🧮 Factors & seasonality因子与季节性?The cross-sectional factors quant equity desks actually trade — computed from SEC EDGAR fundamentals joined to prices, not proxied by an ETF. Each factor is a winsorized cross-sectional z-score (higher = more attractive on that factor); the composite is their equal-weight blend. These are RANKS and context, not a backtested alpha: factors decay and crowd, free fundamentals are sparse for some tags, book/price is weak for intangible-heavy firms, and every value is lagged to the filing period (no look-ahead).量化股票交易台真正使用的横截面因子 — 由 SEC EDGAR 基本面与价格计算，而非用 ETF 代理。每个因子为去极值后的横截面 z 分数（越高 = 在该因子上越有吸引力）；综合为等权混合。这些是排名与背景，而非回测出的超额收益：因子会衰减与拥挤，免费基本面在部分科目上稀疏，市净率对无形资产密集的公司较弱，且所有数值滞后至申报期（无前视）。
Which styles of stock are working — and what this month usually brings哪些风格正在起作用——以及本月的历史规律
Styles — cheap, high-quality, dividend-paying, recent winners — take turns leading. Ranks and history here are descriptions, not buy lists.低估值、高质量、高分红、近期赢家等风格轮流领跑。此处的排名与历史是描述，不是买入清单。
Factor weather因子风向
2026-05
July is usually a rough month for the market's recent winners — the momentum style fell in 7 of the last 10 Julys, including the last 5 straight, while the market itself was up in 100% of them.7月通常是市场近期赢家的困难月份——动量风格在过去 10 个7月中下跌了 7 次（含连续 5 年），而大盘在其中 100% 的时间上涨。
Watch — don't chase this month's hottest stocks.观望——本月不要追逐最热门的股票。
The hottest 10% of stocks right now: 46.0% sit in Information Technology.当前最热的10%个股中，46.0% 集中在Information Technology。
Hot stocks vs the market, last 20 days: -8.3%.热门股相对大盘（近20日）：-8.3%。
?Current-month per-factor seasonal stats — Cheap stocks (Value), Strong earners (Profitability), Careful spenders (Investment), Recent winners (Momentum), Small companies (Size). Stats from Ken French total-US-market academic portfolios (1926+). Definition mismatch: French HML/RMW/CMA are value-weighted long/short over the whole US market, NOT our S&P-1500 winsorized z-scores — this is the style's long-run seasonal CLIMATE, not a forecast of our daily factor leadership. Kept out of every score.当月各因子季节性统计——来自 Ken French 全美市场学术组合（1926年+）。定义差异：French 的多空因子与我们的 S&P 1500 z 分数不同——此为长期季节性气候，非每日因子领先预测，且绝不进入任何评分。
Long-run tendencies and descriptions — not forecasts, and kept out of every score.长期倾向与描述——不是预测，也不进入任何评分。
Factor calendar — what each month usually brings因子月历——每个月的历史规律
Style风格
Jan1月
Feb2月
Mar3月
Apr4月
May5月
Jun6月
Jul7月
Aug8月
Sep9月
Oct10月
Nov11月
Dec12月
Cheap stocks (Value)低估值（价值）
1.9
0.5
0.2
0.8
-0.1
-0.2
1.0
0.5
-0.1
-0.1
-0.2
-0.0
Strong earners (Profitability)高盈利（盈利质量）
-1.0
0.0
0.5
0.7
0.3
0.3
0.7
0.1
0.3
0.5
0.4
0.2
Careful spenders (Investment)稳健投资（投资）
0.6
0.6
0.5
0.2
-0.2
-0.2
0.3
0.3
0.2
0.2
0.2
0.3
Recent winners (Momentum)近期赢家（动量）
-1.5
1.1
1.1
0.2
0.7
1.0
-0.0
-0.1
1.1
0.9
0.8
2.2
Small companies (Size)小盘股（规模）
1.9
0.6
-0.0
-0.2
0.3
0.1
-0.4
-0.1
0.2
-0.8
0.3
0.0
Academic long/short style portfolios (Ken French, 1926+) — a climate read, not our live rankings, and never scored.学术多空风格组合（Ken French，1926年起）——气候参考，非实时排名，绝不计分。Full 100-year deep dive →完整百年数据 →
What is working now当前哪种因子在起作用?For each factor, the trailing-63-day return of its top quintile MINUS its bottom quintile. Positive = that factor has been rewarded recently (descriptive, not a forecast). Compare against the regime above: e.g. value & low-vol leading often accompanies risk-off / late-cycle.每个因子最近 63 个交易日内：顶部五分位减底部五分位的回报。正值=该因子近期获得回报（描述性，非预测）。
