WPM is a more slippery unit than it looks. Two sites can watch you type the same paragraph at the same speed and report numbers fifteen apart, and both can be arithmetically correct.
Words vary in length, so counting them directly would make a test of short words easier than a test of long ones and neither result would be comparable. The convention, standard across typing measurement for decades, is that one word equals five characters including the space that follows it.
So the arithmetic is simply: take the characters you typed, divide by five, divide by the minutes elapsed. Type 300 characters in a minute and you typed 60 words per minute, whether those were sixty three-letter words or twenty long ones.
This is why 'the quick brown fox' style texts and real prose can produce different-looking scores for the same physical typing rate — the character mix changed, not your hands.
Raw WPM counts every character you entered, correct or not. Net WPM counts only the characters that were right on the first attempt. Most sites show one number somewhere between the two, and remarkably few tell you which one you are looking at.
Typepace shows both and treats the gap as the headline diagnostic. It is the cost of your errors, stated in the same unit as your speed, rather than a rounding artefact buried in an accuracy percentage.
| Reading | Meaning | What to do |
|---|---|---|
| Gap under 3 WPM | Errors are costing you almost nothing | Push the speed |
| Gap 3–8 WPM | Normal working range | Hold, and let accuracy catch up |
| Gap above 8 WPM | Speed you have not earned yet | Slow down for a fortnight |
Four decisions, each defensible on its own, and each shifting the result:
Typepace's answers, in order: subtracted; not correct; on the first keystroke; discarded above 1.5 seconds. That last one matters more than it sounds — a four-second pause is reading, not finger travel, and counting it would frame whichever key followed the pause as a weakness it never was.
First-try accuracy judges each position once, at the moment you first reach it, and never revises that judgment. Fix a wrong letter and the text on screen becomes correct; the record still remembers that the first attempt was not.
This reads as harsh and it is deliberate. The skill being built is hitting the right key. A metric that forgives repairs will happily show improvement for months while the underlying error rate does not move at all, which is exactly the trap that makes people plateau without understanding why.
Characters typed divided by five, divided by the number of minutes elapsed. Five characters is the standard definition of a word for typing measurement, which keeps tests comparable regardless of the length of the actual words.
Raw counts every character you entered. Net counts only the characters that were correct on the first attempt. The difference between them is the time cost of your errors.
Indirectly, yes. The correction itself takes time, which lowers your speed, and on Typepace the original mistake still counts against first-try accuracy even after you fix it.
Because sites make different choices about whether errors are subtracted, whether corrections count, when the clock starts, and whether long pauses are included. All four move the number, and most sites do not document which choices they made.
Characters per minute — the same measurement without dividing by five. A CPM of 300 is a WPM of 60. It is more common in non-English typing tests, where the five-character convention fits the language less well.