trading_bot_v4/scripts/sensitivity_analysis.py

#!/usr/bin/env python3
"""
Sensitivity Analysis - Single Parameter Impact Testing
Purpose: Measure exact impact of each parameter in isolation
"""
from __future__ import annotations

import argparse
import sys
from pathlib import Path
from typing import Dict, List, Tuple

PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(PROJECT_ROOT) not in sys.path:
    sys.path.append(str(PROJECT_ROOT))

import pandas as pd
from pathlib import Path
from tqdm import tqdm
from backtester.data_loader import load_csv
from backtester.indicators.money_line import MoneyLineInputs
from backtester.simulator import TradeConfig, simulate_money_line


def test_single_parameter_sensitivity(
    csv_path: Path, 
    symbol: str,
    param_name: str,
    values: List[float]
) -> List[Tuple[float, int, float, float, float]]:
    """
    Vary one parameter while holding others at baseline.
    Returns: [(value, trades, pnl, win_rate, profit_factor), ...]
    """
    data_slice = load_csv(Path(csv_path), symbol, "5m")
    df = data_slice.data
    config = TradeConfig(position_size=8100.0, max_bars_per_trade=288)
    
    results = []
    baseline = MoneyLineInputs()
    
    print(f"\n{'='*80}")
    print(f"TESTING: {param_name}")
    print(f"Values: {values}")
    print(f"{'='*80}")
    
    for val in tqdm(values, desc=f"Testing {param_name}", leave=False):
        # Create inputs with single parameter changed
        kwargs = {param_name: val}
        inputs = MoneyLineInputs(**kwargs)
        
        result = simulate_money_line(df, symbol, inputs, config)
        
        results.append((
            val,
            len(result.trades),
            result.total_pnl,
            result.win_rate,
            result.profit_factor if result.profit_factor != float('inf') else 999.0
        ))
        
        print(f"{param_name}={val:6.2f}: "
              f"{len(result.trades):3d} trades | "
              f"${result.total_pnl:8.2f} PnL | "
              f"{result.win_rate:5.1f}% WR | "
              f"PF={result.profit_factor:.3f}")
    
    return results


def analyze_sensitivity(
    param_name: str, 
    results: List[Tuple[float, int, float, float, float]]
):
    """Analyze parameter sensitivity results."""
    values = [r[0] for r in results]
    trades = [r[1] for r in results]
    pnls = [r[2] for r in results]
    win_rates = [r[3] for r in results]
    profit_factors = [r[4] for r in results]
    
    # Find best by different metrics
    best_pnl_idx = pnls.index(max(pnls))
    best_wr_idx = win_rates.index(max(win_rates))
    best_pf_idx = profit_factors.index(max(profit_factors))
    
    print(f"\n{'='*80}")
    print(f"SENSITIVITY ANALYSIS: {param_name}")
    print(f"{'='*80}")
    
    # Trade count variation
    min_trades, max_trades = min(trades), max(trades)
    if max_trades > 0:
        trade_variation = (max_trades - min_trades) / max_trades * 100
        print(f"Trade Count Range: {min_trades}-{max_trades} ({trade_variation:.1f}% variation)")
    
    # PnL variation
    min_pnl, max_pnl = min(pnls), max(pnls)
    pnl_range = max_pnl - min_pnl
    print(f"PnL Range: ${min_pnl:.2f} to ${max_pnl:.2f} (${pnl_range:.2f} swing)")
    
    # Best values
    print(f"\nBest PnL: {values[best_pnl_idx]:.2f} → ${pnls[best_pnl_idx]:.2f}")
    print(f"Best WR: {values[best_wr_idx]:.2f} → {win_rates[best_wr_idx]:.1f}%")
    print(f"Best PF: {values[best_pf_idx]:.2f} → {profit_factors[best_pf_idx]:.3f}")
    
    # Sensitivity classification
    if pnl_range < 50:
        print(f"\n💡 {param_name}: LOW SENSITIVITY")
        print(f"   PnL swing only ${pnl_range:.2f} - parameter has minimal impact")
    elif pnl_range < 200:
        print(f"\n💡 {param_name}: MODERATE SENSITIVITY")
        print(f"   PnL swing ${pnl_range:.2f} - worth tuning but not critical")
    else:
        print(f"\n💡 {param_name}: HIGH SENSITIVITY")
        print(f"   PnL swing ${pnl_range:.2f} - CRITICAL parameter to optimize")


def run_full_sensitivity_suite(csv_path: Path, symbol: str):
    """Test all parameters for sensitivity."""
    