What the market has rewarded over the past 3 months. Descriptive — leadership rotates.过去3个月市场奖励的风格。仅为描述——领先风格会轮换。
Quality质量+6.3%
Earnings momentum (SUE)盈利动量 (SUE)+3.9%
Profitability盈利能力+3.4%
Investment投资+2.9%
Low accruals低应计+1.7%
Low short interest低做空兴趣+0.3%
Low volatility低波动-3.6%
Low beta (BAB)低贝塔 (BAB)-4.5%
Shareholder yield股东收益率-4.7%
Value估值-9.3%
Where the heat is热度集中在哪
Information Technology46.0%
Industrials20.0%
Materials8.0%
SNDKWDCLITEMUSTXINTCCIENTERCOHRLRCX
When the hottest stocks crowd into one sector, a stumble there hits them all together.当最热个股集中于单一板块时，该板块受挫会同时波及所有热门股。
A descriptive X-ray of where price momentum is concentrated right now — not a score, not a signal, and kept out of every ranking.对当前价格动量集中位置的描述性透视——不是评分，不是信号，且不进入任何排名。
Do these rankings actually predict? — the honest test这些排名真的能预测吗？— 诚实检验?The honest test most factor dashboards never show. At each quarter-end we rebuild every factor AS IT WAS KNOWABLE THEN — no look-ahead into a not-yet-filed 10-K (fundamentals are point-in-time) — then rank-correlate the factor score with the NEXT 63-day return: the Information Coefficient (IC). mean IC = average predictive edge; 0 = none. IC-IR (annualized) = IC ÷ its own volatility; t_HAC = Newey-West t-stat (overlapping forward windows autocorrelate the IC series); q_FDR = Benjamini-Hochberg false-discovery rate across every factor screened. One caveat the deep panel can't escape: prices cover only the currently-listed (delisted names are absent), so the deep read is survivorship-biased — an OPTIMISTIC bound. Judge each number against ZERO.大多数因子面板从不展示的诚实检验。在每个季末，我们按当时可知的信息重建每个因子（无前视、基本面为时点数据），再将因子分数与未来 63 日回报做秩相关，得到信息系数 IC。平均 IC = 平均预测优势；0 = 无优势。IC-IR（年化）= IC ÷ 自身波动；t_HAC = Newey-West t 值（重叠窗口使 IC 序列自相关）；q_FDR = 跨全部因子的多重检验错误发现率。深度面板的一个无法回避的代价：价格仅覆盖现存名（已退市名缺失），故深度读数含幸存者偏差 — 为乐观上界。每个数字都应与零比较。
Straight talk: of everything on this page, only Payout clears our honest prediction test — and only barely, on data tilted in its favour. Treat the rest as descriptions — what is cheap, what is high quality — not as buy lists.直说：本页所有排名中，仅「Payout」勉强通过我们的诚实预测检验——且数据本身对其有利。其余请视为描述——哪些便宜、哪些优质——而非买入清单。
Deep 2011-2026 test · survivor-tilted data, so treat results as an optimistic upper bound.2011-2026 深度检验 · 数据含幸存者偏差，结果应视为乐观上界。?Span 2011-03-31..2025-12-31 · 60 quarterly rebalances · ~1154 names — a DEEP 2011-2026 re-test on a backfilled close panel, not the ~3y rolling cache. The catch: that panel is SURVIVORSHIP-BIASED (delisted names absent — free prices only cover the currently-listed), so it is an OPTIMISTIC bound; a lone, marginal survivor is shown, not a green light. SUE — the strongest factor on the shallow 2023-2025 window — collapses to IC≈0 here. FINRA short-interest is dropped (no point-in-time history) and fundamentals refresh only ~annually, so fundamental factors still have few independent draws.区间 2011-03-31..2025-12-31 · 60 次季度再平衡 · 约 1154 只 — 2011-2026 深度复检。含幸存者偏差：已退市名缺失，为乐观上界；仅展示个别勉强通过者，而非买入信号。
