We read the market’s whole forecast, and show our work.

Prediction markets and sportsbooks put prices on players and games. Most sites show one line from them. This one shows the whole distribution the market prices, turns it into fantasy points under your own scoring, and publishes a record of how every forecast did. When an idea for beating the market is tested, it is registered and the result is published either way; the register says how many have been tested and what each found.

• The whole distribution rather than a single number, wherever the exchange lists a ladder for that player.
• The full record for every player with usage in the exported seasons, whether or not a market exists for them.
• A weekly score of the site’s own published probabilities against what actually happened.

1. 01Facts →WHAT HAPPENEDSource rows as published. A fact may be corrected, and a correction arrives as a new row rather than an edit in place.
2. 02Features →WHAT IS MEASUREDDerived from the facts layer and from nothing else, so every feature can be rebuilt from the record that produced it.
3. 03Beliefs →WHAT IS EXPECTEDDistributions, never scalars. A median is read off the distribution rather than fitted beside it.
4. 04DecisionsWHAT GETS PUBLISHEDImmutable and timestamped, so what a model said at the time survives being wrong.

Method states how a number gets onto a page and what has to be true for it to stay there.

• Every hypothesis published, whichever way it landsEach carries the date it entered the register, so one entered after its first result says so. The register counts what each found.
• No lean removed after the factThe ledger is append-only and hash-chained: the chain holds over 2,540 rows and matches all 4 heads published in the weekly board indexes. Receipts lists every lean.
• No rate without its intervaland the sample it rests on, printed beside it. A bare rate is the one figure that always looks certain.
• Every price names its sourceon the page. Where a figure combines books, the page names them and the rule: the median of de-vigged closes across three named books.

• Recommend a side where money is at stake. The record is published so it can be read against a price; it is not advice, and nothing here is a bet.
• Publish a rate without its interval and the sample it rests on. A bare rate is the one figure on this site that always looks certain.
• Fill a gap with a plausible number. A section without its data is marked, not populated.
• Blend sources without saying how. Where a figure combines books, the page names them and the rule (the median of de-vigged closes across three named books).

And one thing the site commits to rather than refuses: any predictor it publishes ships with its scoring record visible. The one built so far is scored in Research, where the verdict is computed from the file that holds it and carries the interval it rests on. It is never stated here, where nothing could check it.

Names and a contact address wait for the people who make the site to confirm how they want to be listed.