trader.simulation.runner module

Bundles the simulator’s collaborators so run_once can trigger it without importing simulation/, persistence/simulation.py, or SimulationSettings directly (issue #33) — the CLI builds one SimulationRunner and passes it in, keeping the pipeline layer’s import surface unchanged for a feature it merely triggers.

class trader.simulation.runner.SimulationRunner(settings, pipeline_config, discovery_settings, fixed_symbols, bar_repository, decision_repository, simulation_repository)[source]

Bases: object

Steps every configured strategy’s shadow portfolio forward one cycle.

Since issue #42, that is two portfolios per strategy id: the normal one (its own sell decisions honoured) and an Always-be-Buying (AbB) variant that reads the same decisions with sell signals ignored, so only the trailing stop can close a position — isolating whether the sell side is what costs performance, independent of whether the entries are any good.

Parameters:
settings: SimulationSettings
pipeline_config: PipelineConfig
discovery_settings: DiscoverySettings
fixed_symbols: Collection[str]
bar_repository: BarRepository
decision_repository: DecisionRepository
simulation_repository: SimulationRepository
run(*, strategy_ids, now)[source]

Advance both of one strategy’s portfolios per id, all four* failure points isolated from each other.

(*) two per strategy, times as many strategies as strategy_ids holds. One variant of one strategy raising (a malformed inputs_json, a repository error) must not stop its own sibling variant, another strategy’s either variant, or the real cycle this was called from — the same per-unit isolation discipline discover_symbols uses per-theme.

Parameters:

  • strategy_ids (Sequence[str])

  • now (datetime)

Return type:

None