SOL/USDTで大損？「値段の勢いをはかる作戦」の、びっくりな結果と学べること

"""
Relative Spread Strength Trading Signal Generator
異なる期間の移動平均スプレッドから強度を測定
"""
import pandas as pd
import numpy as np


def calculate_rss_signals(df: pd.DataFrame,
                         fast_period: int = 10,
                         slow_period: int = 40) -> pd.DataFrame:
    """
    Relative Spread Strength戦略のシグナル生成
    
    Parameters:
    -----------
    df : pd.DataFrame
        OHLCVデータ
    fast_period : int
        短期EMA期間（デフォルト: 10）
    slow_period : int
        長期EMA期間（デフォルト: 40）
    
    Returns:
    --------
    pd.DataFrame
        シグナルが追加されたDataFrame
    """
    df = df.copy()
    
    # 指数移動平均
    df['ema_fast'] = df['close'].ewm(span=fast_period, adjust=False).mean()
    df['ema_slow'] = df['close'].ewm(span=slow_period, adjust=False).mean()
    
    # スプレッド
    df['spread'] = df['ema_fast'] - df['ema_slow']
    
    # スプレッドの移動平均と標準偏差
    df['spread_ma'] = df['spread'].rolling(window=slow_period).mean()
    df['spread_std'] = df['spread'].rolling(window=slow_period).std()
    
    # 相対スプレッド強度（標準化）
    df['rss'] = (df['spread'] - df['spread_ma']) / (df['spread_std'] + 0.0001)
    
    # RSSのスムージング
    df['rss_smooth'] = df['rss'].ewm(span=3, adjust=False).mean()
    
    # RSSの加速度
    df['rss_acceleration'] = df['rss_smooth'] - df['rss_smooth'].shift(1)
    
    # ダイナミックレベル
    df['rss_upper'] = 2.0
    df['rss_lower'] = -2.0
    
    # トレンド強度
    df['trend_strength'] = np.abs(df['rss_smooth'])
    
    # シグナル生成
    df['rss_prev'] = df['rss_smooth'].shift(1)
    df['is_buy'] = (
        ((df['rss_smooth'] > 0) & (df['rss_prev'] <= 0)) |  # ゼロラインクロス
        ((df['rss_smooth'] > df['rss_lower']) & (df['rss_prev'] <= df['rss_lower']) & 
         (df['rss_acceleration'] > 0))  # 売られ過ぎからの反発
    ) & df['rss_smooth'].notna()
    df['is_sell'] = (
        ((df['rss_smooth'] < 0) & (df['rss_prev'] >= 0)) |  # ゼロラインクロス
        ((df['rss_smooth'] < df['rss_upper']) & (df['rss_prev'] >= df['rss_upper']) &
         (df['rss_acceleration'] < 0))  # 買われ過ぎからの反落
    ) & df['rss_smooth'].notna()
    
    # 不要カラム削除
    df.drop(['ema_fast', 'ema_slow', 'spread', 'spread_ma', 'spread_std',
             'rss_smooth', 'rss_acceleration', 'rss_upper', 'rss_lower',
             'trend_strength', 'rss_prev'], axis=1, inplace=True, errors='ignore')
    
    return df