Equity-sleeve coverage100.0%99.8% covered of 99.8% retained equity weight.
Total-fund coverage99.8%Company fundamentals as a share of 100.0% reported fund weight; no reallocation.
Non-company exposure0.2%Cash, derivatives, and other non-equity rows are excluded · tradingview_etf_holdings · Jul 17, 2026 · current.
Complete mapped company sleeve33/33 mapped companies have stored fundamentals, covering 100.0% of the retained company sleeve and 99.8% of fund weight. The remaining 0.2% is non-company exposure, not missing companies.
KWEB — KraneShares CSI China Internet ETF factor map
Return factors and risk factors built from stored company fundamentals, price tape, themes, event evidence, liquidity, and flow datasets.
Companies33Ranked inside this universe
Factors5034 ready · 16 partial
Coverage98%Average ready-factor coverage
As ofJul 21, 9:30 AM GMT+8Latest stored factor input
Full factor lab detail queuedThe initial factor map is loaded from the compact snapshot; full histories, heatmaps, validation matrices, and lenses hydrate on demand.
Factor workflow
Navigate the factor engine without scrolling the whole report
Start with attribution quality, move into the action book, then inspect individual factor evidence only when needed.
Factor exposure engine
Explain KWEB — KraneShares CSI China Internet ETF through true factor returns
Daily stock-level cross-sectional regressions estimate factor returns from company returns and reusable economic factor exposures. Sector, industry, theme, and country are reporting lenses, not dummy factors used to force-fit the universe return.
For each stored trading day, company returns are regressed against reusable company-level factor exposures. The common market column captures the broad universe tape; style factors explain tilts around that tape.
Lens policyNo fake fit from buckets
Sector, industry, country, and theme are reporting lenses only; they are not dummy variables used to force-fit returns.
Attribution splitWhat is investable versus only explanatory
17 usable factors · 31 diagnostic factors
Common tape+0.84%Broad universe move before active factor interpretation.Rewarded / usable factors-1.25%9 rewarded · 8 risk-controlDiagnostic factors-0.41%29 diagnostic · 2 blockedResidual-0.02%Idiosyncratic return that should flow into company-level research.
Attribution policy. Separate common tape, investable rewarded factors, diagnostic factors, and residual. Diagnostic factors explain movement but are not portfolio-sizing inputs.
Factor action bookWhat to overweight, hedge, block, or research next
Overweight · -0.49% actionable return
Active tilts9Risk controls8Blocked2Research-only29Regime-adjusted9
Small / validation pendingConviction 67 · risk budget 3%
Contribution
-0.35%
Factor return
+1.11%
Exposure
-0.31
Drawdown repair is a rewarded factor with negative contribution; reduce high exposure until payoff repairs.
Regime-aware sizingSmall / validation pending
No stored regime evidence · best n/a · worst n/a · n/a% survival
No stored regime validation yet; keep sizing small until regime behavior is refreshed.
Coverage passValidation watchPayoff watchCrowding watchRisk budget passRegime fit watch
Use the action book as a review order. Active tilts require validation, breadth, uniqueness, controlled crowding, and regime survival; diagnostic rows remain research-only.
Model fit disciplineWhen factor attribution is reliable enough to use
Usable factor fit
Current stock-level fit74%Weighted R-squared across latest company returns.Average daily fit55%How often the model explains day-to-day cross-sectional moves.Fit stability61%Share of stored sessions with usable stock-level factor fit.Residual cost+4.42%Average stock-level error; gross residual +2.85%.
Not an accounting identity. R-squared is a stock-level fit diagnostic, not a promise that factors explain every portfolio return dollar. The residual board remains the source of company-specific work.
Use factors as the first pass, then verify residual groups and company-specific outliers before sizing.
Factor trust dashboardCan this attribution guide positioning?
Usable with checks
Trust score55
Use smaller factor tilts; validate residual clusters and crowded factors before sizing.
Positioning guidanceUsable with checks
Use smaller factor tilts; validate residual clusters and crowded factors before sizing.
Weakest fit dayJul 10, 2026
16% fit · RMSE +0.62%
Residual-dominated days14
Latest fit 80% · average 55%
Required checks before sizing0 passed
Stock-level fitwatch55 · Stock-level fit is usable but should be checked against residual outliers.
Residual loadwatch33 · Residual sleeve is material; check single-name outliers before sizing factors.
Validation depthwatchn/a · Run forward validation before treating factor payoffs as durable.
Crowding controlfail100 · Top factors or redundant factors dominate; cap position size and avoid double counting.
Diagnostic factor loadwatch39 · Diagnostic-only factors are meaningful; use them for triage rather than direct sizing.
Failure modes7 checks
Weak stock-level fitmediumThe cross-sectional model is not consistently explaining stock returns.
Weakest fit dayhigh2026-07-10 had only 16.0% stock-level explanatory fit.
Residual loadmediumIdiosyncratic return is large enough that factor views need company-level confirmation.
Validation depthmediumForward validation is not strong enough to treat every factor payoff as durable.
Crowding controlhighA small number of sleeves or overlapping factors can dominate the attribution.
Diagnostic factor loadmediumToo much movement is explained by diagnostic-only factors that should not be direct sizing inputs.
Factor reliability boardWhich factors are investable, confirming, or noisy
54 average reliability · 0 decision-grade factors
Decision-grade tilts0Validated enough for active factor interpretation
No factor clears all reliability gates yet.
Usable confirmation4Useful when independent evidence agrees
Sentiment reversal setupUse as a confirming factor alongside other independent evidence.
Reliability
72
Validation
n/a
Crowding
60
Drawdown repairUse as a confirming factor alongside other independent evidence.
Reliability
64
Validation
n/a
Crowding
71
ValueUse as a confirming factor alongside other independent evidence.
Reliability
60
Validation
n/a
Crowding
65
Theme alignmentUse as a confirming factor alongside other independent evidence.
Reliability
58
Validation
n/a
Crowding
74
Noisy or incomplete7Needs better validation, coverage, or stability
SizeUse as a descriptive exposure until forward tests are populated.
Reliability
57
Validation
n/a
Crowding
55
Shareholder yieldUse as a descriptive exposure until forward tests are populated.
Reliability
55
Validation
n/a
Crowding
73
Growth accelerationUse as a descriptive exposure until forward tests are populated.
Reliability
53
Validation
n/a
Crowding
60
LeverageUse as a descriptive exposure until forward tests are populated.
Institutional flowUse only as a warning or confirmation; avoid double-counting this factor.
Reliability
47
Validation
n/a
Crowding
99
Thematic demandUse only as a warning or confirmation; avoid double-counting this factor.
Reliability
54
Validation
n/a
Crowding
99
Defense rearmament alignmentUse only as a warning or confirmation; avoid double-counting this factor.
Reliability
52
Validation
n/a
Crowding
99
Revenue acceleration leadershipUse only as a warning or confirmation; avoid double-counting this factor.
Reliability
44
Validation
n/a
Crowding
99
No factor currently clears the decision-grade validation, stability, and uniqueness gates.Some factors are highly correlated with peers, so do not add their attributions together mechanically.Some factors are descriptive only until forward validation or coverage improves.
Tracker