Fastest moving: how the list is made

How the fastest moving lists count filings over the last seven complete days against the thirteen weeks before, and which entities are ranked.

For every district, case type, matter and company, fund or insurer: the count of filings over the last seven complete days, and the thirteen windows of seven days before those seven days, which give the weekly mean. The difference is the count minus the mean, and the percent change is the difference over the mean.

Lists rank by percent change, highest first, and print both numbers and both windows on each row. Counts come from the record; none is an estimate. District counts are calibrated to the official annual counts (each district's observed dockets times its ratio of the Administrative Office's newest twelve month count to the dockets observed in those months), which leaves the percent unchanged.

An entity is ranked only with twenty or more dockets over the ninety-eight days, a thirteen week mean of at least ten a week, and a seven day count at least ten dockets above that mean, so a small base, or a small rise over one, never tops the list. Entities with no dockets in the thirteen weeks before go to New activity, never the ranking. Judges are not ranked here.

The record is the CourtListener docket archive until its last complete day (its late tail is trimmed by weekday medians), plus later days the courts' daily case lists cover completely, plus, for districts, the weeks the district walk has listed completely from the CourtListener search feed. The seven days ranked are the newest fully listed days; each window is counted exactly from complete days and fully listed walk windows, never spread or estimated. A district whose listing of those windows is not yet complete is left out of that week's list and named on the page. The bars on a row in the table are the fourteen windows, the ranked week last; a calendar week in progress, when the record reaches yesterday, is drawn as an estimate (see the projection method) and never ranked.

An earlier ranking method was retired; the methodology history says why.