What is quantitative stock analysis?

Quantitative stock analysis uses mathematical models, statistical methods, and machine learning algorithms to evaluate securities. StoQuant combines 149 features including technical indicators, fundamental ratios, sentiment signals, and macro data to generate alpha scores for 3,500+ stocks.

How does the hidden gem screener work?

The hidden gem screener uses the Benjamin Graham intrinsic value formula (V = EPS x (8.5 + 2g)) to identify undervalued small-cap stocks. It filters for P/E under 15, P/B under 1.5, PEG under 1.0, debt-to-equity under 0.5, and flags stocks trading at least 30% below intrinsic value with fewer than 8 analyst coverage.

What machine learning models power the predictions?

StoQuant uses a 3-model stacking ensemble: LightGBM, CatBoost, and XGBoost with a logistic regression meta-learner. Predictions are calibrated with Venn-ABERS conformal prediction. The system uses 149 features spanning technical, fundamental, sentiment, macro, and alternative data signals.

How does portfolio optimization work?

StoQuant integrates ML price forecasts with the Black-Litterman portfolio optimization model. ML predictions serve as investor views (Q vector) with model confidence intervals mapped to uncertainty (omega matrix). This produces optimal portfolio allocations that blend market equilibrium with AI-driven insights.

Is my portfolio data secure?

Yes. StoQuant uses JWT authentication with timing-safe comparisons, rate limiting (300 req/min per IP), secure headers (HSTS, X-Frame-Options, X-Content-Type-Options), and portfolio ownership verification on every request. No sourcemaps are exposed in production.

What data sources does StoQuant use?

StoQuant aggregates data from SEC EDGAR filings, FMP financial data, social media sentiment (Reddit, StockTwits, Twitter), Google Trends, insider trading records, institutional holdings, options flow, and macro economic indicators from FRED. All data is stored as append-only time series.

Does StoQuant work with Claude Desktop or Claude Code?

Yes. StoQuant ships an official Model Context Protocol (MCP) server, stoquant-mcp, available on npm. One command — npx -y stoquant-mcp install — wires Claude Desktop, Claude Code, or any MCP-compatible client into roughly 30 read-only stock-research tools: prices, Q-Score, ML predictions, hidden-gem screener, insider trades, market regime, and macro dashboards. Power-tier subscription required.

What is a free alternative for deep stock research?

StoQuant is a free-tier quantitative research platform for deep single-stock analysis. It scores US stocks daily with a Q-Score (0-100) whose live out-of-sample record is published on /proof, including when it underperforms, and produces a full evidence-based research memo for any ticker covering valuation, momentum, analyst consensus, insider activity, and machine-learning signals. It is a free alternative to Seeking Alpha, Simply Wall St, and TipRanks for investors who want the underlying proof, not just a rating.

How can I screen for undervalued hidden-gem stocks using alternative data?

StoQuant's hidden-gem screener blends classic value filters — the Benjamin Graham intrinsic value formula, P/E under 15, P/B under 1.5, and a 30% margin of safety — with alternative-data signals most screeners ignore: insider-buying clusters, SEC EDGAR full-text filings, credibility-weighted social sentiment, Google Trends attention, and short interest. This surfaces under-covered small-cap stocks (fewer than 8 analysts) before Wall Street catches on. The screener is free to use.

Q-Score Methodology: IC-Weighted / Fail-Closed Dimensions

Learn how StoQuant's Q-Score (0–100) blends valuation, momentum, sentiment, and machine learning to rate 3,500+ US stocks every day.

Active weights vs fail-closed dimensions

The Q-Score is StoQuant's flagship single-number rating for any stock. It ranges from 0 (avoid) to 100 (strong buy) from IC-weighted dimensions (and fail-closed ones at weight 0). Each dimension is independently scored on a 0–100 scale; only non-zero weights move the composite. Live ranking edge is on /proof — while F / inverted, treat scores as descriptive.

