Sector Tilts Fitted First Half Only

Back to leaderboard · 2015-01-01 to 2025-06-01 · v3_top100_monthly

Overall top-100 hits
30.75
average across horizons
Predicted top-100 avg actual rank
228.75
lower is better
Rank IC
0.0464
picks the better of two stocks 51.5% of the time
IC reliability
4.65
t-stat; IC positive 58.9% of months
not adjusted for overlapping windows

Horizon Breakdown

top 100 only
Horizon Average Top-100 Hits Predicted Top-100 Avg Actual Rank Rank IC
3m 30.54 230.49 0.0444
6m 30.63 229.40 0.0448
12m 31.08 226.37 0.0500

Run Note

test summary
Sector Tilts Fitted First Half Only: 30.75 avg hits@100 (3m 30.54, 6m 30.63, 12m 31.08); avg actual rank 228.75; IC +0.0464 (t 4.65).

Run Details

strategy_test_results #113
id
113
strategy_id
strategy_000103
strategy_title
Sector Tilts Fitted First Half Only
strategy_hash
5f6e03e8b179b88347a469c6e5930118f453f320553233354c7b4d18dcb6a4f7
evaluator_version
v3_top100_monthly
start_date
2015-01-01
end_date
2025-06-01
created_at
2026-07-25 11:13:56
updated_at
2026-07-25 13:17:02

Sample Top-100 Lists

2020-03-31 / 6m / 35 hits

Predicted top 100

AAL, ADBE, ADI, AES, AFL, AIG, ALGN, ALK, AMAT, AMD, AMP, ANET, APTV, AVGO, AXP, BA, BAC, C, CCL, CF, CFG, CI, CMG, CNC, CNP, COF, CRM, CTAS, DAL, DHI, DRI, DVN, DXC, ETSY, EXPE, FCX, FITB, FTNT, GE, GM, GPS, HAL, HCA, HIG, HUM, INTC, INTU, ISRG, IVZ, KEY, KLAC, LEN, LKQ, LNC, LRCX, LYB, LYV, MA, MCHP, MCO, MGM, MOS, MPC, MS, MSFT, MU, NCLH, NOW, NVDA, NWL, OXY, PAYC, PHM, PRU, PYPL, QCOM, RCL, ROST, RTX, SLB, SPG, SPGI, SWK, SYF, SYY, TDG, TER, TPR, TSLA, UAL, UHS, ULTA, UNH, UNM, URI, V, VRTX, VTR, WDC, WYNN

Actual top 100

AAP, AAPL, ABMD, ADBE, ADSK, ALB, ALGN, AMD, AMP, AMZN, APD, APH, APTV, ATVI, AVGO, BBY, BWA, CARR, CDNS, CE, CHRW, CMG, CMI, CPRT, CRM, CTAS, CTLT, CTSH, DD, DE, DHI, DHR, DRI, DXCM, EBAY, EMN, ETSY, EXPE, FCX, FDX, GLW, GPS, HAL, HD, IDXX, IPGP, JCI, KMX, LEG, LEN, LH, LOW, MAS, MCHP, MGM, MOS, NCLH, NKE, NOW, NSC, NVDA, NWS, NWSA, ORLY, PAYC, PH, PHM, PNR, POOL, PVH, PWR, PYPL, QCOM, QRVO, RCL, ROK, ROL, SHW, SIVB, SNPS, SWK, SWKS, SYF, TDG, TEL, TER, TGT, TMO, TSCO, TSLA, TT, UPS, URI, VAR, VIAC, VTR, WHR, WMB, WST, WY

Matched tickers

ADBE, ALGN, AMD, AMP, APTV, AVGO, CMG, CRM, CTAS, DHI, DRI, ETSY, EXPE, FCX, GPS, HAL, LEN, MCHP, MGM, MOS, NCLH, NOW, NVDA, PAYC, PHM, PYPL, QCOM, RCL, SWK, SYF, TDG, TER, TSLA, URI, VTR

