Historical Action outcomes

Historical Action outcomes

Open Review historical outcomes from Actions. Select the broker account, generation window, evidence-availability cutoff and a 7, 30 or 90 calendar-day horizon. Form times are explicitly UTC. Record outcome review reads existing stored evidence and saves a separate immutable result. Opening a retained review is read-only and does not recompute it. No provider calls, backfill, portfolio changes or Action decisions occur.

What this measures

retained-action-quote-outcomes-v1 measures unadjusted observed quote-price change, not total return, realised P&L, strategy performance or a trading backtest. Splits, distributions and other corporate actions can dominate the result. An accepted Action is never assumed to be a filled trade. There is no execution simulation, turnover estimate, spread, commission or tax model.

The horizon is anchored to the retained generation’s evidence_cutoff.evaluated_at. A starting MarketPriceRow must have both its observation time and recorded/availability time at or before that cutoff. It must be within the preceding 72 hours. The starting row pins instrument, listing, provider and currency; the end quote must match all four, occur strictly after the start, and fall within the 72 hours ending at the target horizon. Its availability time must be at or before the requested evaluation cutoff. Both prices must be positive. No observation after either date is substituted; a missing/immature pair remains unavailable, not zero.

Quote change is (end / start - 1) * 100. BUY/INCREASE alignment has the same sign; SELL/REDUCE alignment reverses the sign. This is directional quote alignment, not a short-sale return or a success rate. HOLD/WATCH/REBALANCE have no supported single-instrument directional reference. Prose triggers are retained individually as not_machine_evaluable; quote movement is not a substitute for testing a trigger.

Coverage and limitations

Generation coverage separately retains recommendation, complete no-trigger and incomplete-evidence generations, original rule coverage, policy versions, profiles and evidence. Recommendation coverage uses all emitted Actions in that generation window, including proposed, superseded, accepted, rejected and ignored cases. Status at the requested cutoff comes from suitably dated decisions and generation sequence, not today’s mutable Action status. No-trigger generations are neither wins nor losses. The result shows quote coverage and the complementary share without a measurement; zero recommendations has an undefined percentage, not 100% coverage.

Current holdings and current instrument classifications do not determine eligibility. Sold and delisted instruments remain in the cohort; missing retained instrument identities or historical quotes remain in its denominator. No current alias/listing lookup repairs a missing historical identity. Coverage is conditional on retained records, with retention/survivorship and observation availability bias explicitly disclosed.

Benchmark-relative results, drawdown and FX-normalised returns are not assessable in this methodology. Existing daily bars lack immutable availability provenance, benchmark histories lack durable observation identity, and spot quotes are not a complete adjusted series. The reader does not use these incomplete sources or convert unlike currencies to manufacture a return.

Historical policy replay eligibility is a distinct mode requiring an explicit audit reason. It records unevaluable because the full historical executable-policy/input-availability contract is not retained. No policy replay is performed, and current policy or data is never substituted. An exact historical replay engine, adjusted outcome series, benchmark persistence and governed backfill remain future extensions, not implemented claims.