Methodology

Every number on this site comes from a specific calculation over public data. Here is which one, on what data, and - the part almost nobody tells you - where it stops working. If something cannot be checked, it should not be believed.

Where the data comes from

Published accounts: XBRL filings from the SEC and from the European, Japanese, Korean, British and Brazilian regulators. These are the documents companies are legally required to file.
Prices and market information: public stock-market data providers.
Recommendations and price targets: those published by research firms.

Nothing shown here is our opinion about a company: they are calculations over that data.

The hit rate of each source

This is the site's central measure. It is calculated like this: each past claim by that source about that stock - a price target, a technical signal that fired - is checked against what the price did AFTERWARDS.

Only data available on the date of the claim is used. We never look ahead to decide what to count.
For technical signals the base rate is always shown: how often you would be right without them on that same stock. For the four main sources that base is not measured yet, so the 50% shown is a visual reference and not chance.
The number of cases is shown. A percentage over 30 cases says far less than one over 500.

What this is NOT: a past hit rate is not a future probability. It measures what happened in the available history, in the market that existed then.

The fundamental grade (A-F)

Eight pillars are scored - valuation, profitability, growth, health, cash, shareholder returns, insiders and robustness - over 34 metrics computed from filed accounts. Each metric becomes a PERCENTILE within the peer group (same sector and country), because a 12% margin does not mean the same in banking as in software.

Only one pillar is measured against what happened next. The first seven are accounting judgement: they group ratios that are considered signs of quality. The eighth, robustness, was built the other way round - twelve candidates were tested and the three that predicted were kept: size of the business, margin stability, and how far operating profit covers the debt.

The method: 418 sector-by-quarter comparisons between 2016 and 2025, ranking companies within their sector using only the accounts that had been filed on THAT day, and measuring the return over the following twelve months. The three give a median rank correlation of +0.107 to +0.130 and get the direction right in 72-77% of the comparisons. For reference, operating margin - the best signal among the seven classic pillars - gives +0.132.

Where it does NOT hold: it was measured on companies that file with the US SEC, so outside the United States it is unverified. And the 418 comparisons overlap - twelve-month returns taken every quarter - so they support less than their count suggests.

The letter is assigned by percentile within the universe: A the top 10%, B the next 20%, C the middle 40%, D the next 20%, F the bottom 10%. These are not absolute thresholds - with those, an A was mathematically unreachable.

A pillar can rest on very few metrics. If only two of the nine profitability metrics have data, the pillar is still computed from those two. That is why each pillar on the page shows '5/9': how many of its metrics had data. Two companies with the same grade are not comparable if one rests on 30 metrics and the other on 8.

Measured across the catalogue: the median is 27 metrics out of 34. The weakest pillar is shareholder returns, with one metric or none on 49% of symbols - which is normal, since many companies pay no dividend. Robustness comes next, with 2 of its 3 on half the cases: margin stability needs six quarters of filed accounts and only 36% of the catalogue has them.

What the grade measures, and what it does not: it measures BUSINESS QUALITY. It does NOT measure whether the stock is cheap to buy today. An excellent company can be a bad buy if it already trades expensively.

What this site cannot do

It does not predict prices. The company page shows a “could rise X% over 12 months”: that is an ESTIMATE OF TRAJECTORY —what the measured sources add up to, each weighted by how often it is right— and not a reliable forecast of what the stock will do. It is published next to its confidence and its sources precisely so that it is not read as a promise.
It does not replace anyone's judgement, and it is not financial advice.
The data may contain errors. If you spot one, report it in Contact: it gets fixed.
Everything is measured on the available past. Markets change, and the relationships measured may stop holding.