Strategy Script

5f6e03e8b1
"""Generated V3 stock-ranking strategy."""
from __future__ import annotations

import numpy as np
import pandas as pd


STRATEGY_ID = "strategy_000103"
STRATEGY_TITLE = "Sector Tilts Fitted First Half Only"
DESCRIPTION = """
If sector tilts capture a real persistent effect, tilts fitted on 2015 to
mid-2020 should also help on the untouched 2020 to 2025 period. If they are
period-specific noise they will hurt.
"""
FACTORS_USED = [
    "vol_63d",
    "revenue_growth_pct",
    "dollar_volume_21d",
    "market_cap",
    "free_cash_flow_margin_pct",
    "forward_pe",
    "pe",
    "eps_growth_pct",
    "operating_income_growth_pct",
    "free_cash_flow_growth_pct",
    "volume",
    "sector"
]
PARAMETERS = {
    "vol_63d": 1.0,
    "revenue_growth_top": 0.4,
    "turnover": 0.4,
    "dollar_volume_21d": 0.15,
    "fcf_margin_top": 0.15,
    "fam_value": -0.164,
    "fam_steadiness": 0.0218,
    "fam_growth": -0.065,
    "volume_raw": 0.05,
    "fcf_margin": 0.225,
    "sec_energy": -0.2,
    "sec_health_care": 0.1,
    "sec_utilities": 0.15,
    "sec_real_estate": -0.1
}
COMPLEXITY = 14
PARENT_STRATEGY_ID = "strategy_000102"
GENERATION = 103
RANDOM_SEED = 20260903
CREATED_AT = "2026-07-25T00:00:00+00:00"


def _date_group(df):
    if "trade_date" in df.columns:
        return df["trade_date"]
    return pd.Series(0, index=df.index)


def _rank_pct(df, series):
    """Date-local rank percentile centred on zero; missing values stay neutral."""
    clean = pd.Series(series, index=df.index).replace([np.inf, -np.inf], np.nan)
    ranked = clean.groupby(_date_group(df)).rank(pct=True)
    return ranked.fillna(0.5) - 0.5


def _rank_col(df, column):
    if column not in df.columns:
        return pd.Series(0.0, index=df.index)
    return _rank_pct(df, df[column])


def _zscore_col(df, column):
    if column not in df.columns:
        return pd.Series(0.0, index=df.index)
    clean = df[column].replace([np.inf, -np.inf], np.nan)
    grouped = clean.groupby(_date_group(df))
    mean = grouped.transform("mean")
    std = grouped.transform("std").replace(0, np.nan)
    return ((clean - mean) / std).fillna(0.0)


def _top_half(series):
    """Emphasise only the upper half of a centred rank, leaving the rest flat."""
    return series.clip(lower=0.0) * 2.0 - 0.25


def _turnover(df):
    market_cap = df["market_cap"].replace(0, np.nan) if "market_cap" in df.columns else np.nan
    return _rank_pct(df, df["dollar_volume_21d"] / market_cap)


def _range_252(df):
    return _rank_pct(df, df["from_low_252d"] - df["from_high_252d"])


def _growth_frame(df):
    cols = ["eps_growth_pct", "revenue_growth_pct",
            "operating_income_growth_pct", "free_cash_flow_growth_pct"]
    return df[cols].replace([np.inf, -np.inf], np.nan).clip(lower=-100, upper=300)


def _steadiness(df):
    """Reward growth that is consistent across the four metrics, not one spike."""
    g = _growth_frame(df)
    mean = g.mean(axis=1).fillna(0)
    spread = g.std(axis=1).fillna(0)
    breadth = (g > 0).sum(axis=1) / g.shape[1]
    return breadth - (spread / (mean.abs() + 50)).clip(lower=0, upper=3)


