Winners of the four ranking criteria. Empty only when the grid produced no reports.
Grid point reports, sorted descending by the schema's reportOrder criterion (default sharpe).
Author tracks — one line per (rule x author), deduplicated by grading rule (hold x lock x stop x trailing). This is the RAW continuous track (ideas/hits/hitRate), not a 0/1 ban verdict: the engine grades, userspace decides who to trust. Far more compact than repeating the track on every point's report, and every line is self-contained (carries hold/lock/stop/trailing/author) for grep/jq without a join.
The single report bucket of a run: every grid point graded by the one profit-before-stop metric, its ranking winners, and its author TRACKS. One bucket — there is no per-metric split.