How qualiscore scores a stock

Every company qualiscore covers gets two scores from 1 to 10: bullishness and certainty. They are produced by a language model that runs live web searches and reads company filings, investor relations pages, analyst estimates and news. Every claim in an analysis names its source, and every score carries the date it was produced.

The three figures

Bullishness measures how positively the evidence points for a stock relative to what the market already assumes about it. It is a claim about the share at its current price, not about the business on its own: the same company can score differently at €40 and at €80. A large recent price rise means good news is already priced in, which raises the bar for a high score. Bullishness is not a rating, a recommendation or a price target.

Certainty measures how much is known, whichever way it points. It is computed only from how complete the evidence is, how well different kinds of evidence agree, and how much remains unresolved. The direction of the evidence never moves it, which is why the two scores are never combined.

Moat rating describes how durable a company’s competitive advantage looks: none, weak, medium or strong. If the model’s answer cannot be read, the rating is left out rather than shown as “none”, because those are different claims.

How an analysis is produced

1. Research. A language model runs live web searches and writes a research brief. Its sources are taken from the search’s own record of the pages it retrieved, so the model cannot cite a page it did not read.
2. Key figures are calculated, not copied. Ratios such as the forward price-to-earnings ratio and debt to equity are computed from their inputs, with checks, because the published versions of them are inconsistent.
3. Scoring. A second step reads only the research brief and writes the analysis: both scores, the moat rating, an executive summary, sections on competition, financials and valuation, recent news, catalysts, institutional investors and fund letters, bull and bear points, and the sources it cites.
4. Checks. Before it is saved, the analysis is checked. One that fails is run again or discarded, never corrected by hand.
5. Stored, never overwritten. Each analysis is kept with its date and the version of the scoring rules that produced it, so a company’s score history shows real changes rather than changes to the rules.

What is checked before an analysis is published

• Grounding. Research without the search’s own record of its sources is discarded.
• Citations. Every numbered citation in the text must point to a source the research actually retrieved.
• Catalysts. Upcoming events must appear in the research; the scoring step may not add events of its own.
• Official figures take precedence. A figure from a company’s own filings or reports replaces one found by search.
• No figure rather than a misleading one. When an input is missing or a computed figure fails a check, it is shown as not found instead of estimated, and calculated figures carry a note on how they were worked out.

How often scores change

The 50 largest companies in the S&P 500 and the members of the ATX are re-scored automatically every Monday, and their newest analyses are previewed publicly. Any other covered company is scored when a subscriber requests it, at most once a day. Every score shows the date of its analysis; a score quoted without its date is missing the part that makes it meaningful.

What the scores are not

• Not investment advice. qualiscore is informational tooling. It holds no licence to give advice and does not know your circumstances.
• Not a forecast or a price target. Bullishness compares the evidence with what the price already assumes. It does not predict where the price will go.
• Not reviewed by a person. Analyses are generated automatically and published without human review. Language models make mistakes, and sources can be wrong or out of date. Read the cited sources before relying on a claim.

Questions