⚠ The equal-weight composite has NEGATIVE predictive IC.等权综合因子的预测 IC 为负。Its blended mean IC is -0.0072 over 60 quarters — the naive all-factor blend INVERTS the cross-section on this panel, so it is worse than useless as a stock screen. Do NOT trade the composite. Judge each factor on its own row below: only a factor tagged VALIDATED (survives BH-FDR AND positive mean IC) has earned it, and even that is a marginal survivor on a survivorship-biased panel.其混合平均 IC 为 -0.0072（60 个季度）— 朴素的全因子等权混合在此面板上反转了横截面，作为选股信号还不如不用。切勿交易该综合因子。请按下方每个因子各自的行判断：仅标记为 VALIDATED（通过 BH-FDR 且平均 IC 为正）的因子才成立，且即便如此也只是幸存者偏差面板上的勉强通过者。
factor因子
mean IC平均 IC?average predictive edge; 0 = none平均预测优势；0 = 无优势
IC-IR (ann)IC-IR（年化）?IC ÷ its own volatility, annualized — how consistent the edge isIC ÷ 自身波动（年化）——优势的一致性
tHAC?Newey-West t-stat: accounts for overlapping return windowsNewey-West t 值：修正重叠窗口的自相关
hit命中率?fraction of quarters the IC was positiveIC 为正的季度占比
qFDR?Benjamini-Hochberg false-discovery rate across all factors screenedBH 错误发现率（跨全部因子）
Crowding / overlap拥挤／重叠?Factors that move together are redundant (a 'crowded' tilt). Mean absolute pairwise factor correlation, raw vs after Löwdin de-correlation; the composite (de-correlated) row above strips this overlap. VIF > 5 would flag a factor that is a linear combination of the others.彼此同向移动的因子是冗余的（拥挤的押注）。下为平均 |因子相关性|（原始 vs Löwdin 去相关后）；上方“综合（去相关）”行已剔除该重叠。VIF>5 表示某因子是其它因子的线性组合。:
mean |corr| 0.137 raw → 0.055 de-correlated.平均 |相关性| 0.137 原始 → 0.055 去相关后。low_vol ↔ low_beta · 0.74quality ↔ accruals · 0.71
Market breadth市场广度?Share of index members trading above their own 50-day AND 200-day moving averages, across the S&P 500 / 400 / 600 caps. This is MARKET-MEMBER participation (is the rally broad or narrow?), NOT factor-portfolio breadth. Above ~60% = broad; a small/large divergence flags where participation is rotating. Small & mid breadth history starts 2023-07 (the divergence is on that common window).指数成分中股价同时高于自身 50 日与 200 日均线的比例，覆盖标普 500／400／600。这是市场成分的参与度（涨势是广泛还是狭窄），并非因子组合广度。约 60% 以上=广泛；大小盘背离提示参与度在轮动。中小盘广度数据自 2023-07 起。
index指数
% > 50d MA% 高于50日线
% > 200d MA% 高于200日线
members成分数
Large cap (S&P 500)大盘 (标普500)
64.6%
67.9%
502
Mid cap (S&P 400)中盘 (标普400)
67.9%
71.2%
398
Small cap (S&P 600)小盘 (标普600)
73.8%
74.6%
604
Small-cap vs large-cap breadth: +9.2pp (small minus large, % above 50d). Positive = small-caps participating MORE than mega-caps — a broadening tape; negative = leadership narrowing to the top.小盘相对大盘广度：+9.2 个百分点（小盘减大盘，% 高于50日线）。正值=小盘参与度高于大盘（涨势扩散）；负值=领先向头部集中。
Factor performance — interactive因子表现 — 交互图?Rebased to 0% at the start of the selected window. Four relational views: factors LONG-ONLY (carry with the market), factors vs S&P 500, the Q5−Q1 PREMIUM (market-neutral), and cap-weighted GICS SECTORS vs S&P. Click legend entries to toggle series; pick a range. Powered by TradingView Lightweight Charts.在所选窗口起点重设为 0%。四种关系视图：因子纯多头、因子相对标普、Q5−Q1 溢价（市场中性）、以及市值加权 GICS 行业相对标普。点击图例切换；可选区间。
Broad = S&P 1500 (large+mid+small); Narrow = S&P 500 large-cap. This switch re-ranks only the chart, table & sector monitor below — the IC scorecard, crowding, quilt and breadth panels stay on the broad universe.广义=标普 1500（大+中+小）；狭义=标普 500 大盘。此切换仅重排下方图表、表格与行业监测 — IC 记分卡、拥挤度、拼布与广度面板保持广义口径。 · S&P 500 price history starts 2025-03 (~1y) — curves are shorter and σ/z need ~1y+, so some cells show —.标普 500 价格历史自 2025-03（约 1 年）— 曲线更短，σ/z 需约 1 年以上，部分单元格显示 —。
