guardrailed-llm-agent — Step 3: Feature Engineering: Describing the Market's Mood (snippet 1)

step_3_feature_engineering_describing_th_1.py

feat = df.copy()

# 1. Daily log returns: the foundation of everything
feat['ret'] = np.log(feat['Close']).diff()

# 2. Realized annualized volatility (20-day rolling)
feat['vol'] = feat['ret'].rolling(VOL_WIN).std() * np.sqrt(252)

# 3. Trend score: rolling mean / rolling std
#    Positive = uptrend, negative = downtrend, magnitude = consistency
rmean = feat['ret'].rolling(RET_WIN).mean()
rstd  = feat['ret'].rolling(RET_WIN).std()
feat['trend'] = (rmean / (rstd + 1e-12)).clip(-5, 5)

# 4. Price z-score: how extended is price vs its own rolling mean?
#    Positive = above average (potentially overbought), negative = below
ma = feat['Close'].rolling(ZSCORE_WIN).mean()
sd = feat['Close'].rolling(ZSCORE_WIN).std()
feat['z'] = ((feat['Close'] - ma) / (sd + 1e-12)).clip(-6, 6)

feat['mom63'] = feat['ret'].rolling(63).sum()  # 63-day cumulative return (momentum)
feat['vol5']  = feat['ret'].rolling(5).std() * np.sqrt(252)  # short-term vol (5-day)
feat['ma50']  = feat['Close'].rolling(50).mean()  # 50-day moving average

feat = feat.dropna()