Accuracy matters more than speed, and it is not close. A mistake costs at least three keystrokes to repair - noticing, backspacing, retyping - against the one keystroke it took to make. That asymmetry means a typist at 60 WPM and 98% accuracy finishes real work ahead of one at 75 WPM and 90%, while finding it less tiring. Hold accuracy above 95% and treat speed as the thing that follows.
Consider a 500-character passage. At 90% accuracy you make 50 errors. Each one costs you the wrong keystroke, at least one backspace, and the correct keystroke - three keystrokes minimum, often more when the error is several characters back by the time you notice it.
So the 90% typist spends roughly 650 keystrokes producing 500 characters. The 98% typist spends about 530. Even if the careful typist's fingers move 20% slower, they press far fewer keys - and they finish first.
The correction cost is also understated here, because it ignores attention. Spotting an error interrupts the flow you were typing in; resuming costs more than the keystrokes suggest.
95% is the practical floor and 97–98% is where sustained work gets comfortable. Below 95%, corrections dominate enough that raising accuracy is the fastest available route to raising net speed. Above about 99%, you are probably typing more cautiously than you need to and leaving speed unclaimed.
Every test gives you a diagnostic most people ignore: the distance between raw WPM and net WPM. Raw counts every character you typed; net subtracts the cost of the ones you got wrong. The gap is a direct measure of how much of your speed you are actually keeping.
The case for accuracy is not a case for typing slowly forever. Once you hold 97% comfortably, further accuracy gains have almost no value - the corrections you are removing are already rare - and the constraint moves to pace.
It also matters where the work is thrown away. Chat, search queries, and rough drafts tolerate errors cheaply. Code, commands, and anything a machine parses do not, and there the accuracy bar is higher still.
The practical rule: raise accuracy until it stops being the bottleneck, then raise pace under conditions that do not let you slow down, then check accuracy has not slipped.
Yes. A mistake costs at least three keystrokes to fix against the one that made it, so corrections consume a disproportionate share of your effort. Typists with high accuracy produce more finished text than faster, sloppier typists at the same keystroke rate.
95% is the practical floor and 97–98% is where sustained work becomes comfortable. Consistently above 99% usually means you are being more careful than necessary and could safely push your pace.
During practice, yes - correcting reinforces the right motion. During a timed benchmark it depends on the scoring: if uncorrected errors are penalised, fixing them is usually worth the keystrokes.
Because it should, at first. You are trading a familiar fast-and-sloppy motor pattern for a slower, more deliberate one. The speed returns within a week or two, and it returns on top of the accuracy rather than instead of it.
Raw WPM counts every character typed, correct or not. Net WPM subtracts a penalty for uncorrected errors. The gap between them measures how much of your raw speed survives your mistakes.