How Typing Speed Is Measured: WPM, Accuracy and Tests

Typing speed is measured in words per minute (WPM), and a word is five characters, counting spaces and punctuation ¹. That fixed rule means a typing test does not count dictionary words. Most tests turn keystrokes into five-character blocks ², so speeds can be compared across languages and across tests ³. Three other numbers usually appear beside WPM. Gross WPM counts every word you typed, net WPM counts only the words that were right, and accuracy is the share of the typing that was correct. The same keystrokes can therefore give two different headline numbers, depending on the test's rules.

This guide explains each unit, works one example from start to finish, and compares how the typing tests and exams whose scoring we could read count words and errors. It then covers the measures that researchers use, which are stricter than most websites. For the typical scores people get, see the guide to average typing speed.

A typing word is five keystrokes, not a dictionary word

The standard definition comes from text entry research. For speed purposes, a word is five characters, and the count includes spaces, punctuation and other characters in the typed text ¹. The same five-character rule appears in the 2018 online study of 136 million keystrokes ^{4 5}.

To turn a rate of characters per second into words per minute, you multiply by 60 seconds per minute and divide by 5 characters per word ⁶. That is the same as multiplying by 12. MacKenzie's worked example uses a 43-character phrase entered in 20 seconds. Once the timing is adjusted, the example gives 25.2 words per minute ⁷.

The timing choice matters. The score depends on where the timer starts and stops, so two timers can score the same typing differently, even when both use five characters per word ⁷.

The table below sets out the units you will meet on typing tests and in research. The formulas are the ones used in the sources above, and the glossary links give a short definition of each.

Gross WPM counts everything you typed, and net WPM counts only what was correct

Gross speed is the rate of everything you typed. Net speed is the rate of what was right. A worked illustration from CBSE, India's central school board, shows both for its own typing test. For one candidate, gross speed is 52.1 words per minute ⁸, and net speed for the same run is 40.1 words per minute ⁹. In that example the candidate typed 521 words in 10 minutes, and only 401 of them were correct.

The CBSE formula counts only the correct words for net speed ¹⁰. The incorrect words therefore earn nothing, and they also stay in the gross figure. The two scores are shown side by side below.

The gap between the two scores is the cost of errors in that formula. Accuracy is the net score as a share of gross, which in the example comes to 77 percent ¹¹. The flow below shows how the figures connect.

Net speed has no single agreed formula. The CBSE example counts correct words only. The 2018 online study keeps speed and error rate apart: it reports words per minute from the first to the last keypress ¹², and it reports an uncorrected error rate as a second measure ¹³. Keeping the two separate tells you whether a lower score came from slower typing or from more mistakes. A headline figure can combine the two in a way the test does not state, so read the test's own definition before comparing.

Accuracy is the share of correct text, and tests count it in different ways

Accuracy is usually shown as a percentage of text that is right, but the definition varies. TypingTest.com says accuracy compares the characters or words you typed correctly with the total you typed ¹⁴. The CBSE formula divides net speed by gross speed and multiplies by 100 ¹⁵. TypingTest.com's rule allows characters or words, while the CBSE rule works in words only, so an accuracy figure from one test cannot be read as the same figure on another.

The research measure is stricter about what counts as an error. The uncorrected error rate is the Levenshtein edit distance between the text presented and the text typed, divided by the length of the longer string ¹³. Because it measures what is left in the final text, a mistake you fixed before moving on does not count.

The distinction between corrected and uncorrected errors matters most for dictation. In a phone study, speech input had a corrected error rate of 5.30 percent, against 11.22 percent for the keyboard, so speech produced fewer errors while it was being entered. But speech left more errors in the final text, at 1.30 percent against 0.79 percent ¹⁶. Speech researchers use a related measure, the word error rate, for the same reason.

Some tests make accuracy a pass mark. For court reporters, the Registered Professional Reporter test requires transcription at each speed with 95 percent accuracy ¹⁷. Here accuracy is part of the pass mark, not only speed.

Researchers have also shown that the rules themselves change the result. A 2009 study found that the way errors are handled had a significant effect on every frequently used error metric ¹⁸. An earlier CHI paper found shortcomings in the two error statistics then in use, the minimum string distance error rate and keystrokes per character ¹⁹.

Characters per minute and keystrokes per hour are the same measure in other units

Characters per minute (CPM) needs no conversion beyond the word rule. Because a word is five characters ¹, CPM is five times WPM. A test that reports CPM is reporting the same typing, so you can convert between the two without changing the result.

Keystrokes per hour (KPH) counts every key press over an hour rather than words. One word per minute is 300 keystrokes per hour, because five keystrokes times 60 minutes is 300 ²⁰. The conversion holds only when each keystroke produces one character. A data-entry test that counts shift keys, tabs and numbers in a different way will not match the simple factor.

