Model receipts
How Ball Ranks tests its fantasy football projections
Two football models, thirty-eight season-format tests across fifteen NFL seasons. What the board publishes, how we grade it, and where it has lost.
Ball Ranks started as a fantasy basketball board with one rule: no ranking change ships until it has been replayed against completed seasons.
The same rule applies to fantasy football. The 2026 football draft board is live, with a page for every NFL player. This post explains what the board shows, how we test it, and where it has lost.
Football is a different game, so the math is different
Our basketball board ranks players across nine scoring categories. Football leagues usually use points scoring, where your team's week is one number. The football board publishes three kinds of information:
- Projected season fantasy points under standard, half PPR, and full PPR scoring.
- Value over a replacement-level starter at the same position. A quarterback's 300 points and a tight end's 220 are not the same draft asset. The "VAL" column makes that positional context visible.
- The room's opinion next to ours. Live average draft position from real drafts makes every disagreement easy to see.
The two models
Model Zero is the free baseline model. Every release is graded against it.
Model One is the latest release. It earned that position by reducing the average size of a points miss against Model Zero in 37 of 38 available season-and-scoring-format tests in the current record.
That creates a direct comparison between each model's projections and the final results.
The backtests — all thirty-eight of them
The current record covers every draft year from 2011 through 2025 in standard and PPR, plus 2018 through 2025 in half PPR, where the comparison archive begins. Each replay is time-safe: it uses the version of the board that could have existed before that season's opening kickoff. We score the top 150 against what actually happened and measure the average size of a points miss. Lower is better.
The fifteen-season view overturned part of the original three-season story. A launch claim based on nine replays did not survive the larger sample. We removed that claim and reran the full record. Today's Model One board placed players closer to their actual value-over-replacement finish than a pure-ADP comparison board in 34 of 38 shared-pool replays.
On points accuracy, Model One beats Model Zero in 37 of 38 available season-format replays, with a 7% aggregate reduction in the average miss — against a baseline that itself got stronger in September 2026, when we rebuilt it as a real projection.
We publish losses for the same reason we publish them on the basketball side: a track record should include the bad seasons as well as the good ones.
What this board is not (yet)
No projection is a guarantee. What you get today is a board tested across fifteen completed seasons, with projected points, positional value, confidence, player context, and live ADP.
In-season is where the roadmap points next: weekly rest-of-season ranks and waiver tools that stay useful long after draft night.
Where the data comes from
NFL statistics come from the open nflverse data project. ADP comes from FantasyFootballCalculator's public API of real drafts. Depth charts are the league's published charts, not a stat-derived guess. Ball Ranks is independent and unaffiliated with the NFL.
Draft with receipts: open the 2026 board.