def _overheat(df):
    """Penalise a growth spike that revenue does not support, worse when price is hot."""
    g = _growth_frame(df)
    quality = g[["eps_growth_pct", "operating_income_growth_pct",
                 "free_cash_flow_growth_pct"]].max(axis=1).fillna(0)
    unsupported = (quality - g["revenue_growth_pct"].fillna(0) - 100).clip(lower=0, upper=300) / 100
    explosive = (g.max(axis=1).fillna(0) - 100).clip(lower=0, upper=300) / 100
    hot = (
        (df["return_1m_pct"].fillna(0) > 25).astype(float)
        + (df["return_3m_pct"].fillna(0) > 60).astype(float)
        + (df["from_52w_high_pct"].fillna(-1) > -2).astype(float)
    ) / 3.0
    return explosive * (1 + hot) + 0.7 * unsupported


def _constructive_pullback(df):
    """Distance below the 52-week high, counted only while 6-month return is positive."""
    pull = (-df["from_52w_high_pct"].fillna(0)).clip(lower=0, upper=0.45)
    return pull.where(df["return_6m_pct"].fillna(0) > 0, 0)


def _zscore_series(df, series):
    clean = pd.Series(series, index=df.index).replace([np.inf, -np.inf], np.nan)
    grouped = clean.groupby(_date_group(df))
    mean = grouped.transform("mean")
    std = grouped.transform("std").replace(0, np.nan)
    return ((clean - mean) / std).fillna(0.0)


def _revenue(df):
    return df["revenue_ttm"].replace(0, np.nan) if "revenue_ttm" in df.columns else pd.Series(np.nan, index=df.index)


def _revenue_lag(df, periods):
    """Revenue_ttm shifted back N monthly observations per ticker, trailing only."""
    if "revenue_ttm" not in df.columns or "ticker" not in df.columns:
        return pd.Series(np.nan, index=df.index)
    order = df.sort_values(["ticker", "trade_date"]).index
    shifted = _revenue(df).loc[order].groupby(df["ticker"].loc[order]).shift(periods)
    return shifted.reindex(df.index)


def _sector(df):
    if "sector" not in df.columns:
        return pd.Series("unknown", index=df.index)
    return df["sector"].fillna("unknown")


def _sector_group(df):
    return [_date_group(df), _sector(df)]


def _sector_rank(df, column):
    """Rank percentile within (date, sector) instead of within date."""
    if column not in df.columns:
        return pd.Series(0.0, index=df.index)
    clean = df[column].replace([np.inf, -np.inf], np.nan)
    ranked = clean.groupby(_sector_group(df)).rank(pct=True)
    return ranked.fillna(0.5) - 0.5


def _sector_flag(df, name):
    flag = (_sector(df) == name).astype(float)
    return flag - flag.mean()


def _terms(df):
    return {
        "vol_63d": _rank_col(df, "vol_63d"),
        "revenue_growth_top": _top_half(_rank_col(df, "revenue_growth_pct")),
        "turnover": _turnover(df),
        "dollar_volume_21d": _rank_col(df, "dollar_volume_21d"),
        "fcf_margin_top": _top_half(_rank_col(df, "free_cash_flow_margin_pct")),
        "fam_value": (-0.65 * _zscore_col(df, "forward_pe") - 0.35 * _zscore_col(df, "pe")),
        "fam_steadiness": _zscore_series(df, _steadiness(df)),
        "fam_growth": (0.55 * _zscore_col(df, "eps_growth_pct") + 0.45 * _zscore_col(df, "revenue_growth_pct")),
        "volume_raw": _rank_col(df, "volume"),
        "fcf_margin": _rank_col(df, "free_cash_flow_margin_pct"),
        "sec_energy": _sector_flag(df, "Energy"),
        "sec_health_care": _sector_flag(df, "Health Care"),
        "sec_utilities": _sector_flag(df, "Utilities"),
        "sec_real_estate": _sector_flag(df, "Real Estate"),
    }




def calculate_score(df):
    terms = _terms(df)
    score = pd.Series(0.0, index=df.index)
    for name, weight in PARAMETERS.items():
        if weight:
            score = score + weight * terms[name]
    return score.replace([np.inf, -np.inf], np.nan).fillna(0.0)