tuning.py

import itertools
from typing import Dict, List, Any, Tuple
import pandas as pd

from backtester import Backtester
from strategies.mean_reversion import MeanReversionStrategy
from strategies.momentum import MomentumStrategy
from strategies.market_making import MarketMakingStrategy
from metrics import compute_performance_metrics

STRATEGY_MAP = {
    'mean_reversion': MeanReversionStrategy,
    'momentum': MomentumStrategy,
    'market_making': MarketMakingStrategy,
}


def grid_search_strategy(
    data: pd.DataFrame,
    strategy_name: str,
    param_grid: Dict[str, List[Any]],
    latency_ms: int = 50,
    fee_per_share: float = 0.0005,
    slippage_bps: float = 2.0,
) -> Tuple[pd.DataFrame, dict, dict]:
    """
    Simple parameter grid search returning a summary DataFrame, best metrics, and best results.
    """
    StrategyCls = STRATEGY_MAP[strategy_name]
    keys = list(param_grid.keys())
    combos = list(itertools.product(*[param_grid[k] for k in keys]))

    rows = []
    best_metrics = None
    best_results = None
    for combo in combos:
        params = dict(zip(keys, combo))
        strat = StrategyCls(**params)
        bt = Backtester(data=data, strategy=strat, latency_ms=latency_ms, fee_per_share=fee_per_share, slippage_bps=slippage_bps)
        results = bt.run()
        metrics = compute_performance_metrics(results)
        row = {**params, **metrics}
        rows.append(row)
        if best_metrics is None or metrics['sharpe_ratio'] > best_metrics['sharpe_ratio']:
            best_metrics = metrics
            best_results = results
    summary = pd.DataFrame(rows)
    summary = summary.sort_values('sharpe_ratio', ascending=False)
    return summary, best_metrics, best_results