Typing Speed (WPM) Calculator

This works on typing you have already done; it is not a live typing test. Type a passage in a practice site, word processor or from a printed page while timing yourself, then enter three numbers. Total characters typed includes spaces and punctuation, the figure most character counters report directly. Time taken is in seconds. Uncorrected errors are mistakes still present in the finished text; a typo you backspaced and fixed does not count, although fixing it cost you time.

For 250 characters, 60 seconds and 3 errors, the results panel shows Net WPM 47.0, Gross WPM 50.0, Accuracy 98.8% and Time 1.00 minutes. Net WPM is the highlighted result because it reflects text you could actually use without proofreading.

If a typing site reports gross WPM and accuracy but not net WPM, back out the error count first: errors = total characters × (1 − accuracy ÷ 100). Then enter that count here.

Time (minutes) = seconds ÷ 60.

Gross WPM = (total characters ÷ 5) ÷ minutes.

Net WPM = gross WPM − (uncorrected errors ÷ minutes), never allowed below zero.

Accuracy (%) = (total characters − uncorrected errors) ÷ total characters × 100, or 100% when no characters are entered. The error count is capped at the character count, so accuracy cannot go negative.

Notice that the two error adjustments use different units. In net WPM each uncorrected error removes one whole word, five characters' worth, per minute. In accuracy it removes one character. That is why three errors cost 6% of the gross rate in the example but only 1.2 points of accuracy.

Keep the same gross speed and the same 3 errors, but make it a 5-minute test: 1,250 characters in 300 seconds. Gross WPM is still 50, but net WPM is 49.4 and accuracy is 99.76%, displayed as 99.8%. The three errors are spread over five minutes, so they cost 0.6 WPM instead of 3.

The penalty is errors per minute, not errors per test. A 1,000-character, 5-minute session with 15 errors gives 40 gross and 37 net WPM at 98.5% accuracy: 15 errors over 5 minutes is the same 3 per minute as the first example, so the gap between gross and net is exactly 3 WPM in both. Push errors up to 15 in the original one-minute test and net WPM falls to 35, with accuracy at 94.0%.

The practical rule: compare results only across tests of the same length. A 30-second sprint of 250 characters with 3 errors scores 100 gross and 94 net WPM, which says more about the short timer than about the typist.

Counting real words would reward passages full of short words and punish technical text. A fixed five-character word makes a rate comparable across any passage, which is why typing tests and text-entry research both use it. The largest published typing dataset, Dhakal, Feit, Kristensson and Oulasvirta's "Observations on Typing from 136 Million Keystrokes" (CHI 2018), defines one word as any five characters in exactly this way.

That study measured 168,960 volunteers transcribing 15 English sentences. The average was 51.56 WPM (standard deviation 20.2), or about 258 characters a minute. The fastest 10% typed above roughly 78 WPM, the slowest 10% below roughly 26 WPM, and some participants exceeded 120 WPM. The average uncorrected error rate was 1.167%, and 90% of participants left fewer than 2.66% of characters wrong.

Read those as a reference point, not a norm. Participants were self-selected visitors to a typing-test website, mostly younger and US-based, and speed was timed from first to last keypress on short sentences. The study's speed is not error-adjusted, so compare it with gross WPM: the first example's 50 gross WPM sits just below the 51.56 average, and its 1.2% error rate (3 of 250 characters) is close to the study's 1.167% mean.

Starting the timer too early. The CHI 2018 study described above timed each sentence from the first keypress to the last. If your stopwatch also includes 10 seconds spent reading the passage, the same 250 characters and 3 errors take 70 seconds instead of 60, and the calculator reports 42.9 gross and 40.3 net WPM instead of 50 and 47. Start timing on the first key.

Counting errors by word in one test and by letter in the next. A misspelt word with two wrong letters can be logged as one error or two. Logging the first example's three misspelt words as six letter errors drops the result to 44 net WPM and 97.6% accuracy. Either convention is defensible; switching between them makes your progress chart meaningless.

Entering keystrokes instead of characters. Keystroke counters include every backspace and correction. This calculator expects the characters in the finished text, because corrected mistakes already show up as lost time.

Comparing composition with copying. Typing your own sentences includes thinking time, so it is not comparable with transcribing a passage placed in front of you. Pick one mode and stick to it.

The calculator does arithmetic on the three numbers you supply; it does not observe your typing. Its accuracy depends on an honest character count, a real timer and a careful error count, and it cannot tell a transposed pair of letters from two separate mistakes.

It does not distinguish keyboard layouts, languages, input methods such as phone keyboards, or passages with heavy punctuation and numbers, all of which change what a realistic speed is. Because "words" are five-character units, the WPM shown will not equal your real word count divided by time. Treat an employer's or exam board's own test as the authority when a specific score is required.

Paste what you typed into the word counter calculator to get the exact character count, spaces included, that this calculator's first input expects, rather than estimating it. When results swing between practice passages, run each passage through the readability score calculator: long words and dense technical sentences slow typing independent of skill, so progress is best tracked on passages of similar difficulty.