Methodology

1. Consensus is the spine

There is no projection engine here, deliberately. Markets are driven by vibes, and a board that quietly “improves” on consensus into a tier of bad recommendations is worse than no board at all — flexibility that produces awful picks in some corner of the settings is not flexibility, it is a liability wearing a settings panel.

So the board you see is the published expert consensus board for your format, tiers included, exactly where the analysts put them. FantasyPros publishes five: standard, half-PPR, full PPR, superflex, and half-PPR superflex, each built from more than a hundred analysts. Your settings pick one.

2. What your settings are allowed to change

Within a position, order never changes.Two receivers keep the order the market put them in, whatever your league does. The market’s opinion of two receivers is better than anything computed here.

Across positions, interleaving responds to your roster. Whether the 24th receiver goes before or after the 18th running back genuinely depends on how many of each your league starts, and a published board cannot personalise that for a 14-team superflex league. That shift is a single bounded term:

offset(pos) = 14 × ln( demand_baseline(pos) / demand_yourLeague(pos) )
adjusted    = clamp(consensusRank + offset, ±18 ranks)

It has three properties that are asserted in the build and fail it on violation. It reproduces consensus exactly when your league matches what the board assumes — the ratio is one, the logarithm is zero, the offset vanishes. It is monotone: more demand for a position can only move it up. And it is bounded at 18 ranks, so no combination of settings can manufacture a garbage tier. The sheet tells you, above the board, which board it snapped to and how far your settings pulled away from it.

One assumption is stated rather than buried: a FLEX slot is spent 35% on running backs, 50% on receivers and 15% on tight ends; a superflex slot 75% on quarterbacks. Those are allocation assumptions, not a model.

3. Four markets, compared within position

ADP comes from Yahoo, ESPN, Sleeper and an aggregate of public mock drafts. Each row carries one cell per platform: the number is where that platform actually drafts him, and the tint is whether that is a bargain or a reach against the expert consensus.

The comparison is made within position, and it has to be.The first version differenced each platform’s overall rank against the overall consensus rank, and it painted almost every quarterback as a screaming reach on Sleeper — median gap of eighteen places. That was not a market edge, it was Sleeper’s own ordering: search_rank surfaces quarterbacks far higher than a PPR consensus board ranks them. Comparing within position cancels that bias entirely (median gap: one place) and is the more useful question anyway, since you are choosing between running backs, not against the whole board.

Color thresholds scale with positional depth. Three places between running backs is a great deal at RB5 and noise at RB50, and a flat cutoff lit up a third of the board at full strength. About seven per cent of cells now carry a strong tint.

Two source traps are handled explicitly. ESPN’s ADP saturates: roughly 720 players share a value near 170, which is a clamp rather than a market signal, so it is discarded past that point. And Sleeper publishes no ADP at all — its cell is where Sleeper ranks a player within his position, shown in italics and marked with an asterisk, never presented as a pick number.

Joining four platforms needs a player-id crosswalk, because naive name matching fails on about 15% of the pool — every suffix case and every team defense. Team defenses have no id anywhere, so they are joined on team code.

4. Offseason movement

The arrow at the end of each row is thirty days of ADP movement: up means being drafted earlier than a month ago. It replaced a per-row sparkline, which was honest but cost 28px on every row to answer a question with three possible answers — a bad trade on a board whose whole argument is density.

Thirty days is long enough that one noisy day cannot flip an arrow and short enough to still be news on draft weekend; the neutral band is six picks, against a median move of about five. The underlying series only becomes genuinely daily from around July — measured across the top 150, just 8% of players have any data point by March 1 and 29% by June 1, against 89% by July 1 — so a player with too little history gets no arrow rather than a fabricated flat one.

5. The team arrows — roster plus coaching

Each arrow blends two inputs, and they are kept separate on purpose. Hovering an arrow shows the split.

