9 min
How to Test a Keyboard Properly
A key that seems dead is usually not a broken switch. Test in the right order — registration, rollover, chatter, latency, feel — and you find the real fault.
Why Typing Tests Give Different WPM Scores
Take a typing test on two websites and you will very likely get two different speeds. Neither needs to be wrong. A typing score is a formula applied to what you typed, and the formulas, the passages and the rules about mistakes all differ.
8 min read · updated 15 September 2026 · links to 6 tools
In short
Every typing test counts a “word” as five characters, including spaces and punctuation, so words per minute is really characters per minute divided by five. The differences come after that. Gross speed counts everything you typed. Net speed takes mistakes into account, and tests do that in two common ways: counting only the characters that are correct, or subtracting one word per minute for every uncorrected error each minute. The second is five times harsher per mistake, so a one-minute run of 250 characters with 10 left wrong scores 50 gross, 48 net by the first method and 40 by the second. Accuracy also has two meanings: how much of the final text is right, or how many of the keys you pressed were right, including mistakes you fixed. The passage matters too: code and punctuation-heavy text need far more Shift presses than everyday words. Compare scores only from the same test, the same length and the same material.
One run, three honest scores
Here is a single one-minute run. You type 250 characters, and when the minute ends, 10 of them are still wrong. Three perfectly respectable typing tests can report three different speeds for it.
| Score | Formula | This run |
|---|---|---|
| Gross WPM | characters typed ÷ 5 ÷ minutes | 250 ÷ 5 ÷ 1 = 50 |
| Net WPM, correct characters | correct characters ÷ 5 ÷ minutes | 240 ÷ 5 ÷ 1 = 48 |
| Net WPM, errors per minute | gross − uncorrected errors ÷ minutes | 50 − 10 ÷ 1 = 40 |
A word is five characters
Real words vary in length, so counting them would reward short words and punish long ones. Typing tests settled long ago on a fixed unit instead: five characters, counting spaces and punctuation. Words per minute is characters per minute divided by five, whatever you are typing.
For everyday English the convention is close to reality. The list of common words behind the typing test on this site averages 3.9 letters a word; add the space after each one and a typed word averages 4.9 characters, almost exactly the standard five.
For other material it is not close at all, and that is the point of the convention. The code passage on this site averages 7.2 characters between spaces. At 200 characters a minute, counting real space-separated words would give about 28 words a minute; the five-character rule gives 40, the same as everyday text at the same finger speed, which is what makes the two comparable.
Net speed: two ways to punish a mistake
The correct-characters method simply leaves wrong characters out of the count. Each uncorrected mistake removes one character, which is one fifth of a word, so ten mistakes in a minute cost 2 words per minute.
The errors-per-minute method subtracts a whole word per minute for every uncorrected error, spread over the length of the test. Ten mistakes in one minute cost 10 words per minute; the same ten mistakes spread over a five-minute test cost 2. Per mistake it is five times harsher than the correct-characters method.
Neither is more correct. The errors-per-minute method treats a mistake as the time it would take to find and fix it, which suits jobs where errors are expensive. The correct-characters method measures how much right text you produced. What matters is knowing which one a score used before comparing it with another.
The tests on this site use the correct-characters method, and publish the formula beside every result.
Accuracy has two meanings too
Final-text accuracy compares what is on the screen at the end with the passage. Backspace a mistake and retype it correctly, and it no longer counts: fix all your mistakes and final-text accuracy is 100%.
Keystroke accuracy counts every character key you pressed, including the wrong ones you later deleted. Type a 250-character passage perfectly in the end, but press 20 wrong keys on the way, and keystroke accuracy is 250 correct out of 270 pressed: 92.6%. The time spent fixing those mistakes has already come out of your speed.
A test that shows only one accuracy figure is usually showing final-text accuracy, which flatters people who type fast and correct a lot. Keystroke accuracy is the better guide to how cleanly you actually type. The tests on this site show both.
The passage changes the score
Symbols and capital letters need the Shift key, awkward reaches and more attention, so passages full of them are slower to type at the same skill. The share of characters needing Shift in this site's own passages shows how different the material is:
| Material | Characters needing Shift | Characters between spaces |
|---|---|---|
| Full sentences | 0.7% | 5.4 |
| Punctuation-heavy | 8.3% | 6.0 |
| Code | 12.4% | 7.2 |
| Numbers | 0% | 6.9 |
Numbers need no Shift but are slow for another reason: the number row is far from the home keys, and digit strings have none of the rhythm of words. A score on common words, a score on sentences and a score on code are three different measurements.
Length, pauses and warming up
Short tests usually score higher than long ones. Over 15 seconds you can sprint; over five minutes concentration and stamina start to matter, and most people lose speed. Neither is wrong, but a 15-second score and a five-minute score are not the same measurement, which is why employers who test for real work tend to use longer tests.
Tests also differ in how they handle interruptions. Some stop the clock when you switch away from the page; others keep it running, so a message that pulls you away quietly costs speed. The tests on this site pause while the tab is hidden and report how many seconds were left out.
How to compare typing scores fairly
- Use the same test each time. A different site may use a different net formula.
- Keep the length and the material the same: one minute of common words is not five minutes of sentences.
- Note which speed you are reading, gross or net, and which accuracy, final text or keystrokes.
- Take three runs and use the middle one; single runs vary a lot.
- To improve, look at which characters you miss most rather than at the headline number. The same few keys usually cause most of the errors.
Frequently asked questions
How is WPM calculated?
Characters typed, divided by five, divided by the minutes taken. A “word” is five characters including spaces and punctuation, so 250 characters in one minute is 50 words per minute.
What is the difference between gross and net WPM?
Gross WPM counts everything you typed. Net WPM accounts for mistakes, either by counting only correct characters or by subtracting one word per minute for each uncorrected error per minute. The second method gives a lower score for the same mistakes.
Why is my typing speed different on different websites?
Tests use different net formulas, different accuracy measures, different passages and different lengths. Any one of those can move the score by several words per minute, even when your typing is exactly the same.
Does fixing mistakes lower my WPM?
It costs time, which lowers speed, but on most tests a fixed mistake no longer counts as an error in the final text. Keystroke accuracy still counts it, which is why that figure is lower than final-text accuracy.
Why is my WPM lower on code?
Code needs many more Shift presses and symbols in awkward places, and has no predictable rhythm. In this site's code passage 12.4% of characters need Shift, against under 1% in ordinary sentences.
Is a 15-second typing test accurate?
It measures a short burst accurately, but it usually reads higher than a longer test because stamina and concentration have not come into play. Use a one-minute or five-minute test for a number that reflects real work.
More guides
9 min
Why Two Texts That Look the Same Don't Match
Zero-width spaces, non-breaking spaces, curly quotes, lookalike letters and accents stored two ways make identical text fail to match. How to find each.