"""
Test pandas EMA behavior to understand Wilder's smoothing initialization
"""

import pandas as pd
import numpy as np

def test_pandas_ema():
    """Test how pandas EMA works with Wilder's smoothing."""
    
    # Sample data from our debug
    prices = [16568.00, 16569.00, 16569.00, 16568.00, 16565.00, 16565.00, 
              16565.00, 16565.00, 16565.00, 16565.00, 16566.00, 16566.00, 
              16563.00, 16566.00, 16566.00, 16566.00, 16566.00, 16566.00]
    
    # Calculate deltas
    deltas = np.diff(prices)
    gains = np.where(deltas > 0, deltas, 0)
    losses = np.where(deltas < 0, -deltas, 0)
    
    print("Price changes:")
    for i, (delta, gain, loss) in enumerate(zip(deltas, gains, losses)):
        print(f"Step {i+1}: delta={delta:5.2f}, gain={gain:4.2f}, loss={loss:4.2f}")
    
    # Create series
    gain_series = pd.Series(gains)
    loss_series = pd.Series(losses)
    
    period = 14
    alpha = 1.0 / period
    
    print(f"\nUsing period={period}, alpha={alpha:.6f}")
    
    # Test different EMA parameters
    print("\n1. Standard EMA with min_periods=period:")
    avg_gain_1 = gain_series.ewm(alpha=alpha, adjust=False, min_periods=period).mean()
    avg_loss_1 = loss_series.ewm(alpha=alpha, adjust=False, min_periods=period).mean()
    
    print("Index | Gain | Loss | AvgGain | AvgLoss | RS | RSI")
    print("-" * 60)
    for i in range(min(len(avg_gain_1), 18)):
        gain = gains[i] if i < len(gains) else 0
        loss = losses[i] if i < len(losses) else 0
        avg_g = avg_gain_1.iloc[i]
        avg_l = avg_loss_1.iloc[i]
        
        if not (pd.isna(avg_g) or pd.isna(avg_l)) and avg_l != 0:
            rs = avg_g / avg_l
            rsi = 100 - (100 / (1 + rs))
        else:
            rs = np.nan
            rsi = np.nan
            
        print(f"{i:5d} | {gain:4.2f} | {loss:4.2f} | {avg_g:7.4f} | {avg_l:7.4f} | {rs:4.2f} | {rsi:6.2f}")
    
    print("\n2. EMA with min_periods=1:")
    avg_gain_2 = gain_series.ewm(alpha=alpha, adjust=False, min_periods=1).mean()
    avg_loss_2 = loss_series.ewm(alpha=alpha, adjust=False, min_periods=1).mean()
    
    print("Index | Gain | Loss | AvgGain | AvgLoss | RS | RSI")
    print("-" * 60)
    for i in range(min(len(avg_gain_2), 18)):
        gain = gains[i] if i < len(gains) else 0
        loss = losses[i] if i < len(losses) else 0
        avg_g = avg_gain_2.iloc[i]
        avg_l = avg_loss_2.iloc[i]
        
        if not (pd.isna(avg_g) or pd.isna(avg_l)) and avg_l != 0:
            rs = avg_g / avg_l
            rsi = 100 - (100 / (1 + rs))
        elif avg_l == 0 and avg_g > 0:
            rs = np.inf
            rsi = 100.0
        else:
            rs = np.nan
            rsi = np.nan
            
        print(f"{i:5d} | {gain:4.2f} | {loss:4.2f} | {avg_g:7.4f} | {avg_l:7.4f} | {rs:4.2f} | {rsi:6.2f}")

if __name__ == "__main__":
    test_pandas_ema()