Section membership is rebuilt from fields on the closed row (trust_tier, grade, score, R:R, paper flag, forward sample), so it is a proxy for the live gate — not the gate’s own decision. A class inside a section can carry edge even when the section aggregate does not, which is exactly why nothing here is deleted for being unproven.
Family bucketing is name-based (strategy / source_system). It answers “does this whole surface carry edge?”, not “is this individual trader good?”. Per-trader reverse engineering needs its own ledger.
Loading registry…
Prior backtests are evidence, not verdicts. Without a trade log or a per-symbol price series they cannot be re-scored against a timing null, so nothing on this tab can promote a strategy. The tribunal on the other tabs scores live/forward rows only; this registry exists so old work is never thrown away for being inconvenient.
Row count and fingerprint come from reports/forward_observations.jsonl
as read by the tribunal's forward_gate_block(). A pod at zero rows
cannot promote anything, no matter how good its backtest looks — that is the
point of the gate, not a bug in it.
loading…
| symbol | rule | entry | SL | TP | status | net % |
|---|