Methodology and data sources

Prices and timing

Prices are end-of-day, not real time. The Athens session closes at 17:20 local time, and the closes our data provider has in the first minutes after it can still be from before the closing auction. So we check at 17:30, 18:00 and 18:30 Athens time, all year, and publish a day’s closes only once the exchange’s own official closing files confirm them, usually by 18:00. Where the two differ, the exchange’s figure wins.

Prices, volumes and dividend events come from Twelve Data. Gaps are filled from the exchange’s own official daily closing files. Company figures (P/E, earnings per share, margins, debt, returns, growth) are worked out by us every evening from the company’s reviewed profile, the annual reports it files in the EU’s tagged ESEF format, its results announcements and our closes, and every page and MENTOR show the same figure with the period it covers; the provider’s own figures stand in only for a company we have not worked out yet. If a stock’s last bar is older than the session shown, it is marked stale and left out of the day’s rankings.

Movers and relative volume

Daily return is the change from the previous close. Relative volume is today’s traded volume divided by the stock’s average volume over the previous ten sessions: 3× means three times a normal day. A large move on high relative volume is the signal the front page is built to surface.

The movers universe is common stocks with a market capitalisation of at least €25 million; smaller companies keep their own pages but stay out of the rankings, where a handful of trades can fake a move.

Sectors and peers

Sector performance is the median daily return of the sector’s members — not a cap-weighted average — so one large bank cannot stand in for the whole group. Sectors with fewer than two members are not shown.

A company’s peers are the listed companies with the same business model, from our own map drawn from each company’s reviewed profile, not from the data provider’s industry labels, which file a refiner beside a fuel reseller and a stock exchange beside the banks (a company not yet on the map is grouped by its provider industry for now, marked provisional). The closest come first, nearest in size. A company with no peer of its kind in Athens gets none, rather than the largest companies of its sector; related businesses are offered, marked, and count only when you tick them (an airport’s are the ports).

Each business is compared on the measure it is usually priced on, the same on every member’s page: book value for banks, yield for real-estate investment companies, P/E for retailers (rent sits outside a renter’s EBITDA), EV/EBITDA for industrial and consumer companies. Valuation multiples are shown against the median of the peers that count. A figure that would mislead stays out of that median, marked on the page: net profit above operating profit (likely one-off gains), a P/E or EV/EBITDA too high for the profit beneath it to mean much, a figure under review, a one-off or a recent share issue recorded in the company’s reviewed profile, figures missing an annual report, or a suspended share. A note applies only while the figures shown contain what it describes. Size gaps and parent and subsidiary links are marked too. You can tick any peer in or out and the median follows, and the question suggested for MENTOR follows your choice.

Dividend yield

We compute the trailing yield ourselves: every cash distribution that went ex-dividend in the twelve months to the latest close, divided by that close. We do not use the provider’s ready-made yield field, which we found unreliable for Athens-listed companies, while the underlying dividend events matched company filings.

When a yield cannot be stated honestly — no record, or a dividend paid in a different currency from the share price — we show none.

Company profiles

Each profile is drafted by one AI model from cited primary sources (annual reports, results releases, exchange announcements) and checked by a second, independent model before it is published. It states what the company is, what it is not, how it makes money and what has changed recently, with a numbered source behind each claim. Profiles are refreshed on a rolling basis, at least every six months.

The profiles are the only AI-written text on the site. Prices, ratios and rankings are arithmetic on the data, never generated.

How we check what we publish

Every night, after each data refresh, every published figure is checked against the figures it is built from. The P/E times the earnings per share must land near the share price on the day it was measured, and so must the market cap divided by the shares outstanding. A dividend yield above 30% must be explained by a one-off payment. These are arithmetic checks on our own database: no guesswork, and no AI.

When a check fails badly, the figure stays on the page, marked “under review”. It is left out of rankings, peer medians and what our assistant says until it is resolved. We would rather show the doubt than hide the number.

Once a week an independent review settles each flagged figure: an AI agent working from the companies’ own results releases, financial reports and exchange announcements. A false alarm is cleared with its source; a confirmed error is corrected with its source. Small, well-sourced corrections are applied automatically. Larger or less certain ones are approved by a person first. The same review spot-checks a rotating set of stocks every week and answers every wrong-number report from our beta readers.

Every correction is listed on the corrections page: what the page showed, what it should have shown, when, and the source.