Methodology
Midas Edge brings institutional-style portfolio analytics to individual investors. Our methodology is rooted in academic research and established practices used by quantitative asset managers.
Our Approach
We believe individual investors deserve analytical context comparable to the tooling institutional investors use. Our factor model is based on the Barra USE4 methodology, adapted for accessibility without sacrificing rigor.
Every metric we compute has a clear definition, documented formula, and transparent interpretation. No black boxes.
Documentation
19-Factor Risk Model
Barra-style factor decomposition: portfolio exposures to the market and 18 style factors, plus a GICS sector term in attribution.
Holder IQ Scoring
Quantifying institutional investor skill from 13F filings to identify smart money signals and crowding risk.
Insider Signal Analysis
Scoring insider transactions from SEC Forms 3, 4, and 5 to detect informative vs. routine trading patterns.
Attention Queue Algorithm
Priority scoring system that surfaces positions requiring attention based on factor drift, signals, and events.
Risk Forecast Validation
How IC, bias ratio, and 95% VaR coverage are monitored against realized outcomes and freshness windows.
Anytime-Valid Evidence
How e-values keep continuously monitored research claims statistically valid as new filings and cohorts arrive.
Tax-Loss Harvesting Guardrails
How Midas Edge identifies unrealized losses, suppresses wash-sale-risk suggestions, and documents tax workflow limits.
Red Flags Detection
Automated monitoring of SEC filings for disclosure events that historically correlate with material stock underperformance.
Portfolio Health Score
Composite 0-100 score summarizing portfolio quality across factor balance, smart money alignment, and risk factors.
Method Version Changelog
Concise changes to the model, scoring rules, and prompt templates behind user-facing analytics.
Portfolio health score
health-score:2.1
Insider sentiment counts only open-market purchases and sales, as the stock page and watchlist do. Option exercises, stock awards, tax withholding and gifts no longer read as insider buying or selling.
Affects: portfolio_health, dashboard, score_detail
Portfolio health score
health-score:2.0
Company-specific signals exclude funds, cash and unresolved company identities. History and references compare only the same persisted computation method.
Affects: portfolio_health, dashboard, score_detail
Factor model
v1.2.0
Production Barra-style model pinned for portfolio and ticker analytics.
Affects: factor_exposures, risk_attribution, performance_attribution, portfolio_health
Holder IQ scoring
holder-iq-route:1.0
Initial route-level Holder IQ scoring disclosure with 13F freshness context.
Affects: holder_iq, smart_money, portfolio_health
Insider signal rules
insider-route:1.0
Initial insider-signal scoring disclosure with role and transaction context.
Affects: insider_signals, portfolio_health, alerts
Portfolio health score
health-score:1.0
Composite score disclosure pinned to the current grade-threshold configuration.
Affects: portfolio_health, dashboard, score_detail
Factor model
v1.1.0
Expanded style-factor coverage and stricter provenance checks.
Affects: factor_exposures, risk_attribution, performance_attribution, portfolio_health
Metric Glossary
Short, action-oriented definitions for the core metrics we surface throughout the dashboard and reports.
Insider Activity
Insiders
Recent buying or selling by corporate executives, directors, and major shareholders as reported in SEC Form 4 filings.
Why it matters: Insiders have deep knowledge of their company; unusual buying often precedes positive news, while heavy selling can signal concerns.
Holder IQ Percentile
Smart Money
Percentile rank (0-100) of institutional holder quality based on Holder IQ scores.
Why it matters: Higher percentiles signal overlap with consistently skilled managers; sharp drops can flag crowding unwind risk.
Smart Money Momentum
Smart Money
Direction of net changes by top-decile managers (accumulating, holding, or distributing).
Why it matters: Shows whether skilled managers are adding to or exiting a position.
Exit Pressure
Smart Money
Weighted rate of smart-money exits based on net share reductions among top holders.
Why it matters: Rising exit pressure often precedes volatility or crowded unwind scenarios.
DP Z-Score
Crowding
Z-score of disproportionate popularity using log-odds for over/under-ownership within a segment.
Why it matters: Positive values indicate segment crowding or conviction; negative values indicate avoidance.