Keystrokes per character (KSPC) answers a different question. It measures how many key presses a method needs for each character it produces. MacKenzie defines it as the average number of keystrokes needed to generate each character with a given technique ²¹. It describes how efficient an input method is, not how fast a particular person types.

Test length and text type move the score, so compare only tests that match

Test length is a choice each site makes. TypingTest.com says a one-minute test is good for a quick check, and that 3, 5 or 10 minutes can give a steadier measure of your real speed ²². 10FastFingers uses a one-minute test that reports WPM, accuracy and the counts of correct and wrong keystrokes ²³.

Length matters because speed and accuracy shift during a session. In a study of typing during real-life office work, speed and accuracy both declined later in the working day ²⁴. A long test averages over tired stretches, and a short test can catch a fast burst, so the two numbers answer slightly different questions.

We found no controlled study that gives the same typist's score across test lengths. The advice on length therefore comes from the sites themselves, and a claim that a short test scores higher, or that a long one scores lower, should be treated as untested.

Text matters too. In a crowd-sourced study, keystroke timing correlated with the letter, bigram and trigram frequencies of English text ²⁵. Keystroke timing correlates with the statistics of the language being typed, so a test of ordinary English sentences may measure something different from a test of random characters. When you compare scores, check whether both tests use the same kind of text.

Copying, typing random words and composing your own words are three different speeds

Many published speed figures come from copying tasks. In the 2018 online study, each participant transcribed 15 English sentences ²⁶. The 2019 phone study was a web-based transcription task, in which volunteers first read and memorized each sentence and then typed it ²⁷. Those volunteers averaged 36.2 words per minute with 2.3 percent uncorrected errors ²⁸. The 2018 keyboard average was 51.56 words per minute ⁵. The chart below sets the two averages side by side, though they were measured on different devices.

Composing is different. When you choose the words as you type, you also think, pause and revise. An ETS study of 463 eighth-grade students compared three writing tasks: copying, drafting and editing ²⁹. Copy typing was relatively slow and had little time spent in long pauses, apart from pauses between words, which the authors attribute to looking back at the source text ³⁰. Drafting had relatively fluent typing, but it included significant backspacing, which the authors link to false starts and revision ³¹.

So a copy test measures transcription speed. It does not measure how fast you compose, because the keystroke patterns of the two tasks differ. The same logic applies to dictation. In the 2018 phone study, speech entry reached 153 words per minute for English ³², and that speed counts the finished text, including the time spent fixing errors. For the difference between the two methods, see the guide to dictation versus typing and speaking versus typing speed.

Popular typing tests compared: what each one counts as a word and an error

The table compares the tests whose own pages describe their scoring, along with the national exam and the research measures above. We could not read usable scoring text from the pages of Monkeytype, TypeRacer, Keybr or Human Benchmark, so this guide does not state how those sites calculate their figures.

Two things stand out. Three of the four rows use the same five-character unit, or its keystroke version, and the CBSE exam counts typed words instead. The error rules are not the same. Neither TypingTest.com's FAQ nor the 10FastFingers description gives a net formula, so you cannot tell from those pages how a net score would be worked out. The CBSE document is the only one of these that spells out the net calculation in full.

That leads to the question of which test is most accurate. A test cannot be more accurate than another in general, because accuracy here means matching a definition. A test is most useful when its definition matches the decision you are making. If you need a score for a job application, use the test that the employer or exam names, and copy its rules. If you want to track your own progress, use the same test, the same length and the same kind of text each time. The question on the most accurate typing speed test goes through this in more detail.

Researchers measure text entry with words per minute, error rates and keystrokes per character

Text entry research uses a small set of measures: speed in words per minute, an error rate computed from an edit distance, and keystrokes per character for the efficiency of a method. The 2018 online study reports both speed and an uncorrected error rate ^{12 13}.

Method details can change a published speed, even when the word rule is the same. Two studies that both use five characters per word can report different speeds if they start and stop the clock at different points ⁷. How errors are handled also changes the error figures that studies report ¹⁸.

When you read any typing score, check five things:

1. The word rule. Five characters, or a count of words as the test defines them?
2. The clock. Does it start at the first keystroke, at the first character, or when the text appeared?
3. The error rule. Is an error subtracted, is only the correct text counted, or are errors reported separately?
4. The length and text. One minute, a fixed number of sentences, or a passage? Copied, or composed?
5. The comparison. Compare a score only with others that passed the first four checks.

For the specific questions that follow from a score, the answers are in the pages on how words per minute is calculated, gross and net WPM and converting WPM to keystrokes per hour. To see how the numbers are used in practice, read the guide to what counts as a good typing speed and the guide to typing faster. The full set of guides on typing speed is in the typing speed overview.