Where does your typing speed place you among everyone else? The percentile answers that, but only if you understand what it is comparing and what it quietly ignores.
What a percentile actually says
A percentile answers one narrow question: out of a large population of typists, roughly what share types more slowly than you do? Sitting at the ninetieth percentile means about nine in ten people in that population are slower. It is a position in a distribution, nothing more.
That is worth stating plainly because percentiles are easily misread as grades. The fiftieth percentile sounds like a bare pass and is nothing of the sort: it describes a perfectly competent typist doing exactly as well as most people. The distribution is also lopsided at the top. Moving from forty to fifty words per minute lifts you roughly twenty percentile points, while moving from a hundred to a hundred and ten lifts you less than one, because there are very few people up there to overtake.
The population being compared matters as much as the number. People taking online typing tests are self-selected: they are more likely to be interested in typing and more likely to be practising than a random office worker. That shifts the whole distribution upward, which means an online percentile is generally harsher than the same comparison against the general public.
Why comparison must be within a cohort
Speed is not one measurement, and comparing across different circumstances produces numbers that look precise and mean nothing. Test duration is the clearest case: a fifteen-second burst score is systematically higher than a five-minute sustained score for the same typist, so ranking them together ranks the durations rather than the people.
Language is the other major divide. Word length, accent characters and input methods differ enough that raw character counts are not equivalent. A Korean or Japanese typist composing characters through an input method editor produces output by a fundamentally different route than someone typing English directly, and a single leaderboard combining them would be measuring the writing system.
For the same reason, mobile and desktop scores belong apart. The device sets a hard ceiling that no amount of skill crosses, so mixing them ranks hardware. Any ranking worth reading holds language, duration and device constant.
- Same test duration: burst and sustained speed are different measurements
- Same language: character counts and input methods are not equivalent
- Same device class: a touchscreen ceiling is lower than a keyboard ceiling
- Same mode: prose, numbers and code produce very different speeds
How the percentile on this site is calculated
Your percentile is computed in your browser from a reference distribution of adult typing speeds, interpolated between the points in that table. It is deterministic, which means the same net WPM always yields the same percentile, and it costs no server round trip.
The distribution is an approximation drawn from published typing-speed research rather than a measurement of this site's own users, and it is stated that way deliberately rather than being dressed up as live data. It lives in a single file in the codebase so it can be replaced with real figures once there are enough of them to be meaningful.
The practical consequence is that the percentile is a good rough indicator and a poor precision instrument. Treat the difference between the seventy-second and seventy-fifth percentile as noise. Treat the difference between the fiftieth and the ninetieth as real.
The leaderboard and what it currently shows
The leaderboard on this page displays sample entries. They are there to show how the board behaves — how ranks, countries and languages are presented, and how filtering works — not to report anyone's real result. This is stated on the board itself, because a table of invented names presented as genuine records would be dishonest whatever the intention.
Live rankings require user accounts, since a leaderboard without identity cannot resist someone submitting an impossible score, and accounts require a backend this site deliberately does not yet have. The ranking layer is built behind an interface so that swapping the sample data for a live source needs no changes to any page.
Ranking yourself against yourself
The most useful comparison available today is against your own history, and it happens to be the one that actually drives improvement. Your personal best per mode and duration, your streak, and the trend across your last several tests are all recorded in your browser and shown on the practice page.
A global percentile tells you where you stand once. Your own trend tells you whether what you are doing is working, which is the question that matters if you intend to get faster rather than simply to find out how fast you are.