A typing speed test measures how fast and how accurately you type over a fixed interval. Latin scripts are scored in WPM — five characters counted as one word — while Korean is scored in keystrokes per minute and Japanese in characters per minute, because those are the units those languages actually use. Mistyped keys are counted as you make them, so deleting a mistake does not restore your accuracy.
The timer starts on the first character typed, not when you press Start, so the time spent reading the screen never enters the score.
The word list is built from short, common words of the language. Rare vocabulary would measure reading speed rather than typing speed.
No digits or punctuation appear in the text — their position varies by keyboard layout, which would make the same skill score differently per language.
WPM stands for words per minute, but the «word» is not a dictionary word — it is five characters including the space. Word lengths differ between passages, so counting real words would make the same typist score differently depending on the text.
This tool therefore divides the number of correctly typed characters by five, then by the elapsed minutes. The measure holds whether the passage is full of short words or long ones.
The clock starts on the first keystroke. If the time spent scanning the screen after pressing Start counted, the result would be part reaction test rather than a speed measurement.
On the standard Korean 2-set keyboard, typing 강 requires three keys: ㄱ, ㅏ and ㅇ. Counting characters alone would credit three keypresses as one character, so an equally fast typist scores about 40% of the Latin figure. That is exactly why Korea measures typing in keystrokes.
This tool decomposes the correctly typed text back into jamo and counts real keystrokes. Unicode defines the syllable block 가–힣 as a product of initial, vowel and final, so the jamo can be recovered arithmetically — no lookup table is needed.
Doubled consonants (ㄲ, ㅆ) are one keystroke because they are typed with shift; cluster finals (ㄳ, ㄺ) and compound vowels (ㅘ, ㅝ) are two distinct keys, and ㅙ and ㅞ are three.
In Korean and Japanese the character on screen changes with each key. As ㄱ becomes 가 and then 강, three different characters pass through the same position — but that is one character being built, not three mistakes.
So scoring pauses while composition is active and resumes on the committed character. Without this, only Korean and Japanese users would see accuracy in the thirties, with nothing on screen explaining why.
Browsers disagree about the composing flag on the final event, so composition start and end are tracked directly rather than read off the event.
Accuracy is correct characters divided by correct characters plus errors, where an error is recorded the moment a wrong key lands and is never removed by backspacing.
Comparing only the final text is the alternative, but then anyone typing slowly and fixing everything scores 100%. That measures patience, not accuracy.
Speed only means something alongside accuracy. Errors cost real time to repair in actual work, so accuracy and error count sit next to the speed figure rather than behind it.
Grades use fixed cuts: on WPM, 70 and above is S, then 55, 40 and 25. On Korean keystrokes per minute the cuts are 500, 380, 260 and 150.
Percentiles are deliberately avoided. No response distribution is collected server-side, so showing «top N%» would be inventing a statistic. The cuts sit near commonly published figures and are labelled as reference values.
Hardware is part of the measurement. Key travel and force, wireless latency, browser input handling and refresh rate all land in the number. Comparing across devices is not meaningful; comparing your own results over time on the same device is.
Accuracy first, speed second. Repairing a typo costs far more time than typing the character correctly in the first place, so speed tends to follow accuracy rather than the other way round.
Not looking down at the keyboard makes the largest single difference. While your eyes travel between keys and screen, your hands are idle.
Test in short runs, often. A 15-second run shows your peak once warmed up; a 60-second run shows stamina. A wide gap between the two means endurance, not speed, is the bottleneck.
Five characters, including the space. Real words vary in length, so counting dictionary words would make the same typist score differently on different passages. The five-character convention is the standard in text-entry research and in every mainstream typing test.
In this tool's bands 60 WPM falls in A (55 and above). Treat the bands as a reference: no response distribution is collected on the server, so there is no basis for a percentile, and fixed cuts are used instead.
Yes. An error is recorded at the moment it is typed and deleting it does not undo it. If only the final text were compared, anyone typing slowly and correcting everything would score 100% accuracy — which is not what accuracy means.
While a character is being composed — Korean ㄱ becoming 가 becoming 강, or Japanese kana — nothing is scored. Only the committed character is checked. Without that, Korean and Japanese users would see accuracy around a third of the true value.
No. The passage is generated and scored entirely in your browser and nothing you type leaves the page.
What this tool bases its numbers on, and how far those numbers go.
WPM = (correct characters ÷ 5) ÷ elapsed minutes. Keystrokes per minute = (correct text decomposed into jamo) ÷ elapsed minutes. Accuracy = correct ÷ (correct + errors), where errors are counted as typed and are not undone by deletion. Characters under IME composition are not scored until committed.150 correct characters in 30 seconds gives WPM = (150 ÷ 5) ÷ 0.5 = 60. If that same text were Korean and decomposed to 370 keystrokes, keystrokes per minute = 370 ÷ 0.5 = 740.