Optimization¶

ParameterSweep¶

Parallel grid search using multiprocessing.

from replaybt.optimize.sweep import ParameterSweep

sweep = ParameterSweep(
    strategy_class=MyStrategy,
    data=CSVProvider("ETH_1m.csv"),
    base_config={"initial_equity": 10_000},
    param_grid={
        "take_profit_pct": [0.04, 0.06, 0.08],
        "stop_loss_pct": [0.02, 0.03, 0.04],
    },
    n_workers=10,
)
results = sweep.run()

Constructor¶

ParameterSweep(
    strategy_class: Type[Strategy],
    data: DataProvider,
    base_config: dict,
    param_grid: Dict[str, list],
    n_workers: Optional[int] = None,   # default: cpu_count
)

Methods¶

Method	Returns	Description
`run()`	`SweepResults`	Execute sweep

SweepResults¶

Ranked results from a parameter sweep.

Fields¶

Field	Type	Description
`combos`	`List[dict]`	Each combo has params + metrics

Methods¶

Method	Returns	Description
`best(metric, n)`	`List[dict]`	Top N by metric (descending)
`worst(metric, n)`	`List[dict]`	Bottom N by metric
`filter(**kwargs)`	`SweepResults`	Filter by param values
`to_dataframe()`	`pd.DataFrame`	Export as DataFrame
`summary(metric, top_n)`	`str`	Formatted summary table

Combo Keys¶

Each dict in combos contains all parameter keys from param_grid plus:

Key	Type	Description
`net_pnl`	`float`	Net PnL
`net_return_pct`	`float`	Return percentage
`max_drawdown_pct`	`float`	Max drawdown
`total_trades`	`int`	Trade count
`win_rate`	`float`	Win rate
`profit_factor`	`float`	Profit factor