Speed or accuracy: which to train first
Why accuracy comes first
Accuracy is the skill itself: hitting the intended key without a correction. Speed is what happens when the skill becomes automatic. At 95 percent accuracy, one character in twenty is wrong, and in real writing that means a fix every few words: notice the error, stop, backspace, retype. The cleanup costs more time than the typing saved by hurrying. Clean strokes, repeated, become automatic; automatic strokes become fast. Train in that order and you get both.
The 98 percent floor
Use a simple gate: do not chase pace below 97 to 98 percent accuracy. Below the gate, faster practice is error practice, and errors automate just as reliably as correct strokes. Every typing test that shows both numbers can be used this way: run it, read accuracy first. If it is below the floor, the run does not count as speed training, whatever the WPM says.
How speed follows accuracy
Motor learning has a boring, well-reproduced rule: skills consolidate around what you repeat. Repetition at 98 percent builds clean automation, and once a stroke pattern is automatic, the floor of its speed rises on its own. This is why typists who slow down for two weeks often jump 10 WPM afterwards without any speed work. The improvement was already earned; the clean repetitions unlocked it.
When to push pace
Once accuracy sits above 98 percent in normal sessions, add deliberate overspeed: run one or two tests per week as fast as you physically can, accuracy be damned. These runs look bad on paper and they raise your ceiling, because they make the current pace feel slow. Keep them separate from clean practice and never let them become the default; a week of only overspeed runs drags accuracy down within days.
Reading the numbers
Most tests report gross WPM (raw keystrokes) and adjusted WPM (errors penalized). Judge pace runs by adjusted WPM, which is the honest number. Judge accuracy sessions by accuracy alone and ignore the speed line entirely. And compare like with like: a 60-second run and a two-minute passage do not produce comparable numbers, and mixing them makes every plateau look like a mystery.
A simple weekly split
Four days of clean practice: real passages at a pace where accuracy stays at or above 98 percent. One day of overspeed runs. One day of a fixed benchmark test, always the same conditions, logged. One rest day. That split moves both numbers steadily for months; when accuracy climbs while speed stalls, you are due for an overspeed block, and when speed climbs while accuracy falls, you are training too fast.
Tools
Type Sprint shows gross WPM, adjusted WPM and accuracy for every run, with a chart of how the run developed, so the weekly split needs no spreadsheet. Its history page keeps the trend on your own device, and the weak-key panel turns error counts into tomorrow's drill list.