[ Docs — reading signals ]

Reading signals

Every card in the feed answers three questions: what happened, how much it should matter, and whether you're early. The class tells you where in the chain the future leaked; the score and fields tell you the rest.

The five classes

Confirmed

A signed award on the public record — contract awards, subawards, federal grants. The fact is certain; the only question is whether the market has priced it yet, which is what the timing field answers.

Predicted

An award that doesn't exist yet: a contract nearing expiration or an open solicitation, with the likely winner named and a stated win probability. Wrong sometimes, by design — the probability is the honesty.

Budget

Appropriated money that hasn't been spent. Budget lines and program funding traced forward to the incumbents positioned to receive them — the earliest, slowest-burning class.

Press

The company said it first — an 8-K or press release announcing a contract. We verify the claim against the federal record and score the gap between what was said and what's filed.

Market

The tape already believes something — unusual options activity or a sentiment spike around a ticker with federal exposure. Read it as confirmation or as a warning that the edge has faded.

The fields on a card

FieldHow to read it
Score0–100, one number for how much this should matter. Built from materiality, historical analogues, and regime. 72+ is the high-conviction bucket — see the track record for why exactly there.
TimingHow fresh the read is: same-day, or flagged "likely priced in" when the market has had time. A stale flag on a big award is the signal doing its job.
MaterialityThe award's size against the company's market cap and revenue — a $50M award is noise for a prime, transformative for a microcap.
ActionThe model's read, in plain words: buy watch, monitor, or fade. It is a research label, not a recommendation.
Win probabilityPredicted-class only: the model's stated probability that the named company wins the award.
Score breakdownTerminal only: each component's contribution to the final score, so you can see exactly where the model's conviction comes from — and where to disagree.

Where the numbers come from — and the realized results behind them — is covered in Methodology.