Factor Lab Daily Brief 2026-09-30

Data Cut-off and Sample

• Price data cut-off: 2026-09-28 (latest date in prices_cache.parquet)
• Beijing time: 2026-09-30 06:30; EDT 2026-09-29 18:30 (after Tuesday close; Monday 9/29 market holiday)
• Valid stocks: 502 (fundamental factors: 474–502)
• Evaluation window: Cross-sectional factor value → next 21-day return (Spearman rank correlation IC)
• Valid observations: 60 cross-sections for fundamental factors; 36 cross-sections for price factors (MOM/VOL/SIZE)
• ⚠️ Note: The most recent 21 trading days (approx. 09-03 to 09-28) cannot yet fully evaluate 21-day forward returns; IC means below use all available historical cross-sections

Running-Day Data Continuity Check

This run (09-30) produces identical results to the previous run (09-29) – all factor IC means, ICIRs, and p-values match digit by digit. Compared to the progressive changes seen in 09-26/25/24 (MOM from -0.147 → -0.129 → -0.109 → -0.076 → -0.048; VOL from +0.086 → +0.093 → +0.102 → +0.110 → +0.117), this run introduced no new price data.

Conclusion: Price data effectively stops at 09-28; 09-30 output is a continuation of the existing state.

Seven-Factor Summary Table

⭐ = p < 0.05 (statistically significant; ⚠️ not corrected for overlapping-sample autocorrelation)

Key interpretations:

• MOM negative but insignificant: Momentum factor broadly ineffective; p=0.255 fails to reject “no relationship.” Continued recovery from -0.147 (09-22 to 09-30), deviating from the positive-momentum baseline but not yet stable.
• VOL stable positive and highly significant: ICIR=1.05, the highest among all factors. High-volatility groups underperform low-vol (Q1-Q5 = -4.85%), confirming the low-vol anomaly persists.
• EP / FCF Yield / ROE all negative and significant: High-earnings, high-FCF, high-ROE groups underperform – “value” and “quality” factors are inverted in this cross-section.
• BP and SIZE completely ineffective: p > 0.8, IC near zero.

Sector Momentum Table (MOM by Sector)

Coverage: 7 GICS sectors (this framework covers momentum-computable sectors among the 11 GICS; Real Estate, Energy, Materials do not produce independent rows in this framework).

Sector momentum changes

• 5 sectors flipped positive→negative: IT, Utilities, Financials, Industrials, Utilities. Per ic_history.json, these sectors’ positive-momentum baseline dates back to 05-29. IT flipped from +0.309 to -0.232 on 08-19 and has stayed negative – this is a continuing regime shift, not a 09-30 anomaly.
• Health Care flipped negative→positive: Historical mean -0.108, current +0.080 (z=+4.68), significantly偏离. Low-momentum groups outperform (LS -4.72%), suggesting “catch-up/reversal” dynamics within healthcare.
• Consumer Discretionary & Staples: From significant positive to weak positive; p-values 0.329/0.382, no longer significant.

Strategic Implications

1. Low-vol strategy is currently the most reliable signal: VOL ICIR > 1.0, p ≈ 0. A long-low-vol / short-high-vol portfolio recorded -4.85% mean return in-sample (Q1 low-vol -2.37% vs Q5 high-vol +2.48%). ⚠️ This is a paper portfolio return, unadjusted for transaction costs – not an implementable strategy return.
2. Momentum in regime transition: Overall MOM IC is negative and insignificant; five sectors have also flipped. Pure momentum strategies are not advisable in this cross-section. IT’s reversal from +0.31 to -0.23 is large and persistent – the “strongest get stronger” dynamic within AI-narrative tech stocks has reversed.
3. Value/quality factors inverted in current cross-section: EP, FCF Yield, ROE all negative and significant. This does not mean “value rotation confirmed” – these factors exhibit sharp regime switches across different periods.
4. BP and SIZE have no predictive power in this sample: Exclude from strategy construction for now.

Conclusions That Cannot Be Drawn

• ❌ MOM negative ≠ market will fall. MOM tests cross-sectional ranking vs. forward returns within S&P 500, not market direction.
• ❌ EP/FCF Yield/ROE negative ≠ “value stocks about to outperform growth.” This is a cross-sectional snapshot; these factors flip regimes sharply over time.
• ❌ SIZE negative ≠ small-cap rally incoming. SIZE shows no significant rank correlation with future returns in this test.
• ❌ Sector reversal alerts ≠ single-day突发事件. Per ic_history, most sector momentum reversals began 08-19 to 09-03; this run did not change established facts.
• ❌ p < 0.05 ≠ profitable strategy. t-tests do not handle overlapping-sample autocorrelation; p-values marked “uncorrected.” ICIR is IC mean / IC std, not a strategy Sharpe ratio.
• ❌ Quantile portfolio returns are pre-cost. Q1-Q5 LS is a paper calculation; slippage, borrow costs, and rebalancing frequency materially affect implementable returns.

Statistical and Data Limitations