Why multiple dimensions? Because no single metric captures the full investment opportunity. Valuation alone ignores momentum (a cheap stock can get cheaper); momentum ignores fundamentals (a stock in a dying sector may rally anyway); sentiment ignores earnings quality (optimistic headlines can mask accounting tricks). By blending nine independent viewpoints, the Q-Score becomes more robust than any single approach. It is also more interpretable: you can open the score card and see which dimensions are driving the recommendation.

The nine dimensions are:

Valuation: Is the stock cheap relative to earnings, book value, cash flow, and the Benjamin Graham formula? Targets stocks with P/E < 15, P/B < 1.5, and 30%+ margin of safety.

Momentum: Is the stock gaining or losing relative momentum? Tracks 5-day, 20-day, and 63-day technical momentum plus sector-relative strength. Momentum confirms valuation (a cheap stock with rising momentum is a strong signal).

Technical: Are there chart patterns or breakout signals? Looks for volume-confirmed breakouts, squeeze setups (Bollinger Band compressions), and price-pattern reversals.

Sentiment: What do news headlines and SEC filings say? A hybrid keyword-plus-FinBERT classifier grades headlines as bullish, bearish, or neutral. Sentiment from recent earnings transcripts is also scored.

Analyst: What do Wall Street analysts recommend? Tracks consensus rating (buy/hold/sell), recent upgrades/downgrades, and estimate revisions. Analyst upgrades often precede price moves.

Insider: Are company insiders buying or selling? Large insider buys ($5M+) are one input among the nine. This page does not state a hit rate or a forward return for them. Read the published track record on /proof.

Earnings Quality: Is the company's reported earnings "real"? Checks free cash flow vs. net income, accruals (accounting adjustments), R&D capitalization, and Piotroski F-Score. Filters out companies using accounting gimmicks.

Machine Learning: What do the 149-feature ML ensemble predictions say? LightGBM, CatBoost, and XGBoost predict 30-day forward returns. The ML dimension captures non-linear relationships and feature interactions that traditional scoring misses.

Gem Discovery: Is the stock a hidden gem? Looks for small-cap undervaluation, analyst coverage gaps (< 8 analysts despite quality fundamentals), and insider confidence. Hidden gems often have disproportionate upside when discovered by the market.

How it works

Score each dimension independently — For each stock, compute 0–100 scores for every dimension the scorer knows. Only dimensions with non-zero IC weight enter the composite; fail-closed dims stay at weight 0. Momentum depends on recent price trends and sector relative strength; DCF/valuation uses margin-of-safety style inputs when available.
Weight by reliability — Not all dimensions are equally predictive in all market regimes. In Bull markets, momentum is reliable; in Bear markets, value is reliable. StoQuant learns these regime-conditional weights from historical walk-forward data. Weights change daily based on the current HMM regime.
Read the score as a rating, not a probability — A Q-Score of 75 is a point on the 0–100 scale. It is not a 75% probability of outperformance. Do not treat the number as calibrated odds. The published track record is on /proof — judge that page instead of inferring skill from the score.
Map to recommendations — Map the final Q-Score to a recommendation category using the same bands the scorer uses: 75+ Strong Buy, 58+ Buy, 42+ Hold, 28+ Sell, otherwise Strong Sell. Also show confidence (high/medium/low) based on the spread of dimension scores.
Display the driver scorecard — Show the user not just the final Q-Score but also the top 3 contributing dimensions and their scores. For example: "Q-Score 78 (Buy). Top drivers: ML 88 (buy signal), Insider 82 (mega-buy detected), Valuation 72 (30% margin of safety)."

Weighting and Adaptive Regimes

The Q-Score weighs each dimension based on historical out-of-sample predictive power. This is where the walk-forward validation described in our backtesting article becomes crucial: StoQuant measures the Information Coefficient (correlation with actual forward returns) for each dimension every 20 days. Dimensions with higher IC get higher weights.

