The Opportunity Finder surfaces evidence, not picks. It scores every S&P 500 company on two lenses — business quality and market doubt — ranks the companies that score high on both, then reads the news and insiders' own trades as two further lenses on the top companies. The same inputs always produce the same score; there is no black box.
No factor screen reliably beats the cap-weighted S&P 500 on a risk-adjusted basis. This one does not either. Two pre-registered out-of-sample tests found no selection edge (2026-07-23). A third measurement, of the score's own components, explains why (2026-09-19). We publish the tests alongside the tool. So this is a research instrument, not an alpha engine. It measures business quality and market doubt for every company in the universe, point-in-time, and lays the evidence out for your judgment.
We score each company against its sector peers on durable-quality measures: return on invested capital, gross and operating margins and their direction, gross profitability, three-year revenue and EPS growth, free cash flow, debt load, and share-count change. The blend becomes a quality percentile against the universe.
We then measure how much the market doubts the name, two ways. Cheapness versus its own history: today's EV/EBITDA and EV/FCF against the company's trailing five-year range. And price distress: drawdown from the one-year high, position versus the 200-day average, depth within the year's range, and any early upturn in margins or returns. These combine into one cumulative doubt score; the underlying signals stay visible as separate tags. The three distress measures are not independent. Measured across 90 months they correlate between 0.74 and 0.89 — three readings of one fact, that the price fell (component audit, 2026-09-19). Read the combined score as one price-decline measure with a valuation tilt, not as four signals agreeing. We report them separately because each one is checkable against the tape.
Quality and doubt are pure numbers; they cannot tell whether today's doubt is earned. So for the top-ranked names we read the recent news and classify it. A language model judges whether the doubt reflects realized deterioration (earned — we penalize the score; a cheap price on a deteriorating business is a value trap), transient sentiment or price-action noise (unearned — we keep or lift it), a constructive catalyst (lift), or a real but unresolved debate (left about unchanged). The base quality and doubt scores never change — the overlay adjusts only a separate opportunity score, and the verdict is shown on each name's page. The weekly and monthly editions recompute it; the daily report carries the latest read forward. The read is repeatable: the model classifies, but each verdict moves the score by a fixed, pre-set amount.
Quality and doubt describe how the market sees a company; the people who run it can disagree with the market in public. For the top-ranked names we read their own Form 4 filings — the open-market shares corporate officers and directors buy or sell in their company. Buying is the signal: an insider purchase is costly and voluntary, and a cluster of insiders buying into a drawdown is the people closest to the business betting against the doubt. Selling we treat as noise — insiders sell for taxes, diversification, and scheduled plans — so it barely moves the score and never symmetrically. Unlike the news read, this lens uses no model at all: it is arithmetic on disclosed trades, so the same filings always produce the same factor. As with the news overlay, the base scores never change, the factor sits in its own column, and Form 4s are filed within about two business days — so the read is current, not a delayed disclosure.
High quality and high doubt together is the zone this report flags: a strong business the market is currently marking down. The flag is a description, not a prediction. We tested the claim it rests on — that doubt predicts returns better among high-quality businesses than weak ones. Across 90 months it does not. The difference runs slightly the other way and sits within noise (component audit, 2026-09-19). The flag tells you where to look. It does not tell you that looking there pays. A flag means look closer. It never means buy.
The top-ten table ranks the opportunities. The quality-versus-doubt map plots the full universe. A detail page for each top name shows its fingerprint, the doubt evidence, the news read, a plain-English summary, and a suggested research next step — confirm the cause, await a confirmed inflection, or watch for a better entry.
Every input is point-in-time: filings count only as of their filing date, prices only through the as-of date. The universe is survivorship-safe — it includes companies later removed from the index. Prices are dividend-adjusted. Nothing here is investment advice, and the tool never places an order. Confirm each setup in primary sources before acting.