SIG
Stock Directional Forecaster (Sign)Configurate Symbol
Runs the per-symbol configurator for one ticker: fetches its daily-wide derivative dataset via ORK's TDM proxy, establishes a no-feature directional- accuracy baseline, then walks every available feature method one-at-a-time and keeps the ones that strictly improve the measured accuracy. The resulting optimal feature set plus its measured accuracy is persisted as JSON so the per-symbol forecast picks it up on its next run. Single-flight per job kind — a forecast on a different symbol does NOT block this kickoff. Requires the symbol's derivative data to already exist in TDM's local store.
Forecast Symbol
Generates the directional forecast for one ticker's next NYSE trading day. Loads the symbol's saved configuration when one exists; otherwise trains on the default no-feature baseline. The forecast lands under the per-date individual forecast directory keyed by the next trading day's ISO date. Single- flight per job kind — a configurate-symbol run does NOT block this kickoff. Requires the symbol's derivative data to already exist in TDM's local store.
Configurate All Symbols
Runs the per-symbol configurator over every ticker in TDM's daily-wide derivative inventory. Sequential- path processes one symbol at a time; parallel-path uses a thread pool sized by Max workers (0 = auto-detect from CPU). Per-symbol failures are best-effort — logged + counted, do not abort the whole run. Hours of runtime; default concurrency- gate setting (off) holds the system exclusive while it runs. No upstream dependency is declared — the bulk configurator is a building block of the daily routine rather than the daily routine itself, so it can be fired against partial / ad-hoc inventories for testing without an ORK-side gate.
Forecast All Symbols
Bulk forecast routine. Pulls TDM's derivative inventory, configurates every symbol that does not yet have a saved configuration (the configurate phase always skips already-configured symbols — that filter is forced because the configurator is hours of work and re-running it on the same symbol set has no upside), then forecasts every symbol that is both in the derivative inventory and has a configuration on disk after the configurate phase. All per-symbol forecasts in one run share a single frozen entry date (the next trading day), so the consolidated report filename matches every per- symbol row's date column. When "Only missing symbols" is on, a per-symbol forecast file that already exists on disk for the target entry date is skipped during the forecast phase — saves hours when resuming a partial run or iterating during testing without re-doing completed symbols. Defaults off so a fresh run regenerates every per-symbol forecast. After every symbol is forecasted, writes a consolidated report to forecasts/full/<entry_date>.txt with one row per forecasted symbol carrying (symbol, entry date, direction, configurator accuracy, classifier probability) — the probability column is consumed by magnitude / winner-selection downstream alongside accuracy. Sequential-path processes one symbol at a time within each phase; parallel-path uses a thread pool sized by Max workers (0 = auto-detect from CPU) and the same pool size is used for both phases. Per- symbol failures are best-effort within each phase (logged + counted, do not abort the run); a symbol whose configurator failed simply does not carry forward into the forecast set. Hours of runtime worst case; default concurrency-gate setting (off) holds the system exclusive while it runs.
Rebuild Distributions
Refreshes the cached distributions snapshot the Distributions tab reads from. The cache lives in two tables in SIG's local store and serves the Distributions page without walking the saved per-symbol configuration set on every load. Configurate All Symbols and Forecast All Symbols already refresh this cache at the end of their pipelines, so this action exists for the operator who wants to refresh the snapshot without running a full bulk job (or who's bringing up a fresh deployment with no cache populated yet). Fast — seconds even on a universe-sized inventory. Default concurrency-gate setting (off) holds the system exclusive while the rebuild writes both tables under one transaction.