However, dimension predictiveness varies by regime. In our audit logs, we found that the Valuation dimension has IC = +0.12 in Bull/Range markets but IC = -0.05 in Bear markets (meaning overweighted cheap stocks actually underperform in downturns—a phenomenon known as "value traps"). To adapt, StoQuant computes regime-conditional weights: in Bull regimes, Valuation gets 28% weight; in Bear regimes, it drops to 15%, and other dimensions (Quality, Insider, Earnings) get boosted.

A Q-Score is a rating, not a probability. A score of 75 does not mean a 75% chance of outperformance, and it is not calibrated odds. Read /proof for the published track record and judge that page yourself.

Sentiment and Analyst dimensions require special handling. Sentiment can be noisy (one negative headline can spike bearish sentiment even if fundamentals are sound). To reduce noise, StoQuant applies a multi-day exponential smoothing: recent headlines are weighted more, but extreme swings are dampened. Analyst consensus has its own lag—upgrades often lag technical momentum by weeks. StoQuant models this lag and credits upgrades slightly less in the first 5 days after they occur.

The Gem Discovery dimension is perhaps the most unique. While traditional ratings focus on the best-known, highest-quality stocks, StoQuant deliberately highlights small-cap anomalies: profitable companies with $250M–$10B market caps, less than 8 analyst coverage, and valuation metrics suggesting 30%+ upside. These are the "hidden gems" that Wall Street has overlooked. The Gem dimension gets boosted when insider buying, earnings surprises, and technical breakouts all align—signals that the market may be about to discover the stock.

Related on StoQuant

Use the Q-Score in action: Best Stock Screener (stoquant.com/best-stock-screener) and AI Stock Picks (stoquant.com/ai-stock-picks). See today's scores: Today's Top Q-Score (stoquant.com/today/top-q-score).

FAQ

What does a Q-Score of 75 mean?

A Q-Score of 75 is a point on the 0–100 scale. The scorer maps 75+ to Strong Buy. It is not a 75% probability of outperformance. The dimension breakdown shows which inputs are high or low. For any track record, read /proof.

How often does the Q-Score update?

After each market close the engine rescores a rotating slice of the 3,500+ US stocks we cover, cycling the whole list over about a week; opening a stock's research page recomputes its score on the spot. Each dimension's input data is refreshed on its own cadence: stock prices (intraday), sentiment (continuously), analyst ratings (daily when published), and fundamentals (quarterly, but cached).

What is the difference between Q-Score and analyst consensus?

Analyst consensus is a simple average of Wall Street ratings (buy/hold/sell). Q-Score blends IC-weighted dimensions (technicals, momentum, quant_analytics contrarian, analyst, ml, earnings_quality, dcf); gem/social/moat/catalyst are fail-closed at weight 0 until residualized IC recovers. Analyst consensus lags; Q-Score updates on the daily rescoring cycle. Judge live edge on /proof.

Can Q-Score be gamed or manipulated?

Q-Score is computed from public data (prices, filings, headlines, fundamentals) and walk-forward validated, so it is resistant to manipulation. However, earnings manipulation can fool the Earnings Quality dimension, and coordinated social media campaigns can spike sentiment. StoQuant uses multiple independent dimensions so that manipulation of one (e.g., fake social buzz) is outweighed by real signals (e.g., insider buying, technical breakouts).

Why does my favorite stock have a low Q-Score?

A low Q-Score does not mean the stock is bad—it means the currently active (non-zero-weight) dimensions are not aligned in its favor. Fail-closed dimensions still appear on research cards but do not move the composite. Read /proof for whether the live ranking has edge.

Does StoQuant recommend selling when Q-Score drops below 40?

No. Q-Score is a rating, not a trade signal. A Drop from 75 to 50 is meaningful but does not mandate a sell. Q-Score is best used as part of a broader process: review the driver scorecard, check if your original thesis has changed, and decide whether to hold or rebalance.