Roster is derived. It compares the current consensus value of the players a team added against the players it lost at that position. A running back who left is priced at what the market thinks of him today, on his new team. That is reproducible arithmetic.

Coaching is authored, and weighted heavily.A new play-caller can matter more to a position’s production than any single signing — a pass-first coordinator can lift a whole receiving corps without the roster changing at all, and a run-heavy hire can quietly end a receiver’s season as a WR2. A sheet that stayed silent on coaching would be missing most of the story, so it is not a footnote here.

Coaching cannot be derived. The one free structured source for 2026 coaching is partially stale — it catches roughly six of the ten head-coach changes, still lists departed coaches and misspells others — and nothing free covers offensive coordinators, where about twenty teams changed. So each entry is hand-authored, carries its source, and is validated at build time. A team with no entry contributes zero to the arrow rather than a guess.

The impact score is an editorial judgement in the range −2 to +2, informed by reported scheme and public expectation, and it is capped: one full step is worth 35 value points against thresholds of 28 and 90, so coaching can carry a position across one boundary but can never manufacture a big move that the roster contradicts. Where a hire is good news for one position and bad for another — a pass-first coordinator is not equally good news for a running back — the score is set per position.

Coaching is only written up where it plausibly moves production — ten teams, one sentence each. An earlier version led all thirty-two cards with a head coach, a coordinator, a scheme lineage and a paragraph, which buried the players. Every team still carries its staff behind the scenes so the arrows stay honest; the card just does not spend your attention on a coordinator swap that changes nothing.

The three letter grades on each card are derived and each says exactly one thing. Offenseis the consensus value of that team’s projected skill starters — only the starters, since a third running back does not make an offense better. Defense is where the market ranks that team defense. Schedule is the average strength of the defenses that offense has to face, so an A is the easiest schedule. All three are graded against the other 31 teams, so a C means mid-league rather than bad, and none of them is a power ranking.

What none of it can see: scheme fit for a specific player, a return from injury, a second-year leap, or a rookie the market has not priced yet.

6. The Top 200, and what a blurb is

Every blurb on the player page is composed from a signal that exists in the data: a change of team, a new play-caller, how far the market has moved him in thirty days, whether the expert panel is unusually split on him, his depth-chart spot, his injury designation. Nothing is borrowed from someone else’s analysis and nothing is invented. Where a player genuinely has no news, it says so rather than manufacturing a story.

That reads thinner than a human analyst would write, and it should. What it buys is coverage: two hundred players, all current as of the last rebuild, with no stale July takes hiding among them.

Upside and worst case are the genuinely sourced part.They are not a model’s spread — they are the most and least optimistic of the 100+ experts on the consensus panel, and the earliest and latest a player has actually been taken in public mock drafts. Those are real people’s real opinions and real picks.

Injury tags come from Sleeper’s designation. The tooltip leads with the injury itself rather than the label, because Sleeper lists a player who has had ACL surgery as “Questionable” and reading that first would mislead you. Sleeper publishes no return date, so none is invented: the text says what the designation guarantees under the roster rules and stops there.

7. Your data

There are no accounts. Your settings, stars and removed players live in your own browser’s storage and are never sent anywhere. That is the right trade for a thing you use once a year in a basement on bad wifi — but it means clearing site data, or opening the sheet on a different device, starts fresh. The “Share settings” button encodes your league setup into a link so you can carry it across devices.

Removed players drop out before replacement level is computed, so marking twenty keepers genuinely moves where each position runs dry. That is the one thing a printed sheet from the internet can never do for you.

8. Sources

• Expert consensus rankings and tiers — FantasyPros.
• ADP — Fantasy Football Calculator (free for personal and commercial use with attribution), Yahoo Fantasy, and ESPN Fantasy public endpoints.
• Player metadata and search rank — Sleeper.
• Cross-platform player id crosswalk — DynastyProcess.
• Rosters, draft picks and team marks — nflverse. Marks served by ESPN’s CDN.