DP Odds Ratio
Crowding
Odds ratio of a filer segment holding the security versus the rest of the market.
Why it matters: Highlights when a specific cohort disproportionately favors a security.
Crowding Score
Crowding
Normalized crowding risk score (0-100) derived from ownership concentration and smart-money overlap.
Why it matters: Higher scores indicate crowded ownership and elevated unwind risk.
Crowding (HHI)
Crowding
Herfindahl-Hirschman Index of institutional ownership concentration (0-10,000 scale).
Why it matters: High concentration raises risk of forced selling and correlated exits.
Churn Rate
Flows
Quarterly holder turnover: (entries + exits) / (2 * average holders).
Why it matters: High churn suggests unstable ownership and higher price volatility.
Whale Flow
Flows
Quarter-over-quarter change in exposure among the largest filers.
Why it matters: Large whale flows can materially impact liquidity and price pressure.
Value vs Presence Divergence
Flows
Stealth score capturing when position value changes faster than holder count.
Why it matters: Signals stealth accumulation or distribution even when holder counts look stable.
Style Drift
Factors
Change in factor exposures over time relative to prior snapshots or benchmarks.
Why it matters: Large drift means the portfolio is behaving differently than intended.
Factor Imbalance
Factors
Degree of tilt away from neutral or benchmark exposures across the 19-factor model.
Why it matters: Imbalances drive unintended risk and can dominate performance.
Portfolio Health Score
Portfolio
Composite 0-100 score summarizing factor balance, diversification, smart-money alignment, and risk flags.
Why it matters: Use the score as a triage aid, not a trade signal: low scores tell you where to inspect exposure, concentration, or risk first.
Attention Score
Portfolio
Priority score from 0-100 that ranks positions by signal severity, position impact, recency, and actionability.
Why it matters: Higher scores help you focus review time on positions with the strongest combination of risk, change, and portfolio relevance.
Data Freshness
Portfolio
How recently the underlying brokerage, market, filing, or factor data was refreshed for the metric you are viewing.
Why it matters: Freshness tells you whether a number reflects current portfolio state or a stale upstream dataset that should be refreshed or treated cautiously.
Net Value
Portfolio
What the account would be worth if everything were settled today: holdings you own plus cash, minus the market value of your shorts and anything borrowed from your broker.
Why it matters: Net value is your own money in the account. Gains and losses on everything you hold land on it, so it moves more than the market when you borrow.
Margin Loan
Portfolio
Money borrowed from your broker against the securities in a margin account. Brokers report it as a negative cash balance, also called a margin debit.
Why it matters: A margin loan adds to what you hold without adding to your own money, accrues interest, and is repaid before you see any proceeds.
Portfolio Leverage
Portfolio
The market value of what you own divided by your net value. 1.0x means you hold exactly your own money's worth; 1.5x means every $1 of your money controls $1.50 of holdings. Gross leverage also adds the size of your shorts.
Why it matters: Leverage multiplies moves in both directions: at 1.5x, a 10% fall in what you own takes about 15% off your net value.
AI Confidence
AI
0-100 estimate of signal consistency across the data sources used to generate an AI-assisted explanation.
Why it matters: Confidence helps calibrate how much context to demand before acting; it is not a probability that the recommendation is correct.
Information Coefficient
Validation
Correlation between a model signal and later realized returns over the validation window.
Why it matters: Positive IC means the signal has historically ranked future outcomes in the intended direction; weak or negative IC lowers trust in that signal.
Bias Ratio
Validation
Calibration check comparing realized outcomes against the model's expected distribution or directional claim.
Why it matters: Large bias ratios indicate the model may be systematically overstating or understating an effect and should be treated conservatively.
95% VaR Coverage
Validation
Share of observations that fall inside the model's 95% value-at-risk or prediction interval.
Why it matters: Coverage far below 95% means risk is understated; coverage far above 95% may mean the model is too conservative to be useful.
R-Squared
Validation
Fraction of return variation explained by the factor model or validation regression.
Why it matters: Higher R-squared means the model explains more of the observed movement; low values imply more idiosyncratic or missing drivers.