Performance table表现一览?Every factor across horizons, watchlist-style. SCOPE toggles factor long-only portfolios vs the Q5−Q1 premium; VALUE toggles raw return, return vs the S&P, and σ (the move z-scored vs that series' own trailing year — fixed horizons only). Click a column to sort.各因子横跨时间跨度的自选股式表格。SCOPE 切换纯多头与 Q5−Q1 溢价；VALUE 切换原始回报、相对标普、σ（相对自身过去一年的 z，仅限定窗）。点击列排序。
Sector monitor — cap-weighted GICS vs S&P行业监测 — 市值加权 GICS 相对标普?Each cap-weighted GICS sector's return RELATIVE to the S&P over each horizon, colored by how unusual that move is vs the sector's own trailing-year distribution (z). Deep color + |z|≥2 = an outsized rotation. The small number is the relative return; the big number is the z.每个市值加权 GICS 行业相对标普的回报（各时间跨度），按该走势相对其自身过去一年分布的异常程度（z）着色。颜色越深、|z|≥2 = 异常轮动。小字为相对回报，大字为 z。
Factor performance — point-in-time portfolios因子表现 — 时点组合?Each factor built into a daily PORTFOLIO at every month-end, as it was knowable then (no look-ahead): long-only = cap-weighted top quintile with a 5% single-name cap; long/short = equal-weight top-minus-bottom quintile (a paper factor premium, NOT a shortable book). Long-only carries with the market (compare SPY); the long/short isolates the premium. Free fundamentals are ANNUAL, so fundamental factor ranks refresh only ~once a year per name — most rebalance turnover is the price legs (low-vol/low-beta). History is ~3 years; every Sharpe ships with a block-bootstrap CI that mostly straddles zero, and NO factor clears FDR on the IC scorecard. Descriptive context, not validated or tradeable alpha.在每个月末，将每个因子构建为日度组合（无前视）：纯多头=市值加权顶部五分位（单股5%上限）；多空=等权顶部减底部五分位（纸面因子溢价，非可做空组合）。Free fundamentals are ANNUAL, so fundamental factor ranks refresh only ~once a year per name — most rebalance turnover is the price legs (low-vol/low-beta). History is ~3 years; every Sharpe ships with a block-bootstrap CI that mostly straddles zero, and NO factor clears FDR on the IC scorecard. Descriptive context, not validated or tradeable alpha.
Rebased over 2023-06-13→2026-07-17 (38 month-ends). Free fundamentals are ANNUAL, so fundamental ranks refresh only ~yearly — most turnover is the price legs (low-vol/low-beta). Every Sharpe ships with a block-bootstrap 95% CI that mostly straddles zero, and at most a lone, marginal factor clears FDR (see the IC panel above). Descriptive, not tradeable.区间 2023-06-13→2026-07-17（38 个月末）。免费基本面为年度数据，故基本面排名约每年才刷新一次 — 多数换手来自价格类因子（低波动/低贝塔）。每个夏普均附 block-bootstrap 95% 置信区间（多数跨越零），且至多个别勉强因子通过 FDR（见上方 IC 面板）。仅作描述，不可交易。
factor因子
long-only trend纯多头走势
LO cum纯多头累计
LO Sharpe纯多头夏普
Q5−Q1 cum多空累计
Q5−Q1 Sharpe (95% CI)多空夏普（CI）
P(Sharpe>0)P(夏普>0)
ValueValue
+85%
1.09
+30%
0.69 [-0.31, 1.66]
0.911
ProfitabilityProfitability
+26%
0.51
-25%
-0.82 [-2.1, 0.39]
0.103
QualityQuality
+46%
0.76
+7%
0.37 [-0.84, 1.63]
0.719
InvestmentInvestment
+39%
0.69
+13%
0.71 [-0.42, 1.82]
0.895
PayoutPayout
+55%
0.89
+1%
0.07 [-0.9, 0.97]
0.541
Low volatilityLow volatility
+28%
0.83
-44%
-1.08 [-2.25, 0.01]
0.026
Low betaLow beta
+22%
0.61
-43%
-0.71 [-1.79, 0.27]
0.079
CompositeComposite
+54%
0.93
-6%
-0.11 [-1.1, 0.84]
0.399
SPY (cap-wtd market)SPY（市值加权市场）
+74%
—
—
—
—
