Projections come from player stats retrieved 2026-10-14, covering play through week 5 of 2026; roster facts beside them — who was cut, who is on a reserve list, who changed team — come from rosters retrieved 2026-10-14.

The bridge to the game model is game script: projected margin and total decide how many plays each team gets and of what kind. A receiver on a 26% target share in a game projected 17–27 down is worth more than in one projected 28–17 up, because the trailing team throws more.

Where the model departs from recent form

The bar is the weighted average of a player’s last six games, which anyone can compute in two minutes. Above, the six that game script and matchup favor most this week; below, the six they penalize. If this chart had bars in only one direction, the model would be adding nothing over the baseline.

What to expect from this

What is published is a blend of the model and the player’s recent form, and that is deliberate: measured out of sample, the model on its own does not beat a weighted average of the last six games at any position. The blend does — on error at all four, and on rank correlation at three of the four. Quarterback is the exception and it is stated plainly: the blend lowers the error but orders slightly worse than the simple average. Weekly fantasy is mostly noise; anyone promising more than this has not measured it. The Matchup column is the opponent-defense adjustment, corrected for the quality of offenses it has faced and damped to 45%: a real signal, far smaller than it is usually sold as.

The largest correction in development was the quarterback: a team’s pass attempts are not its quarterback’s, and without discounting sacks and scrambles the position came out 28% high.