Broad market
Factors
The stock model's common market component, with a base loading of 1 for each stock: the market intercept, exactly 1.0 for every security on every date, so its factor return is the move all stocks share that day. Portfolio loadings also reflect position direction and any supported leverage adjustment. It does not vary across stocks; how strongly each stock has moved with the market is the separate Historical Beta (BETA) factor.
Why it matters: Separates the common market component from differences in historical market sensitivity. A factor loading is not a percentage of your portfolio or a forecast return.
Size (Market Cap)
Factors
Company size measured by log market capitalization (the natural log of market cap on the model date), scored against the cap-weighted average company of the stock's own GICS sector. Positive values are larger than that sector average and negative values smaller, so a mega-cap in a sector of mega-caps scores near 0, below a smaller company that is large for its own sector.
Why it matters: Small caps historically outperform but with higher volatility. Size tilts drive meaningful return differences over time.
Size Non-Linear
Factors
Non-linear size: a Gaussian bump exp(-s²/2) of the scaled log market cap s (log cap minus its cross-sectional median, divided by its standard deviation), re-centered on its median and orthogonalized against SIZE with market-cap weights. It peaks for mid-size companies.
Why it matters: Extreme size exposures (micro-caps or mega-caps) behave differently than mid-caps. This factor captures those edge effects.
Book-to-Price
Factors
Value: book equity divided by market capitalization, capped at ±5 (negative book equity is kept). Higher values indicate value stocks, lower values growth stocks.
Why it matters: Value vs growth is one of the most persistent factor tilts. Value stocks often outperform long-term but can underperform for extended periods.
Earnings Yield
Factors
Earnings relative to price: 75% forward FY1 EPS / price (point-in-time analyst estimates), 15% trailing-12-month operating cash flow per share / price and 10% trailing-12-month EPS / price, renormalized over the legs a name has. Financials drop the cash leg; real-estate names have no value in v1.2.0. Known data discontinuity (2026-09-28): until 2026-09-25 only 234 of 3,868 universe names (6%) had a point-in-time forward estimate vintage (from a late-June 2026 snapshot), so EARNYILD was renormalized onto trailing cash flow and trailing EPS for the other 94% of names; daily forward-vintage capture began on 2026-09-26 (58% forward coverage on 2026-09-28, 89% on 2026-10-01, 98% by 2026-10-23), causing a one-time cross-sectional rank shift on 2026-09-28 (day-over-day rank correlation 0.845 vs 0.999 median).
Why it matters: High earnings yield stocks are often undervalued and may offer better risk-adjusted returns. Low yields signal growth expectations.
Dividend Yield
Factors
Trailing-12-month dividends per share divided by price. Confirmed non-payers are 0 and missing dividend or price data stays missing. v1.2.0 has no point-in-time forward dividend estimates, so the trailing yield carries the full weight.
Why it matters: Dividend stocks provide income and often exhibit lower volatility. High dividend tilts can indicate defensive positioning.
Growth
Factors
Earnings and revenue growth: 70% forward EPS growth (FY2 vs FY1 analyst estimates), 20% five-year historical EPS growth (CAGR) and 10% revenue growth (forward estimates where available, otherwise trailing-12-month year over year), renormalized over the legs a name has. Known data discontinuity (2026-09-28): until 2026-09-25 only 234 of 3,868 universe names (6%) had a point-in-time forward estimate vintage (from a late-June 2026 snapshot), so GROWTH was renormalized onto five-year historical EPS growth and revenue growth for the other 94% of names; daily forward-vintage capture began on 2026-09-26 (58% forward coverage on 2026-09-28, 89% on 2026-10-01, 98% by 2026-10-23), causing a one-time cross-sectional rank reshuffle on 2026-09-28 (day-over-day rank correlation 0.204 vs 0.998 median).
Why it matters: Growth stocks can deliver outsized returns but often trade at premium valuations, creating downside risk if growth disappoints.
Leverage
Factors
Debt load: 75% market leverage, net debt / (net debt + market cap) with net debt = total debt minus cash floored at 0; 15% total debt / total assets; 10% total debt / book equity (positive book only); renormalized over the legs a name has.
Why it matters: Leverage amplifies both gains and losses. High leverage portfolios face greater risk during economic stress.