Recent moves — z vs trailing year近期变动 — 相对过去一年的 z?Each horizon's compounded return z-scored against that series' own trailing-252 distribution of overlapping returns (|z|≥2 = a notable move, marked ●). Green/red follows the sign of the move. Toggle the long-only portfolio vs the Q5−Q1 premium.各时间跨度的复利回报，相对该序列自身过去 252 日重叠回报分布做 z 标准化（|z|≥2 为显著，标 ●）。绿/红表示方向。可切换纯多头与多空溢价。
Factor crowding因子拥挤度?Correlation of the factor LONG/SHORT daily return series. Deep blue = the two premiums move TOGETHER (a crowded, redundant pair); red = they hedge each other. This is return co-movement (distinct from the static z-exposure overlap in the IC panel). With annual fundamentals the price legs (low-vol/low-beta) co-move most.因子多空日度回报序列的相关性。深蓝=两种溢价同向（拥挤、冗余）；红=互相对冲。这是回报的同向性（区别于 IC 面板中的静态暴露重叠）。
Quilt — Q5−Q1 monthly ranks拼布 — 多空月度排名?The periodic-table-of-returns: each column ranks the factor Q5−Q1 premiums best→worst that month (and a YTD column), cells colored by factor IDENTITY so you can trace one factor's path — hover a cell to highlight it across every column. ~13 months on the free cache; descriptive, not a regime-robust signal.收益周期表：每列按当月多空溢价从优到劣排名（含 YTD 列），单元格按因子着色以便追踪某一因子 — 悬停某格可高亮其在各列的位置。免费缓存约13个月；描述性，非稳健信号。
07
08
09
10
11
12
01
02
03
04
05
06
07
YTD
Payout+1%
Value+8%
Quality+1%
Quality+2%
Low beta+4%
Value+3%
Value+3%
Investmen+4%
Value+5%
Quality+2%
Quality+7%
Profitabi+1%
Low beta+14%
Investmen+10%
Value+1%
Composite+3%
Investmen-0%
Investmen+0%
Composite+3%
Investmen+2%
Investmen+2%
Low volat+0%
Quality+3%
Investmen+1%
Investmen-0%
Investmen+1%
Composite+12%
Value+6%
Profitabi+1%
Investmen+3%
Payout-2%
Composite-3%
Value+3%
Composite+2%
Payout-0%
Quality-0%
Payout+2%
Value-2%
Profitabi-4%
Value-1%
Value+10%
Quality+6%
Investmen-1%
Payout+2%
Profitabi-2%
Payout-4%
Investmen+2%
Payout+2%
Quality-1%
Low beta-0%
Composite+2%
Payout-2%
Value-4%
Payout-1%
Payout+9%
Payout+0%
Composite-1%
Low beta+1%
Value-3%
Value-4%
Low volat+2%
Quality+0%
Composite-2%
Composite-1%
Investmen+0%
Profitabi-3%
Payout-5%
Quality-3%
Low volat+9%
Composite-4%
Quality-2%
Quality-1%
Low beta-4%
Profitabi-4%
Payout+1%
Low volat-2%
Low volat-3%
Payout-2%
Low volat-2%
Composite-6%
Composite-5%
Low volat-3%
Profitabi+6%
Low volat-18%
Low volat-5%
Profitabi-1%
Composite-4%
Low volat-6%
Profitabi+1%
Profitabi-3%
Low beta-3%
Value-3%
Low beta-3%
Low volat-12%
Low volat-9%
Composite-3%
Investmen+2%
Profitabi-19%
Low beta-5%
Low volat-4%
Low volat-5%
Low beta-7%
Quality-0%
Low beta-4%
Profitabi-10%
Profitabi-5%
Profitabi-5%
Low beta-14%
Low beta-12%
Low beta-6%
Quality-1%
Low beta-24%
Multi-factor composite多因子综合?The RANKING here uses only the factor legs that carry a POSITIVE measured information coefficient on the deep-history (leak-free) scorecard, IC-weighted with a sign constraint — negative-IC / FDR-failing legs (e.g. low-vol, investment) are EXCLUDED from the order because they rank names in an anti-predictive direction (the blind equal-weight composite's own scorecard grades it ic_ir -0.049). The per-leg leaderboards below still show every leg, badged with its IC/FDR verdict. Validate-before-weight: the order a trader sizes off cannot be moved by a leg the scorecard rejects.此处排序仅采用在深度历史（无泄漏）评分卡上信息系数为正的因子腿，按 IC 加权并施加符号约束——负 IC／未通过 FDR 的腿（如低波动、投资）被排除，因为它们会使排序反向。下方各腿榜单仍展示全部腿并标注其 IC/FDR 结论。· ranked by排序依据: Value · Quality · Profitability · Shareholder yield · excluded (negative IC)已排除（负IC）: Investment, Low volatility
