Fantasy football, wrapped weekly.
Nobody has this figured out, least of all me. Everything here shows its work so you can disagree with it properly.
Davis Allen played 69% of the Rams' snaps in week 7 against a 43% average, and his target share went from 3% to 17%, with no teammate out. The rest of the week's role moves, the work that hasn't paid yet, and the heaviest loads are below.
Read the pieceBucky Irving went from 55% of Tampa Bay's snaps to 85% in week 6 with no teammate out, and turned 22 touches and targets into 5.9 points. The rest of the week's role moves, the work that hasn't paid yet, and the heaviest loads are below.
Nobody in Jacksonville's backfield missed week 5, and Bhayshul Tuten's carry share still went from 56% to 81%. The rest of the week's role moves, the work that hasn't paid yet, and the heaviest loads are below.
Kyle Monangai took 30 carries and 54% of Chicago's snaps in week 4 while D'Andre Swift's carry share fell from 52% to 28%, and no one was out. The rest of the week's role moves, the work that hasn't paid yet, and the heaviest loads are below.
The model doesn't start with points. It starts with who is getting the work, how that has moved over the last few weeks, and how little it should trust the yards that came with it. Here is that layer, the idea I tested and left out, and what to do with it on a Sunday morning.
Fantasy points are the last thing that happens, not the first. The numbers I lean on, the ones I only read with context, and the ones that get you into trouble when you read them alone.
Most trade offers die because nobody attached a reason to them. The lowball isn't the problem — the silence after it is, and the manager trying to win the trade is the reason the darts get thrown in the first place.
Every stat here either carries its source or is labelled a placeholder. There is no third category. A confident number you cannot check is worth less than no number at all, because it costs you a roster move before you find out.
Takes get revisited when they turn out to be wrong, and the revision says it was a revision. Quietly editing a bad call is the same category of dishonesty as inventing a stat.
Rankings arrive with their tiers, their scoring format and how far each player moved. Calculators explain their formula instead of hiding it, so you can argue with the method rather than guess at it.