How the Greediness Score rates a filing: bands, factors, and limits

The Greediness Score is a 0 to 100 rating of how much structural conviction a public filing shows. It reads the things a document can actually establish - who traded, how unusual the trade was for that person, how quickly it was disclosed, and who traded alongside them - and turns them into a single comparable number so you can decide which filings to open first.

It is a prioritization aid, not a forecast. Our own historical validation has not demonstrated predictive power, and we say so on every page rather than implying an edge we cannot support.

Bands exist so you can triage a feed without staring at raw numbers. A Strong score is a reason to open the primary document, not a reason to trade. Two Strong filings can tell very different stories once you read the codes and footnotes.

A Form 4 insider purchase, a congressional periodic transaction report, and a Schedule 13D activist stake are different documents with different tells. Scoring them on one shared checklist would flatten what makes each informative, so each type gets its own calibrated factor set.

Form 4 filings are scored on insider role, cluster activity, trade size versus that person's own history, position impact, filing urgency, and direct ownership. Congressional filings are scored on disclosure timeliness, trade size, committee relevance, and chamber context. Schedule 13D filings are scored on stake size, filing urgency, accumulation velocity, and filer history.

Factors need data. A first-time filer has no trade history to compare against; a filing that omits a field cannot be scored on it. When a factor's input is missing, that factor is excluded and its weight is redistributed across the others.

Scores built on thin data are capped and labeled low confidence. A number assembled from missing inputs should not look as sure as one where every factor had something to read. When you see the low-confidence label, treat the score as a weak prior and lean harder on the primary document.

The default weights are one opinion about what matters. The Greediness tab shows every factor, its default weight, and worked example filings computed step by step, and it lets you change any weight to match your own thesis and apply it to the whole feed on your device.

If you think cluster activity should outweigh filing urgency, say so and watch the feed re-rank. The point is a model you can inspect and argue with, not a black box.

The score ranks the row in front of you. It does not re-label the transaction code. In the Strong-band export from 29 September 2026, Scorpio Tankers' chief executive and president each appeared as a buy of 240,495 shares near $81.16, scored 84 and 86, with the explanation citing a senior role and a seven-insider cluster. The Form 4s are code A awards under an equity plan, and most of the shares reported afterward were unvested. The same export's Total Return Securities row, 9,000 shares at $5.77, scored 80 with high confidence for role and a stake change, and that Form 4 is code P.

Use the band to choose what to read. Then open the filing: code, security title, footnote, and holdings after the trade. High confidence means the factors had data. It does not mean the row is an open-market purchase.

The score does not predict returns, does not infer motive, and does not prove anyone traded on non-public information. It describes filing structure. For the factors that most often drive a Form 4 score, see the cluster buy and direct-versus-indirect guides; for the buy-versus-sale asymmetry, see the companion explainer. Then open the live feed at /app?band=strong and verify the primary document.