Insider conviction内部人信念SEC Form-4 ·SEC Form-4 · 6mo to 2026-03?Net OPEN-MARKET insider buying minus selling per company from SEC Form-4 filings (purchases code P, sales code S; grants and option exercises excluded). The SEC ships these as a quarterly bulk dataset, so this is the most recent COMPLETED quarter — a slow conviction read, not real-time. Clustered open-market BUYING is the more informative side (insiders buy for one reason; they sell for many).来自 SEC Form-4 的每家公司净公开市场内部人买入减卖出（买入代码 P、卖出 S；排除授予与期权行权）。SEC 按季度发布，故为最近一个完整季度 — 慢速信念信号，非实时。
Value估值?Cheap on fundamentals — a blend of earnings yield, book/price, sales/price and cash-flow yield (the cross-section Fama-French/Dimensional trade). Volatile and regime-dependent; book/price is weak for intangible-heavy firms.基本面便宜 — 盈利收益率、市净率倒数、市销率倒数与现金流收益率的综合。top前 25
Quality质量?A quality composite: high ROE, LOW accruals (cash-backed earnings — Sloan), and low leverage. AQR Quality-Minus-Junk style; tends to hold up in drawdowns.质量综合：高 ROE、低应计、低杠杆（AQR QMJ 风格）。top前 25
Profitability盈利能力?Gross profitability = gross profit / total assets (Novy-Marx). About as powerful as book/price historically, and it tilts the OPPOSITE way (toward growthier, higher-quality names).毛利润率 = 毛利润／总资产（Novy-Marx）。top前 25
Low volatility低波动?The low-volatility anomaly: low trailing total volatility has historically delivered market-like returns with less risk (USMV/SPLV capture this).低波动异象：低历史波动率。top前 25
Earnings momentum (SUE)盈利动量 (SUE)?Earnings momentum (SUE = standardized unexpected earnings). Latest quarterly EPS vs the same quarter a year ago, standardized by its own history — the post-earnings-announcement-drift effect: beats keep drifting up. Built point-in-time from free SEC EDGAR quarterly EPS frames; in our leak-free IC scorecard it is the STRONGEST positive factor and the one that survives the FDR control (validated on the 2023-2025 price window, like the rest of the zoo). ⚠️ A deep 2011-2026 re-validation (survivorship-optimistic) collapsed this edge to ~zero (IC 0.0005, HAC t 0.06, L/S Sharpe 0.09) — it held only on that shallow window (PEAD post-publication decay), so treat SUE as descriptive, not a validated scored factor.盈利动量（SUE = 标准化超预期盈利）。最新季度 EPS 对比去年同季、以自身历史标准化 — 盈利公布后漂移效应。基于 SEC EDGAR 季度 EPS 逐期点时构建；在防泄漏 IC 评分中（浅窗口）为最强正向因子且通过 FDR。⚠️ 深度 2011-2026 复验显示该边际接近零（IC 0.0005，t 0.06），仅在浅窗口成立 — 视为展示性因子，非已验证计分因子。top前 25
Low beta (BAB)低贝塔 (BAB)?Betting-Against-Beta (Frazzini-Pedersen): low-beta stocks have had high risk-adjusted returns because leverage-constrained investors bid up high-beta names.Betting-Against-Beta：低贝塔。top前 25
Low short interest低做空兴趣?Low short interest. High days-to-cover (short interest / avg daily volume, FINRA bi-monthly) negatively predicts the cross-section (Hong-Li-Rajan-Sherman) — leaders here are the LEAST shorted; the laggards are the crowded shorts.低做空。高回补天数（FINRA 双月）在横截面上负向预测。top前 25
