How a Typing Speed Test Turns Keystrokes Into a Number

A typing speed test counts characters rather than words. One word means five characters, spaces and punctuation included, so the WPM figure on the screen is the character count divided by 5 and then by the minutes you spent. Everything that makes two tests disagree about the same passage sits on top of that single rule.

What counts as a word in a typing test?

Five characters, whatever those characters happen to be. A research note kept by I. Scott MacKenzie at York University sets out the arithmetic plainly: entry speed in characters per second, multiplied by 60 seconds per minute and divided by 5 characters per word. A 43 character phrase entered in 20 seconds gives 2.15 characters per second, which converts to 25.8 words per minute.

The five character convention is older than the personal computer. It comes from typing competitions in the early 1900s, where a five stroke word equivalent made results from different passages comparable, a history the same note traces back to a 1980 survey of typewriters and typing methods by Hisao Yamada. Strokes include letters, spaces, punctuation and everything else you enter.

The consequence is useful. A passage of short common words and a passage of long technical terms produce the same figure for the same amount of finger work, which is exactly the point of counting strokes instead of dictionary words.

When does the clock start and stop?

The clock starts on your first keystroke, not when the page appears. Time spent reading the passage, settling your hands or finding the text box costs you nothing, and a test that started on page load would be measuring reading speed alongside typing.

That choice carries a small arithmetic consequence most sites ignore. If timing begins on the first character, the preparation before that character is missing from the measurement, so the character count should drop by one before dividing. In the 43 character example, the corrected figure is 42 divided by 20, which is 25.2 words per minute rather than 25.8, a difference of 2.3 percent. The method is described in a research note on calculating text entry speed.

Stopping depends on the mode. A fixed 15, 30 or 60 second run ends when the time is up, wherever you are in the text. A full passage run ends on the last character, so its duration belongs to you. The final figure uses the milliseconds actually recorded rather than the nominal length, which is why a 60 second run occasionally reports 61 seconds.

Why does one run produce five different numbers?

Take a single 60 second run in which 300 characters were entered, 12 of them not matching the source, spread across 9 words, in a passage averaging six characters per word including the space. Every line below describes the same fingers doing the same work.

Whole-word marking counts only the words typed exactly, so 50 words less the 9 that contain a wrong character leaves 41. The spread across that table is 19 WPM on one set of keystrokes, and none of those lines is wrong. This is the whole reason two sites report different results for the same person on the same afternoon.

What does backspace actually cost?

In our typing test backspace moves the cursor back one position, clears the mark on that character so a retype can count as correct, and adds one to a separate corrections counter. What it does not do is return the time already spent, and it does not remove the mistake from your accuracy.

Accuracy here is keystrokes minus total wrong characters, divided by keystrokes, so it describes the entire session rather than the final state of the paragraph. Repairing one character costs one backspace, one retype and the pause before both, roughly three keystrokes of time for one character of finished text. That price is the mechanism behind a plateau that looks inexplicable on paper.

What does the largest typing study say about measurement?

The CHI 2018 study Observations on Typing from 136 Million Keystrokes, run at Aalto University with Cambridge, recorded 168,000 volunteers and reported something that matters for how tests are built. Rollover typing, where the next key goes down before the previous one has come back up, turns out to be common, and it is more frequent among faster typists.

A test that insisted on one key at a time would therefore understate precisely the people it claims to measure. The same study sorted participants into eight typing styles that differ in speed, accuracy and finger use, which is a reminder that one figure describes a rhythm rather than a rank. Overlapping presses also depend on the board itself, and a keyboard that drops one of two simultaneous keys is a hardware matter, covered in ghosting and n-key rollover.

When the number does not mean what people think

• Repeating the same passage measures familiarity with that passage, because the second run benefits from knowing what is coming.
• A 15 second run measures a sprint. Short runs flatter fast starters and penalise anyone who reads ahead.
• Comparing a figure between sites without knowing the marking method compares two different quantities, as the table above shows.
• A single session is a sample of one, and ordinary variation between runs is a few words per minute in either direction.
• If one particular character keeps failing rather than one particular finger, the cause may be the board rather than the hands. The keyboard test settles that in under a minute.

Who should measure, and who can skip it

Measuring is worth the minute if a job advert states a speed, if your work is scheduled or paid by volume, or if you want a starting point before practicing. The unit employers publish is frequently not WPM at all, which is set out in what WPM means and in CPM and keystrokes per hour.

Measuring is not worth much if the figure is being chased for its own sake, or set against someone else’s result from another site under other rules. The comparison that carries information is with your own earlier run, at the same length and under the same marking method.

Start with one 60 second run on plain prose, then the same length on a passage dense with semicolons and capitals. The gap between those two figures says more about your typing than either of them alone.

FAQ

Why is a word five characters?

Because typing competitions in the early 1900s needed results from different passages to be comparable, and counting strokes rather than dictionary words achieves that. The convention passed from competitions into research and from research into every online test.

Does backspace lower my speed?

It lowers your speed and it does not repair your accuracy. The seconds spent correcting stay in the denominator, and the wrong character stays in the error count for the session even after the text on screen looks perfect.

Why does another site give me a different number?

Because it very likely counts differently. Whole-word marking, refusing wrong input, and the treatment of a trailing partial word each move the result by several words per minute on identical typing.

Which test length is the most honest?

Sixty seconds on a passage you have not seen before. Fifteen seconds measures a sprint, and a passage you have typed already measures memory as much as technique.

Is it WPM or wpm?

Research literature writes it in lower case, on the grounds that words per minute is a unit rather than a name. Job ads and test sites write WPM, and both refer to the same arithmetic.

Then take one deliberate run at something demanding. Walt Whitman’s There was a child went forth every day, from the first Leaves of Grass in 1855, is 225 characters of short repeated phrases, the plain end of the scale. Michael Faraday’s account of the shadow of a candle flame, taken down at his Christmas lectures and printed in 1861, packs 460 characters into two sentences, each one broken by a semicolon and interrupted by commas around a phrase. Washington Irving’s page on going into the country, from The Sketch Book of 1820, runs 499 characters of the same punctuation heavy kind. Typing one after another shows where your rhythm goes when a sentence stops being predictable. All three are waiting in the typing test, each with its author, year and source on its own page.