Factor Lab Daily Brief 2026-09-16

Data & Sample

Factor Overview

⭐ marks p < 0.05.

Key observations:

• MOM remains negative (-0.1855); only 12.8% of 39 cross-sections positive. Winners significantly underperform losers—reversal regime strong, momentum ineffective.
• EP, FCF Yield, ROE all persist negative and identical to yesterday (diff = 0). This is not a “factor reversal today”—the fundamental factor cache was not refreshed. Price cache ends at 09-15, but TTM panel ICs carry forward unchanged.
• BP remains ineffective (IC ≈ 0, p = 0.73).
• VOL holds positive IC (low-vol anomaly persists); SIZE holds negative IC (small-cap premium within-sample).

Status Quo Continuation

EP, BP, FCF Yield, ROE fundamental factors show zero change vs yesterday (diff = 0.0000). The ic_history.json shows these factors’ entries written identically each day since 09-08. This means:

• No “new” fundamental factor reversal today—market data from 09-08 to 09-16 had no effect on fundamental ICs (cache reused verbatim).
• ic_history.json contains大量 duplicate date entries (same date written 2-7 times), which inflates baseline distribution weights and affects z-score calculations in anomaly detection. Deviation alerts at ±2σ should be treated as continuation of existing regime, not daily breakpoints.

Sector Momentum Map (MOM by Sector)

Key observations:

• IT momentum IC = -0.477 is the strongest signal this run—nearly all 39 cross-sections negative (only 7.7% positive). Q5 (high momentum/winners) mean return +0.03%, Q1 (low momentum/losers) +14.36%. This is not “tech stocks falling”—it is strong within-sector reversal in technology.
• Financials and Industrials also show significant reversal (IC < -0.2, p < 0.001). | Real Estate is the only sector with positive IC (+0.126), but N=31, p=0.136, not statistically significant.
• Health Care IC = -0.034, p = 0.465—momentor factor entirely ineffective in healthcare, no pattern within-sample.

Implications for Strategy

1. Momentum factor currently not usable. Full-market MOM IC = -0.186, with IT/Industrials/Financials all deeply negative. Traditional “chase winners” strategies would lose money persistently in this regime. Reversal approaches (sell recent winners, buy recent losers) show stronger within-sample performance, but note:

   • Long-short direction is Q1 (low momentum/losers) minus Q5 (high momentum/winners); positive long-short (+5.65%) means losers outperform winners.
   • Transaction costs not deducted; actual implementable returns below stated values.

2. Fundamental factors (EP/FCF Yield/ROE) all negative in this run. This does not directly imply value factor失效—the IC is the mean of 60 historical cross-sections, not a daily IC. Negative values mean that across 60 past cross-section dates, high-factor groups underperformed low-factor groups. Caveats: -本期 EP/FCF Yield/ROE ic_history entries repeated identically since 09-08, meaning the actual as-of date of these numbers may predate 09-08. Do not treat as today’s valid signal.

   • SIZE IC = -0.074 across 39 cross-sections, highly significant (p < 0.0001)—small caps outperformed large caps within-sample. This is a cross-sectional correlation, not a “full-market small-cap rally” forecast.

3. Low-vol anomaly persists. VOL IC = +0.061 significant positive (p = 0.021)—high-vol groups underperform low-vol groups. Classic low-vol anomaly, still effective in this regime. Long-short Q1-Q5 = -2.42%, meaning Q5 (high vol) overall underperforms Q1 (low vol).

Conclusions That Cannot Be Drawn

• ❌ Cannot say “MOM flipped from positive to negative today”—ic_history shows momentum turned negative since late August; today IC = -0.186 is continuation of the existing reversal regime, not a new flip.
• ❌ Cannot say “EP/FCF Yield/ROE reversed today”—all three diffs = 0, cache not updated, no new information.
• ❌ Cannot say “fundamental factors negative = value investing无效”—IC mean is historical cross-section average, not future return; negative only means high-factor groups lagged low-factor groups over the past 60 cross-sections.
• ❌ Cannot say “SIZE negative = small caps will rise”—this is cross-sectional correlation, not a time-series forecast.
• ❌ Sector negative count (7/8 negative) does not equal “7 new sector reversals today”—baseline contains大量 June-July positive-momentum records; z-score alerts mark deviation, not daily breakpoints.

Statistical & Data Limitations