How this research feeds the system — technical status本页研究如何接入系统——技术状态?Factor intelligence is context-only inside the Neural Web: it explains regime context and flags attention candidates for the cortex to consider, but it never ranks, scores, or originates signals (RUL-NW7). The BH-WITHHELD chip is mandatory until the 5-hypothesis family FDR sweep completes (per RUL-NW7, display-only interim).因子智能在神经网络中仅作为背景：解释市场体制背景，为皮层标记关注候选项，但从不排名、评分或发起信号（RUL-NW7）。BH-WITHHELD 标签为必须项，直至五假设家族 FDR 整体检验完成。
ACCRUING累积中
For the curious: what this factor research may influence elsewhere. Short version — it explains and observes; it never picks stocks or moves rankings.给好奇者：本页因子研究可影响系统的哪些部分。简言之——只解释与观察，从不选股或改动排名。
As of数据日期2026-07-20
· 10 dates个日期 · latest最新 2026-07-17floor pending基础待建Style regime风格体制mixedleader领先因子composite IC -0.0072Pair G (contradictions)G 对（矛盾）?Pair G is the Neural Web contradiction detector: it flags factor-vs-NW signal mismatches above a severity threshold. Dormant until the factor panel exceeds 60 dates (config/reflexes.yml §555). Severity policy: note is the current ceiling (H2 gate-passed required for tension; RUL-NW12).G 对是神经网络矛盾检测器：当因子与神经网络信号之间的不一致超过严重程度阈値时，进行标记。DORMANT休眠 · severity ceiling: note (H2 gate required for tension)严重程度上限：note（tension 需 H2 关卡）Factor attention因子关注度?The factor_attention reflex fires when a factor signal warrants cortex attention. Authority tier shows how much influence the attention reflex has been granted (A0/A1 = shadow only; granted = full attend allowed). Influence is earned through graded probation, never self-asserted.因子关注反射在因子信号值得皮层注意时触发。权限等级显示该关注反射被授予的影响力。tier等级: A0/A1 shadow · 0 firings次触发 · 0 graded已评级refused未授权Hypotheses H1–H5假设 H1–H5?H1=factor_weather to NW conditioning, H2=Pair G severity elevation, H3=cross-section alibi on board fires, H4=factor_attention earn-in, H5=factor_leader macro alignment. All carry BH-WITHHELD status until the 5-hypothesis family FDR sweep runs (RUL-NW6). PRE-FDR INTERIM = registered, accruing data, but not actionable before family BH.H1=因子天气→神经网络条件，H2=G 对严重程度提升，H3=买盘触发的横截面依据，H4=因子关注赚取，H5=因子领先者宏观对齐。H1: BH-WITHHELDH2: BH-WITHHELDH3: BH-WITHHELDH4: BH-WITHHELDH5: BH-WITHHELDMandatory interim: BH-WITHHELD on all 5 hypotheses until family FDR sweep completes (est. ≥2027). PRE-FDR INTERIM data is accruing.强制过渡期：所有 5 个假设的 BH-WITHHELD 状态将保持至家族 FDR 整体检验完成（预计≥2027）。PRE-FDR INTERIM 数据正在累积。Allowed influence允许的影响?What factor intelligence is and is not permitted to do inside the Neural Web, per RUL-NW7/NW9 and the constitution. Explain and shadow-attend are always permitted for context. De-escalation and ranking are permanently prohibited at this tier.根据 RUL-NW7/NW9 和章程，因子智能在神经网络内被允许和禁止的操作。解释和影子关注始终允许用于背景分析。去升级和排名在当前等级下永久禁止。explain — always解释 — 始终允许attend — shadow only关注 — 仅影子de-escalate — no去升级 — 禁止rank — never排名 — 永不
Source: SEC EDGAR XBRL 'frames' API (free, keyless) for FY2025 fundamentals, joined to the S&P 1500 price universe; 1497 names with a full composite as of 2026-07-17. Fundamentals refresh weekly; ranks recompute daily with prices. Factors are decayed/crowded and free fundamentals are imperfect — treat as a research lens, not a buy list. See research/QUANT_FACTOR_EXPANSION.md and LIMITATIONS.md.来源：SEC EDGAR XBRL 'frames' API（免费、无需密钥）的 FY2025 基本面，与标普 1500 价格域连接；截至 2026-07-17 有 1497 只具备完整综合分。基本面每周刷新，排名随价格每日重算。因子已衰减／拥挤且免费基本面并不完美 — 视作研究视角，而非买入清单。Generated生成于 2026-07-20 19:15 UTC