We are a team of engineers, quants, and investors building the research platform we wished existed — open methodology, calibrated machine learning, and an out-of-sample track record you can inspect.
Most stock research tools fall into two camps: pretty dashboards that tell you a stock is undervalued, or black-box AI that tells you to buy without showing the work. Neither camp gives investors the two things that actually build conviction — a transparent model and a published track record. StoQuant was built to close that gap. We combine Benjamin Graham value principles with modern machine learning, conformal prediction, and Bayesian portfolio construction. The result is a daily Q-Score for 14,000+ US-listed stocks, a hidden-gem screener grounded in intrinsic value, and a Black-Litterman optimizer that knows whether the current regime favors value, momentum, or quality. Our north star is simple: every signal should be explainable, every model should be validated out-of-sample, and every investor should be able to see the evidence before making a decision.
Transparency is not a marketing word at StoQuant; it is an architectural choice. Our data is stored append-only, so historical signal values are never overwritten. Our model changelog is public, so you can see when a model changed and why. Our proof page shows the actual forward-return distribution of Q-Score quintiles, not a cherry-picked backtest. We also publish /llms.txt and /llms-full.txt so AI crawlers and answer engines can ingest our methodology in plain text. If a model underperforms, you will see it on /proof before we do.
StoQuant is built on peer-reviewed and widely cited work: Benjamin Graham and David Dodd's Security Analysis (1934) for intrinsic-value discipline; Fischer Black and Robert Litterman's Global Portfolio Optimization (1991) for Bayesian allocation; Vladimir Vovk's Venn-Abers predictors (2015) for calibrated probabilities; Joseph Piotroski's F-score (2000) for earnings-quality measurement; and Messod Beneish's M-score (1999) for forensic accounting detection. We link to the original papers and explain how each is adapted in our pipeline.
StoQuant is an independent, self-funded team. We ship daily, read every piece of feedback, and publish our roadmap transparently. You can reach us through the site chat, on X at @stoquant, or via the MCP community channel. Related pages: Methodology (stoquant.com/methodology), Proof (stoquant.com/proof), Model Changelog (stoquant.com/model), Pricing (stoquant.com/pricing), and the MCP server (stoquant.com/mcp).
StoQuant is built by a small team of engineers, quants, and investors with backgrounds in systematic trading, machine learning, and platform engineering. Team bios and roles are listed above.
Yes. StoQuant is self-funded and independent. We do not receive payment for stock coverage, ratings, or placement in the hidden-gem screener.
We offer a free tier with Q-Score, hidden gems, and public research pages. Pro and Power subscriptions add advanced analytics, unlimited screens, API access, and the MCP server.
Visit /proof for walk-forward out-of-sample metrics and /model for the daily model changelog. Both are public.
Yes. The stoquant-mcp package on npm exposes roughly 30 read-only tools to Claude Desktop, Claude Code, and any MCP-compatible client. See /mcp for install instructions.