Skip to content

replaybt

Analysis

mitrionxyz/replaybt

Analysis¶

MonteCarlo¶

Monte Carlo simulation for robustness analysis. Runs shuffle (permutation) and bootstrap (resampling) simulations on completed trade results.

from replaybt import MonteCarlo

mc = MonteCarlo(results=backtest_results, n_simulations=1000)
mc_result = mc.run()
print(mc_result.summary())

Constructor¶

MonteCarlo(
    results: BacktestResults,
    n_simulations: int = 1000,
)

Methods¶

Method	Returns	Description
`run()`	`MonteCarloResult`	Execute simulations

MonteCarloResult¶

Field	Type	Description
`n_simulations`	`int`	Simulations run
`n_trades`	`int`	Trades in source
`initial_equity`	`float`	Starting equity
`shuffle_pnl_percentiles`	`Dict[int, float]`	PnL at 5/25/50/75/95th pct
`shuffle_max_dd_mean`	`float`	Avg max drawdown
`shuffle_max_dd_percentiles`	`Dict[int, float]`	DD at 5/50/95th pct
`bootstrap_pnl_mean`	`float`	Expected PnL
`bootstrap_pnl_std`	`float`	PnL std deviation
`bootstrap_pnl_percentiles`	`Dict[int, float]`	PnL at 5/25/50/75/95th pct
`bootstrap_max_dd_mean`	`float`	Avg max drawdown
`bootstrap_max_dd_percentiles`	`Dict[int, float]`	DD at 5/50/95th pct
`ruin_probability`	`float`	Probability of total loss

Method	Returns	Description
`summary()`	`str`	Formatted report

WalkForward¶

Rolling walk-forward optimization. Trains on one window, tests on the next, repeats.

from replaybt import WalkForward

wf = WalkForward(
    strategy_class=MyStrategy,
    data=data_provider,
    base_config=config,
    param_grid=grid,
    n_windows=4,
    train_pct=0.60,
    metric="net_pnl",
    anchored=False,
)
result = wf.run()
print(result.summary())

Constructor¶

WalkForward(
    strategy_class: Type[Strategy],
    data: DataProvider,
    base_config: dict,
    param_grid: Dict[str, list],
    n_windows: int = 4,
    train_pct: float = 0.60,
    metric: str = "net_pnl",
    anchored: bool = False,
    n_workers: Optional[int] = None,
)

Methods¶

Method	Returns	Description
`run()`	`WalkForwardResult`	Execute walk-forward

WalkForwardResult¶

Field	Type	Description
`windows`	`List[WindowResult]`	Per-window results
`n_windows`	`int`	Number of windows
`anchored`	`bool`	Anchored mode
`train_pct`	`float`	Train fraction
`metric`	`str`	Optimization metric
`oos_net_pnl`	`float`	Combined OOS PnL
`oos_total_trades`	`int`	Combined OOS trades
`oos_win_rate`	`float`	Combined OOS win rate
`oos_max_drawdown_pct`	`float`	Combined OOS drawdown
`param_stability`	`Dict[str, List]`	Params per window

Property	Type	Description
`params_consistent`	`bool`	Same params >50% of windows

Method	Returns	Description
`summary()`	`str`	Formatted report

WindowResult¶

Field	Type	Description
`window_index`	`int`	Window number
`train_start_idx`	`int`	Train start index
`train_end_idx`	`int`	Train end index
`test_start_idx`	`int`	Test start index
`test_end_idx`	`int`	Test end index
`best_params`	`dict`	Best params from train
`train_metrics`	`dict`	Training metrics
`test_result`	`BacktestResults`	OOS test results
`all_combos`	`int`	Combos evaluated