    test_configs = [
        ("flip_threshold_percent", [0.3, 0.4, 0.5, 0.6, 0.7, 0.8]),
        ("ma_gap_threshold", [0.15, 0.25, 0.35, 0.45, 0.55, 0.65]),
        ("momentum_min_adx", [15.0, 18.0, 21.0, 24.0, 27.0, 30.0]),
        ("momentum_long_max_pos", [60.0, 65.0, 70.0, 75.0, 80.0]),
        ("momentum_short_min_pos", [15.0, 20.0, 25.0, 30.0, 35.0]),
        ("cooldown_bars", [1, 2, 3, 4, 5]),
        ("momentum_spacing", [2, 3, 4, 5, 6]),
        ("momentum_cooldown", [1, 2, 3, 4, 5])
    ]
    
    all_results = {}
    
    print(f"\n{'='*80}")
    print("RUNNING FULL SENSITIVITY SUITE")
    print(f"{'='*80}")
    print(f"Testing {len(test_configs)} parameters")
    print()
    
    for param_name, values in tqdm(test_configs, desc="Overall progress"):
        results = test_single_parameter_sensitivity(csv_path, symbol, param_name, values)
        all_results[param_name] = results
        analyze_sensitivity(param_name, results)
    
    # Summary ranking
    print(f"\n{'='*80}")
    print("PARAMETER SENSITIVITY RANKING")
    print(f"{'='*80}")
    
    # Calculate PnL ranges for ranking
    rankings = []
    for param_name, results in all_results.items():
        pnls = [r[2] for r in results]
        pnl_range = max(pnls) - min(pnls)
        rankings.append((param_name, pnl_range))
    
    rankings.sort(key=lambda x: x[1], reverse=True)
    
    print("\nOptimization Priority (by PnL impact):")
    for i, (param_name, pnl_range) in enumerate(rankings, 1):
        if pnl_range > 200:
            impact = "🔴 CRITICAL"
        elif pnl_range > 100:
            impact = "🟡 HIGH"
        elif pnl_range > 50:
            impact = "🟢 MODERATE"
        else:
            impact = "⚪ LOW"
        print(f"{i}. {param_name:30s}: ${pnl_range:8.2f} swing - {impact}")


def main():
    parser = argparse.ArgumentParser(description="Single-parameter sensitivity analysis")
    parser.add_argument("--csv", type=Path, required=True, help="Path to OHLCV CSV")
    parser.add_argument("--symbol", default="SOL-PERP", help="Symbol name")
    parser.add_argument("--param", help="Test specific parameter only")
    args = parser.parse_args()
    
    print("="*80)
    print("V9 SENSITIVITY ANALYSIS")
    print("="*80)
    print(f"Data: {args.csv}")
    print(f"Symbol: {args.symbol}")
    
    if args.param:
        # Test single parameter
        test_values = {
            "flip_threshold_percent": [0.3, 0.4, 0.5, 0.6, 0.7, 0.8],
            "ma_gap_threshold": [0.15, 0.25, 0.35, 0.45, 0.55, 0.65],
            "momentum_min_adx": [15.0, 18.0, 21.0, 24.0, 27.0, 30.0],
            "momentum_long_max_pos": [60.0, 65.0, 70.0, 75.0, 80.0],
            "momentum_short_min_pos": [15.0, 20.0, 25.0, 30.0, 35.0],
            "cooldown_bars": [1, 2, 3, 4, 5],
            "momentum_spacing": [2, 3, 4, 5, 6],
            "momentum_cooldown": [1, 2, 3, 4, 5]
        }
        
        if args.param not in test_values:
            print(f"❌ Unknown parameter: {args.param}")
            print(f"Available: {', '.join(test_values.keys())}")
            return
        
        results = test_single_parameter_sensitivity(
            args.csv, args.symbol, args.param, test_values[args.param]
        )
        analyze_sensitivity(args.param, results)
    else:
        # Test all parameters
        run_full_sensitivity_suite(args.csv, args.symbol)
    
    print(f"\n{'='*80}")


if __name__ == "__main__":
    main()