How statistics are counted
The statistics page shows open figures for this site: what is read, where readers come from, what devices they use and how many robots came. This article explains where those figures come from and what each of them means. What happens to your data along the way is described in detail on the Privacy page.
What we count
We count statistics ourselves, on our own server: there are no third-party counters here. We write nothing into your browser for counting — no cookies, no entries in its storage.
A visitor is a notional mark that lasts one day. All statistics are anonymous: we keep neither IP addresses nor browser details. When there is too little data to show it anonymously, we fold it into a common group.
Statistics count days in Coordinated Universal Time (UTC).
If you do not want to be counted, turn on your browser’s request not to be tracked (Do Not Track or Global Privacy Control; where to find them is explained on the Privacy page). We then keep neither the mark nor what the visitor read, and count only how many pages readers with this setting opened and roughly how many of them there were. On the statistics page this is the “Asked not to be counted” sector; its figures are rounded.
“Visitors”, “Page loads” and “Devices”
The pie charts show visitors who presented themselves as users. Not all of them are. Whom we tell apart, and how:
- Asked not to be counted — a rounded number of browsers that asked not to be tracked.
- Scripts — programs that download pages, present themselves as a browser and send browser telemetry, yet a number of signs tell us this is unlikely to be a live person: how it names itself, what it reports about itself and how fast it opens pages. Programs that do not pose as a browser do not land here — they appear in the robots section.
- People, Desktop and Phone — live users.
- People? and Desktop? — an estimate. These visitors are no different from ordinary people, but their behaviour shows an anomaly. From it we estimate how many of them are probably not people and show that number separately. We do not flag any particular visitor.
By our estimate, the real number of users lies between people and people + people? + asked not to be counted. The “Readers” line shows this range.
Robots
Robots name themselves honestly: Googlebot, Bingbot, programs of companies that train AI, link previews in messengers. We count them separately, in the robots section at the bottom of the page, and they are not included among the readers. For robots we count requests, not visitors.
Many companies publish lists of the addresses their robots come from. For such robots we check the address: a robot is counted only if the request came from that company’s network. For other robots there is nothing to check the address against, and we recognize them by name alone.
The site does not serve pages to some robots at all, for example to some SEO analytics services and scanners. They do not appear on the statistics page.
Pretended to be a robot, a browser or a visitor
These marks on the statistics page flag those who, by all appearances, pass themselves off as someone else:
- pretended to be a robot — named itself as a well-known robot but did not come from its network. Which names are used this way is shown in the “Scripts posing as robots” list;
- pretended to be a browser — named itself as an ordinary browser, but a real browser does not behave like that;
- Posed as visitors — downloaded a page like an ordinary browser, but the view was never confirmed. Programs that fake a browser look like this, but real browsers land here too: with JavaScript turned off, or in the first three cases listed under “What is not counted”.
The remaining scripts are gathered in the robots section, in the “other scripts” sector.
How the figures add up
The Visitors and Page loads charts are the overall totals that analytics services would probably show. In the first chart, scripts are probably not users, and the ones marked with a question mark raise our doubts because of anomalous figures. The real number of people lies between people and people + people? + asked not to be counted (scripts are most likely not live users).
All other sections count people only: translations, languages, books and chapters, devices, time of day, countries. They include “people?” too: the estimate is not tied to particular visitors, so it cannot be subtracted from the tables. Apart from the charts at the top, scripts appear only in the robots section, where a chart splits them into “pretended to be a robot”, “pretended to be a browser” and “other scripts”.
Visitors over the last 30 days is the sum of visitors by day: users are simply added up day after day.
A page load and a page view are the same thing: a page that was shown on screen. The “Pages served” line counts something else: every page the server gave out to anyone, robots included. That is why it is much larger.
Why the total by country is smaller than the overall total. The visitor total includes scripts and those who asked not to be counted, while countries are counted for people only, together with “people?”. The country may also be unknown: it could not be determined from the address, or the request came from a data-centre network, where the country says nothing about the reader. Such visitors are gathered in the “unknown” row. The map has neither this row nor the “others” group, so the map total is smaller still.
Small numbers are not shown. Rows with too few visitors are gathered into an “others” row with a combined total, so that no individual reader can be recognized by a rare value.
What is not counted
Readers and page views do not include:
- a page that was closed right away, before reading began;
- views in a browser that itself reports being controlled by a program;
- views whose counting was blocked by a browser extension;
- anything that is not a page: images, styles, downloadable files, as well as error responses and redirects;
- robots the site refuses.
Pages from the first three cases are still included in “Pages served”, and some of them show up in the “Posed as visitors” line.
Back to statistics or to the documentation.