#!/usr/bin/env python3
"""Debug MA gap scoring to understand parameter insensitivity."""

import sys
from pathlib import Path
import numpy as np
import pandas as pd

# Add backtester to path
sys.path.insert(0, str(Path(__file__).parent.parent))

from backtester.data_loader import load_csv
from backtester.indicators.money_line import MoneyLineInputs, ema

def analyze_ma_gap_distribution(csv_path: str):
    """Analyze MA gap score distribution to understand parameter behavior."""
    
    print("=" * 80)
    print("MA GAP DISTRIBUTION ANALYSIS")
    print("=" * 80)
    
    # Load data
    data_slice = load_csv(Path(csv_path), "SOL-PERP", "5m")
    df = data_slice.data  # DataSlice.data is the DataFrame
    print(f"Loaded {len(df)} bars\n")
    
    # Calculate EMAs
    df["ema_fast"] = ema(df["close"], 50)
    df["ema_slow"] = ema(df["close"], 200)
    
    # Calculate raw MA gap percentage
    ma_gap = 100.0 * (df["ema_fast"] - df["ema_slow"]) / df["close"]
    
    print("RAW MA GAP (%) DISTRIBUTION:")
    print(f"  Min: {ma_gap.min():.4f}%")
    print(f"  25th: {ma_gap.quantile(0.25):.4f}%")
    print(f"  Median: {ma_gap.median():.4f}%")
    print(f"  75th: {ma_gap.quantile(0.75):.4f}%")
    print(f"  Max: {ma_gap.max():.4f}%")
    print(f"  Std: {ma_gap.std():.4f}%\n")
    
    # Test different thresholds
    for threshold in [0.2, 0.3, 0.35, 0.4, 0.5]:
        gap_score = np.tanh(ma_gap / threshold)
        
        print(f"\nTHRESHOLD = {threshold} (default: 0.35):")
        print(f"  gap_score min: {gap_score.min():.6f}")
        print(f"  gap_score max: {gap_score.max():.6f}")
        print(f"  gap_score median: {gap_score.median():.6f}")
        
        # Check flip_threshold comparisons
        for flip_pct in [0.4, 0.5, 0.6, 0.7]:
            flip_threshold = flip_pct / 100.0  # Convert to decimal
            
            # Count how many bars exceed threshold
            long_signals = (gap_score > flip_threshold).sum()
            short_signals = (gap_score < -flip_threshold).sum()
            total_signals = long_signals + short_signals
            
            print(f"    flip_threshold={flip_pct}% → {total_signals} potential signals")
            print(f"      (LONG: {long_signals}, SHORT: {short_signals})")

if __name__ == "__main__":
    if len(sys.argv) != 2:
        print("Usage: python debug_ma_gap.py <csv_path>")
        sys.exit(1)
    
    analyze_ma_gap_distribution(sys.argv[1])