MEASURED–DISCLOSED
FANTASY POINTS
We combine the five props into a fantasy total, keeping the correlations between them rather than adding up five separate guesses. Below is what that total is worth knowing — and, at more length than you will be used to, what it is not.
SCORING FORMAT
1× receptions · 0.1× receiving yards · 0.1× rushing yards · 6× touchdowns · targets and carries score 0 — they are inputs, not score components
Full PPR is the format we measured against a bar set in advance. Its accuracy result — beating both fantasy baselines — was tested against an acceptance frozen before we saw the answer. Its range was not; see below.
MEASURED–DISCLOSED
WHAT IT DID ON 2025
AVERAGE MISS
4.42fantasy points per player-week — 13.5% better than his last-four average and 12.1% better than his season average, over 3,753 player-weeks.
OUR 80% FANTASY RANGE CAUGHT
75.7%of outcomes in 2025, against a nominal 80%. We are reporting what it caught, not claiming it hit a target — no target was ever set for this range.
OUR 50% FANTASY RANGE CAUGHT
49.4%against a nominal 50% — so this band runs slightly narrow too, not wide.
loading this week…
Which weeks this covers. Fantasy covers weeks 2–17; the props archive covers 1–17. Week 1 has no fantasy line because the composition needs prior-week usage to draw from, and week 1 has none. Compare against the props archive →
NOT CLAIMED
WHAT THIS NUMBER IS NOT
Five things, all of them real, none of them in small print. A composed number hides its assumptions better than a single projection does, so these are stated at full size.
1
This does not cover quarterbacks
A QB fantasy score needs passing yards and passing touchdowns, and we model no QB passing anywhere in this project. So a quarterback’s fantasy points cannot be computed at all — not estimated badly, not computed. They are absent rather than quietly left out, and we would rather tell you the surface has a hole than let you assume it does not.
2
Fumbles are not modelled
Standard scoring subtracts 2 points for a fumble lost. We do not model fumbles, so every number here is biased slightly high — by exactly the fumbles that would have cost points.
3
One dependence structure for every player, and it is imperfect
Fantasy points combine several correlated outcomes, and we use a single pooled correlation structure for everyone. Measured on training data it varies materially by player type: the link between receiving yards and touchdowns runs 0.116 to 0.294 across volume terciles, and between rushing yards and touchdowns 0.342 to 0.054 across depth-of-target terciles. The range is least well matched at the extremes — the highest-volume and most unusual players.
4
Components we did not price count as zero
If we published no rushing projection for a receiver, his rushing yards enter the total as zero rather than as an estimate. On blind data, 4.0% of receiving-only weeks had any actual rushing yards, averaging 0.9 yards — about 0.1 of a point. Small in expectation, real in the tail: the largest was 151 yards.
5
Both ranges come out slightly narrow
The 80% range caught 75.7% of outcomes and the 50% range caught 49.4% — both a little under what their labels suggest. Real weeks land outside these ranges somewhat more often than the numbers imply, so treat the edges as softer than they look. Under contract v1.2 we described the narrower band as running wide; the corrected measurement shows the opposite, and this is the figure of record.
HOW IT IS BUILT
The five props are drawn together from one frozen pool that carries how they move with each other, then scored. Changing format rescores the same draws — it never refits anything, so the three formats are the same underlying week seen through different coefficients.
pool ebb90a7f78582052 · 15,028 draws · 557 players
covers WR / TE / RB only
tier MEASURED–DISCLOSED · contract v1.2
The five underlying props, and the intervals that were measured against a bar set in advance: the receipts →