Factor Lab Daily Brief 2026-09-11

Data Cutoff

2026-09-10 (US market close). prices_cache: 499/502 stocks have data through 09-10; remaining 3 at 08-10 to 08-21 (incremental fetch anomaly, to be investigated). Cron ran at 06:30 CST (18:30 ET 09-10), capturing 09-10 close prices only, not including 09-11 (Friday) data.

Sample and Evaluation Methodology

• Universe: S&P 500 constituents, 502 valid stocks (474 for ROE)
• Fundamental factors (EP / BP / FCF Yield / ROE): based on combined_panel.pkl daily TTM earnings panel, n_ic_obs = 60 cross-sections
• Price factors (MOM / VOL / SIZE): n_ic_obs = 39 cross-sections
• MOM: Price[t-21]/Price[t-273] - 1, 21-day forward return
• IC: Spearman rank correlation, factor value vs next-period return (average across cross-sections)
• Long-Short: Q1 long − Q5 short (code sorts Q1 = lowest factor value group), before transaction costs
• p-value: Ordinary t-test, not corrected for overlapping return autocorrelation
• ICIR: IC mean / IC std, not strategy Sharpe
• Long-short direction: Must verify actual Q1-Q5 vs Q5-Q1 ordering in code; uncosted returns ≠ implementable strategy returns

Current vs Previous and 5 Recent Runs Comparison

Note: EP/BP/FCF Yield/ROE are identical across 09-07 to 09-11 (to 4 decimal places), indicating no new price data propagated to fundamental factor computation since 09-07. This is inherent to quarterly financial panel lag, not a bug. The 09-11 run added only one new trading day (09-10) for price factors; small changes in MOM/VOL/SIZE reflect single-day price movement impact on rolling statistics.

Status quo persists. No new data-driven regime change.

Complete 7-Factor Table

⭐ = |IC|/σ > 0.5 or p < 0.05

Key Interpretations:

• MOM −0.20: Negative momentum regime persistent since late August; stable between −0.020 and −0.22 from 09-05 to 09-11, with mild convergence (ICIR from −1.96 → −1.49), but sign unchanged.
• SIZE −0.08: Small cap underpressure continues, ICIR −1.06, highly significant (p uncorrected for overlap).
• EP/FCF Yield/ROE all negative: Value/profitability factors inversely correlated in this sample window. Does not equal “value rotation confirmed”—these are time-averaged cross-sectional IC, not any strategy’s absolute return.
• BP ineffective: IC ≈ 0, p = 0.73, no predictive power.
• VOL +0.05: Low-volatility anomaly direction (high VOL group underperforms), but p = 0.10, not at 0.05 threshold.

Sector Momentum Table (MOM by Sector)

Coverage: 8 of 11 GICS一级 sectors. Health Care MOM p = 0.19, not significant; Utilities p = 0.04 but |IC| only −0.06.

Anomaly Alert Interpretation (against ic_history last 10 unique run dates):

• 7 sectors with “positive-to-negative” MOM IC alerts (🔴)—Consumer Discretionary, Consumer Staples, Financials, Industrials, Information Technology, Utilities—all are existing regime markers, not today’s novelty. ic_history shows these sectors’ MOM IC was positive from 05-29 to mid-August (mean +0.12 to +0.37), turned negative from late August and has persisted. Reports should flag “continuation” not “new.”
• Real Estate is the only sector with positive MOM IC (+0.17); its historical mean in ic_history also hovers near zero, not a reversal.
• Health Care IC = −0.06 is elevated relative to historical mean −0.12 ± 0.03 (z = +2.03), but p-value dropped from 0.044 to 0.191 (no longer significant), possibly reflecting sample noise rather than regime change.

Strategy Implications

1. Momentum reversal is the strongest signal but not a new event: MOM −0.020, IT sector −0.49 is continuation of regime established late August. IT long-short spread 19.6% (Q1 winners − Q5 losers), extreme magnitude but seen at similar levels multiple times within the sample window.
2. Value factors uniformly ineffective: EP/FCF Yield/ROE all negative IC, unchanged from 09-07 to 09-11. Do not use “value reversion” as strategy basis before next quarterly panel update.
3. Small cap pressure persistent: SIZE IC −0.08 deepening daily, consistent with negative MOM—small cap + low momentum combo weakest in this regime.
4. Real Estate isolated positive momentum: Only sector with positive MOM IC (n=31, ic_obs=21), small sample warrants caution.

Conclusions That Cannot Be Drawn

• ❌ MOM negative ≠ market must fall. MOM measures cross-sectional winner vs loser relative performance, not bound to market direction.
• ❌ SIZE negative ≠ small cap rally imminent. It only means size factor cross-sectional regression was negative in this sample window.
• ❌ EP/FCF Yield/ROE negative ≠ value rotation confirmed. These are time-averaged cross-sectional IC means, not any long-short strategy’s actual return.
• ❌ Red light count ≠ independent risk event. 6 of 7 sector “positive-to-negative” alerts are continuation signals per ic_history baseline, not breakouts.
• ❌ p < 0.05 ≠ “not random” (uncorrected for overlapping samples). MOM/VOL/SIZE n_ic_obs = 39 are rolling overlapping windows; standard t-test overstates significance.
• ❌ ICIR is not strategy Sharpe. ICIR = IC_mean / IC_std measures factor stability, not risk-adjusted return.
• ❌ Long-short returns before costs; do not call “implementable strategy returns.”

Statistical and Data Limitations

1. ic_mean is multi-cross-section rank correlation, not daily IC, not future return forecast, not drop probability.
2. MOM uses 21-day forward return by default; the most recent 21 trading days (08-21 to 09-10) cannot be fully evaluated (t-21 factor computation needs data before 08-21).
3. p-values not corrected for overlapping return autocorrelation. 39 observations are non-independent; true significance may be lower than reported.
4. Four fundamental factors unchanged for 4 runs: combined_panel.pkl is quarterly TTM earnings; next update requires new earnings season release.
5. 3 stocks prices_cache lagging 20+ days (EA 08-10, EQR 08-21, one at 08-14); incremental fetch logic may incorrectly short-window data.
6. Incomplete sector coverage: Health Care (p=0.19), Utilities (weak signal) momentum unreliable; 8/11 sectors statistically meaningful.

Charts: /charts/factor-ic-2026-09-11.png · /charts/sector-mom-2026-09-11.png