Show the decision clearly. Preserve the boundary underneath it.

Weekly Fantasy Rankings · Beta v0.1

One football projection, three fantasy scoring systems

The Weekly Fantasy Rankings estimate a player-level football stat line for the selected NFL week, conditional on the player being active. That same projected football state is then scored under Standard, Half-PPR, and PPR rules.

This keeps the football expectation consistent across formats. Changing the scoring setting changes how the accepted stat projection is scored, not the underlying player projection.

What Beta v0.1 uses

Beta v0.1 does not incorporate realized 2026 player production into the projection itself. Its player-production history comes from prior seasons, while current schedule, roster, and depth-chart information can update the accepted player-week context. Beta v0.1 remains the governing public methodology until a successor is prospectively validated and explicitly promoted.

Ranks and tiers answer different questions

A rank preserves order. A tier groups nearby players into deterministic projection clusters that are more useful for lineup decisions than pretending every adjacent rank is meaningfully different.

Beta v0.1 tiers are score-proximity clusters, not confidence probabilities. The current public product does not manufacture tier-confidence percentages or adjacent-tier probabilities that the admitted model does not yet support.

What the board shows

Each public player row can show rank, team and opponent, tier, projected fantasy points, movement when genuine prior public history exists, and an expandable Why / Basis explanation derived from the accepted ranking output.

The Why surface describes the main football-stat contributors to the projection. The Basis surface describes the admitted evidence context supporting the player state without exposing private model weights or internal pipeline machinery.

Availability and Week 1 provenance

Beta v0.1 rankings are conditional on the player being active. The board's Ranked state means the player is included in the accepted ranking population; it is not a live injury-status claim. A later methodology version may incorporate a separately governed participation probability, but the current board does not pretend that model already exists.

The first public snapshot has no legitimate prior public fantasy snapshot, so movement is displayed as unavailable rather than manufactured. Games that were already underway before a governed pre-kickoff state existed are not retrospectively presented as pregame recommendations. Publication is also blocked if any player still in the active ranking pool has already reached kickoff; in that case a new candidate must be generated before release.

Accepted Player-Week State → Football Projection → Scoring Transformation → Ranking / Tier Output → Immutable Snapshot → Website.

Fantasy Data Attribution

Open NFL data, transformed into 135 Sports rankings

Fantasy Beta v0.1 uses public data from the nflverse project for schedule, player, roster, historical-stat, and depth-chart evidence. Core nflverse-data releases used by the model are published under the Creative Commons Attribution 4.0 International license.

135 Sports modifies and transforms those source datasets into normalized player-week states, proprietary projection outputs, scoring transformations, ranking boards, tiers, and explanations. nflverse is an upstream data source and does not sponsor or endorse 135 Sports.

CFB Top 30

Ranking predictive team strength as the season develops

The 135 Sports CFB Top 30 is a versioned estimate of predictive team strength that starts from an audited launch view and updates as governed game evidence accumulates.

Rank order is deterministic within each accepted snapshot. Adjacent positions can be close, so rank numbers should not be read as equal intervals or as a claim that every adjacent separation is substantively large. No public tiers or uncertainty values are shown because those objects have not been separately validated for this product.

The AP Top 25 and US LBM Coaches Poll are shown only as public comparison points. They do not determine the 135 ranking.

Historical snapshots preserve the methodology version that produced what the public actually saw. When a successor methodology is adopted, it receives a distinct immutable snapshot rather than rewriting an earlier release.

Governed Team-Strength State → Normalized CFB Result Object → Immutable Snapshot → Website.

The browser does not consume or reconstruct protected upstream model states, scores, evidence counts, weights, parameters, formulas, candidate schedules, audit machinery, source-custody controls, or private provenance.

What the website does not do

The public website does not create rankings, resolve canonical identity, choose data rights, calculate projections, assign tiers, manufacture movement, or alter accepted 135 model artifacts. It renders accepted result objects and immutable snapshots.

Development fixtures remain explicitly labeled when used. Production ranking snapshots are admitted through an explicit non-fixture contract and real publication provenance.

Accountability over time

Published snapshots preserve the metadata needed to identify what the public actually saw and when it was published. Historical ranking states should remain inspectable rather than being silently overwritten after the fact.