strategy_000001

"""Generated V3 stock-ranking strategy."""
from __future__ import annotations

import numpy as np


STRATEGY_ID = "strategy_000001"
DESCRIPTION = """
Ranks S&P 500 candidates using medium-term momentum, quality growth, valuation,
and a short-term risk adjustment. It rewards constructive pullbacks from 52-week highs
only when 6-month momentum is positive.
"""
FACTORS_USED = [
    "return_6m_pct",
    "momentum_12_1_pct",
    "eps_growth_pct",
    "revenue_growth_pct",
    "forward_pe",
    "vol_63d",
    "from_52w_high_pct",
    "from_200d_ma_pct",
    "market_cap",
]
PARAMETERS = {
    "momentum_weight": 0.482186814567,
    "quality_weight": 0.209719609938,
    "value_weight": 0.221719583982,
    "risk_penalty": 0.274342007265,
    "drawdown_bonus": 0.044126602183,
    "market_cap_penalty": 0.078049788131,
    "min_price_sma": -0.058217647682,
}
COMPLEXITY = 5
PARENT_STRATEGY_ID = None
GENERATION = 1
RANDOM_SEED = 42
CREATED_AT = "2026-07-24T22:42:18+00:00"


def _zscore(series):
    clean = series.replace([np.inf, -np.inf], np.nan)
    std = clean.std()
    if not np.isfinite(std) or std == 0:
        return clean * 0.0
    return (clean - clean.mean()) / std


def eligibility_filter(df):
    # Avoid severely broken trends; the evaluator already applies price/volume
    # filters, so this strategy-level filter is deliberately narrow.
    return df["from_200d_ma_pct"].fillna(0) >= PARAMETERS["min_price_sma"]


def calculate_score(df):
    # Momentum combines raw 6-month return with 12-minus-1 momentum so the score
    # favors both recent leadership and sustained prior trend.
    momentum = 0.65 * _zscore(df["return_6m_pct"]) + 0.35 * _zscore(df["momentum_12_1_pct"])

    # Quality growth rewards companies where revenue and EPS are both improving.
    quality = 0.55 * _zscore(df["eps_growth_pct"]) + 0.45 * _zscore(df["revenue_growth_pct"])

    # Lower forward P/E is better, but missing valuation data is neutralized by
    # the evaluator's rank fill below rather than treated as automatically cheap.
    value = -_zscore(df["forward_pe"])

    # Penalize short-term instability; jumpier names need more momentum/quality
    # to stay highly ranked.
    risk = _zscore(df["vol_63d"])

    # Interaction: constructive pullbacks receive a bonus only when momentum is
    # positive. Deep drawdowns without momentum do not get rewarded.
    pullback = (-df["from_52w_high_pct"]).clip(lower=0, upper=0.35)
    constructive_pullback = pullback.where(df["return_6m_pct"].fillna(0) > 0, 0)

    # Mild size penalty keeps mega-cap stability from dominating a gain forecast.
    size_penalty = _zscore(np.log1p(df["market_cap"].clip(lower=0)))

    score = (
        PARAMETERS["momentum_weight"] * momentum
        + PARAMETERS["quality_weight"] * quality
        + PARAMETERS["value_weight"] * value
        - PARAMETERS["risk_penalty"] * risk
        + PARAMETERS["drawdown_bonus"] * constructive_pullback
        - PARAMETERS["market_cap_penalty"] * size_penalty
    )
    score = score.where(eligibility_filter(df), -1e9)
    return score.replace([np.inf, -np.inf], np.nan).fillna(score.median()).fillna(0.0)