A public, timestamped, append-only forecast record open to anyone in the Federation. Falsifiable claims only. No price predictions. Scored when they resolve, and never deleted — including the wrong ones. Especially the wrong ones.
Anyone can claim they called it. Almost nobody publishes the calls they got wrong. The Federation Record exists to make the second thing normal.
You submit a claim about the future that can only turn out true or false. If it is admissible, it is published with a timestamp, cryptographically anchored so the timestamp cannot be backdated, and left alone. When the resolution date arrives it is scored against what actually happened. It is never edited and never removed. Your standing is the whole record, wins and losses together.
This is the same discipline Dossier applies to its own forecasts at dossiertrack.co/forecasts, opened up to the Federation. Dossier’s record is the product’s. This one is yours.
A track record that can be edited is not a track record. Anchoring means anyone can verify a forecast existed before its outcome was known, without trusting us. Append-only means a bad call cannot quietly disappear. Those two properties are the entire product.
A submission needs all four of these. Most first submissions fail on the first one, and that is expected — writing a genuinely falsifiable claim is a skill, and it is most of the skill.
Things that qualify. Whether a protocol ships a named feature by a date. Whether a bill clears a chamber. Whether an on-chain metric crosses a threshold in a stated window. Whether a named project’s TVL sits above or below a figure on a date. Whether an audit is published. Whether a listing happens.
These are hard exclusions, inherited from the Dossier forecast specification. They are not judgement calls and they are not appealable.
Submissions that do not qualify are declined with the reason stated publicly, without naming the submitter. The reasons are the point. A record where everything gets in would tell you nothing about the people in it.
Probabilistic forecasts are scored with a Brier score: the square of the gap between what you said and what happened. Say 80% and it happens, you score 0.04. Say 80% and it does not, you score 0.64. Lower is better, zero is perfect, and 0.25 is what a coin flip earns. Plain yes/no claims are scored as correct or incorrect and counted separately.
Standing is reported three ways once there is enough volume to mean anything: your Brier average across resolved forecasts, your count resolved, and your calibration — whether the things you called at 70% actually happen about 70% of the time. Calibration is the hard one and the only one worth boasting about.
The standings table below opens once anyone reaches five resolved forecasts. Ranking on one lucky call is not a track record.
Standing in this record is a public reputation for being right, held by you, verifiable by anyone, and portable to wherever else you want to point at it.
Standing in the Federation Record does not convert to tokens, airdrop eligibility, governance weight, Founding Member NFT utility, Dossier discounts, or any other economic benefit — now or later. There is no points balance, no redemption, and no plan for one. Participating is not an investment and earns nothing but the record itself. This is a permanent condition of the Record, not a current phase of it.
That rule exists for two reasons. It is the honest description of what this is. And a reputation system that pays out is a reputation system people farm — which would destroy the only thing that makes the record worth reading.
Submissions are reviewed by hand for admissibility. Expect a few days, and expect to be declined the first time or two. Nothing you submit is published until it has been reviewed.
The Record opened on 2026-07-27. The first admitted forecast will appear here, with its timestamp and anchor. Nothing has been backfilled and nothing will be.
| HANDLE | RESOLVED | BRIER | CALIBRATION |
|---|---|---|---|
| Standings open when a participant reaches five resolved forecasts. | |||
If it is full of correct calls, be suspicious — ask what was excluded. If it contains plainly wrong calls, still published, still attributed, then it is doing what it was built for.