First-party analysis · v0-fixture

How the desk computes a week

This is a working sketch, not a finished or backtested model. Weights stay labeled placeholders until someone calibrates them. The stamp is methodology v0, subject to change.

The job is narrow. Produce comparable weekly fantasy point estimates so start/sit, rankings, and add/drop are driven by numbers first. A later explain layer may restate those numbers. It may not invent a status, a return, or a stat line that was not supplied.

Inputs

• Trailing weekly counting stats stored as structured rows, not prose.
• Derived usage shares: snap, target, carry, route.
• Listed injury or participation status. The model never guesses a designation.
• Bounded matchup factor, cap ±0.08.
• Scoring format: PPR, half-PPR, or standard.

The mean

1. Base rate: decaying average of trailing format-adjusted fantasy points. Current decay weights, most recent first: 1, 0.7, 0.5.
2. Usage overlay: current versus prior usage, capped at ±0.15.
3. Availability gate: OUT, IR, or INACTIVE forces the projection to 0. Questionable and doubtful apply soft discounts. The model does not invent a return.
4. Matchup factor: a bounded multiplier so one heuristic cannot run the desk.
5. Uncertainty: wider floor and ceiling when the sample is thin or the status is dirty.

Mean is the ranking number. Floor and ceiling are a model range, not a promise. All three sit on player and comparison pages.

Certainty score

Start/sit pages also print a 0 to 96 certainty score. It is not a probability of winning a week. It is a desk grade of how much the call can lean on the math in front of it.

• A wider mean gap raises the score.
• A toss-up (under 1.5 estimated points) keeps it thin.
• High uncertainty, a status discount, or an availability zero cuts the score.
• Three trailing weeks on both sides adds a little confidence.

Bands: thin, lean, clear, strong. Read them next to the status labels, not instead of them.

Waiver radar

The waiver board adds an urgency tag (hot, rising, stash, fade) from the same estimates plus snap-share change and listed status. It is not a claim that a name is available on any host platform.

Format

• Standard: yards and touchdowns only.
• Half-PPR: +0.5 per reception.
• Full PPR: +1.0 per reception (provisional default, NEED JOSHUA INPUT).

What stays distinct on the page

• Reported status: the listed designation with a last-verified time.
• Model projection: estimated fantasy points, not official.
• Editorial note: short desk copy. It may only restate computed fields.

What stays blocked until licensed data

• Live injury zeros presented as official current-week availability.
• Indexable production rankings.
• Public backtested accuracy claims.

Licensed ingest waits on Joshua approving API spend. See the independent disclaimer and the about page.