Historical Beta
Factors
The model's standardized historical-beta characteristic: a 252-trading-day regression of the stock's excess return on the SPY benchmark's (IVV, then VOO, as fallbacks), Dimson-adjusted with one lag and one lead for asynchronous trading. v1.2.0 applies no shrinkage. This, not the broad-market intercept, is the factor that carries market sensitivity.
Why it matters: Historical beta captures how a stock has moved with the market over time. It's a key input for estimating expected returns and portfolio risk.
Beta Non-Linear
Factors
Non-linear beta: the square of the scaled beta b (beta minus its cross-sectional median, divided by its standard deviation) minus 1, re-centered on its median and orthogonalized against BETA. It is high for both very high- and very low-beta stocks.
Why it matters: Extreme beta stocks (very defensive or very aggressive) have unique risk characteristics not captured by linear beta.
Residual Volatility
Factors
Stock-specific volatility: 75% the 252-day standard deviation of residuals from the stock's market regression, 15% Parkinson high-low price-range volatility over 252 days and 10% an EWMA of residual volatility (42-day half-life), renormalized over the legs a name has.
Why it matters: High residual volatility means more stock-specific risk. These positions need more attention and may require larger position limits.
Liquidity
Factors
Share turnover: a blend of the log of total volume divided by average float shares over 21 days (35%), 63 days (35%) and 252 days (30%). Higher values mean more actively traded shares.
Why it matters: Illiquid positions are harder to exit during stress. High liquidity exposure typically indicates capacity for larger trades.
3-Week Momentum
Factors
Short-term momentum: the sum of the stock's daily log returns over the 15 trading days ending the day before the model date (the model date's own return is skipped).
Why it matters: Short-term momentum often reverses (mean reversion), making high exposure risky. Useful for tactical timing.
11-Month Momentum
Factors
Intermediate-term price momentum: the sum of daily log returns over 231 trading days (about 11 months) ending 42 trading days (about two months) before the model date, so recent weeks are excluded from the measurement to reduce short-term reversal effects.
Why it matters: Momentum is one of the strongest and most persistent factors. Winners tend to keep winning over 6-12 month horizons.
5-Day Reversal
Factors
Recent price reversal, using the negative of a five-day price return before the model date: minus the sum of daily log returns over the 5 trading days ending the day before. This factor is not your portfolio's five-day return.
Why it matters: Distinguishes exposure to recent price declines from exposure to recent winners. Reversal is a model characteristic, not a promise that a price move will reverse.
Short Interest
Factors
Shares sold short as a fraction of float shares, from the latest short-interest record known on the model date, capped at 150%.
Why it matters: Heavily shorted stocks can squeeze higher or validate bear cases. Understanding short exposure helps manage crowding risk.
Hedge Fund Ownership
Factors
Fraction of shares outstanding held by hedge-fund filers, from 13F holdings classified by filer segment (archived provider holder lists known by the model date as the fallback), capped at 150%.
Why it matters: High hedge fund ownership creates crowding risk - when hedge funds exit together, selling pressure can be severe.
Passive Ownership
Factors
Fraction of shares outstanding held by passive vehicles (index funds and ETFs), from 13F holdings classified by filer segment (archived provider holder lists known by the model date as the fallback), capped at 150%.
Why it matters: High passive ownership provides stable demand but can reduce price discovery. Index inclusion/exclusion events create significant flows.
Share Issuance
Factors
Share issuance: 70% the 12-month log change in float shares plus 30% the 12-month change in float shares divided by the 12-month change in market cap. A name with a split in the trailing 12 months has no value, and one with a buyback in that window is set to 0, so share reductions are not scored; higher means more dilution. Known data gap: the stored v1.2.0 exposures are 0 on every row for all 385 dates from 2024-11-04 through 2026-05-21 (measured 2026-10-21), so the factor carried no information in that window.
Why it matters: Issuance transfers value from existing holders, and net issuers have historically lagged companies that shrink their share count.
Important Disclaimer
The methodologies described here are for educational and research purposes. Factor exposures, alpha estimates, and signal scores are analytical tools, not trading recommendations. Past performance of any factor or signal does not guarantee future results. Always consult a qualified financial advisor before making investment decisions.