How TypeBurst Measures Your Typing

What counts as a word

A word is five characters, including the space that follows it. It is not an English word. That convention comes from typewriter-era testing, and it exists so that two people typing two different passages get comparable scores: without a fixed unit, a passage of short words would inflate a score and a passage of long ones would deflate it.

So 300 correct characters in one minute is 60 WPM, whatever those characters happened to spell. Characters per minute (CPM) is the same count without the division by five, which is why CPM on this site is always exactly WPM multiplied by five.

Which WPM variant TypeBurst reports

The headline figure is correct characters divided by five, divided by minutes elapsed. Characters you typed that do not match the passage contribute nothing to it. They are not subtracted twice either, which is what separates this from classic net WPM.

Gross WPM would count every character you typed, right or wrong, divided by five, per minute. Net WPM starts from gross and subtracts uncorrected errors divided by the minutes elapsed, and it is the figure most typing certificates and employment assessments quote. Both are calculated from the same run and shown on the results screen, so you never have to guess which number an employer means.

Worked through on a one minute run of 350 characters with 15 uncorrected errors: gross is 70 WPM, net is 55 WPM, and the TypeBurst figure is 67 WPM. Same typing, three defensible numbers.

How accuracy and errors are counted

Accuracy is correct characters divided by all characters you typed, as a percentage. The error count is the number of characters left standing that do not match the passage, and it uses the same denominator, so the two numbers always agree with each other.

Two details change the result and are worth stating plainly. First, the space that finalises a word counts as one correct character, but only if the word itself was correct; a wrong word costs its space as well. Second, only the final state of a word counts: a character you typed, deleted and retyped correctly inside the word you are on is not held against you, while a mistake you leave standing is. In other words, TypeBurst counts uncorrected errors, not keystrokes you regretted.

When the clock starts, and what it is

The timer starts on your first keystroke, not when the page loads and not when you click into the box. It stops when the chosen duration has elapsed, and the final score divides by that full duration even if you stopped typing early, which is why abandoning a test halfway does not flatter the number.

Time is read from performance.now(), a monotonic clock, rather than from the system time. A monotonic clock cannot jump backwards or forwards when the machine's clock is corrected, so an elapsed time on this site is always a real elapsed time. One honest caveat: browsers throttle timers in background tabs, so a test left running in a hidden tab finishes later in wall-clock terms than it should. The elapsed time it scores against is still measured exactly, so the WPM stays correct.

Input delay is a real, if small, source of error on any browser-based test. Keystrokes are processed as they arrive and the metrics are derived from character counts rather than from per-key timings, which keeps the measurement stable, but a heavily loaded machine will always read a little slow. That applies to every typing test in a browser, including this one.

What backspace and paste do

Backspace works normally inside the word you are typing. It is blocked at the very start of a word, so once a word has been finalised with a space it is locked and cannot be revisited. That keeps the score comparable between runs and stops a slow, meticulous editing session from reading as fast typing.

Pasting into the test is blocked outright, and a visible notice says so when you try. A test that accepts a paste measures your clipboard, not your hands.

How the passage is built

Each test draws its words at random from the word list belonging to that page, with the one rule that the same word never appears twice in a row. The standard list holds 240 of the most common English words, and the kids test uses a separate list of 100 short, familiar words. Which list a page uses is stated in plain sight underneath the test.

The passage rendered into the page's raw HTML before you start is a fixed sample, generated from a seed derived from the page's own address, so that a search engine or an AI crawler that runs no JavaScript still sees real text rather than an empty box. The moment the page loads, that sample is replaced by a freshly randomised passage.

Methodology version 1.0.0

These rules are versioned. The scoring described on this page is version 1.0.0, in force since 5 August 2026. Any change to how a score is calculated raises the version number and is listed with its date on the changelog, so a score you recorded in the past can always be matched to the rules it was measured under.

Version 1.0.0 is the first published version of these rules. It documents behaviour that was already in the code rather than changing it, with two exceptions made at the same time: pasting is now blocked, and the timer now reads from a monotonic clock instead of the system clock.

What is stored, and where

Two things are stored, both in your own browser. Your personal best per duration, under keys of the form typeburst:pb:<page>:<seconds>, each holding a single number. And a history of finished runs under typeburst:history:v1, holding the date, WPM, net WPM, accuracy, CPM, duration, word list and page for the 200 most recent tests, which is what the progress page reads. Nothing at all is sent to a server: no analytics call carries your score, there is no account and there is no upload. The text you typed and the timing of individual keystrokes are not stored anywhere at all. Clearing your browser data clears the lot, and there is no copy anywhere else, which is why the progress page offers an export.

That also means TypeBurst has no distribution of scores to compare you against. Where other sites show a percentile, this one shows the figures from published research instead, because a percentile without collected data would be invented.

The same run under three scoring rules

One minute, 350 characters typed, 15 of them wrong and left standing. Arithmetic, not a claim.

Sources

• Words per minute, on the five-character word convention and net WPM
• performance.now(), MDN Web Docs, on the monotonic clock used for timing