Three things are being timed at once
Each word's clock starts when the word is painted and stops when a correct submission arrives, and three separable processes sit inside that interval. Recognising that a word has appeared takes something like 200 to 300 milliseconds of pure visual reaction. Reading it and converting it into a motor plan takes longer for unfamiliar letter sequences than for common ones. Then there is the typing itself, which for a seven-letter word is seven keystrokes plus a return, and which is the only part a conventional words-per-minute test measures.
This is why the number here is not comparable to a typing test result. A copy-typing test gives you a paragraph you can read ahead in, so the recognition and planning costs overlap with the typing and largely disappear. Here each word arrives cold and nothing overlaps. Expect a considerably worse effective words-per-minute figure than your usual one, and treat the gap between the two as roughly the size of the per-word startup cost.
Why a typo costs more than it looks
A rejected submission does not stop the clock. The word stays up, you retype it, and the whole episode lands on that word's time. At a two-second baseline, one typo typically adds one to two seconds, which means a single mistake costs about as much as an entire additional word. Over a ten-word run, three typos are worth roughly a 30 percent penalty on the mean.
That asymmetry is the argument for the standard typing advice: speed above the point where accuracy starts dropping is worth nothing. If your accuracy is below about 90 percent, slowing down until it recovers will almost always lower the mean rather than raise it. If accuracy is already high, there is room to push.
Mean against median
Both are reported because they answer different questions. The mean is what your total time divided by word count actually was, and it is what a typo run wrecks. The median is the middle word and is barely moved by one disaster, so it describes your typical word better. A large gap between the two means one or two words dominated the run — usually the ones you mistyped, occasionally a word whose letter combination is awkward on your keyboard layout.
The spread between fastest and slowest is the same story in raw form. A tight spread means consistent attention. A wide one usually means you drifted, or that a long word turned up early before you had settled into the rhythm.
Comparing runs honestly
The word list is fixed at sixty entries and each run draws without replacement, so no word repeats inside a run but the mix of long and short words varies between runs. That variation is real: a run that happens to serve up sandwich, notebook and umbrella will be slower than one full of river, wagon and pocket, through no fault of yours. Twenty words averages more of that away than ten does, which is the only reason to prefer the longer setting.
Beyond that, the hardware matters as much as it does on any browser timing test. Keyboard scan rate, display refresh and browser scheduling all sit between your finger and the recorded time. Comparing your morning run to your evening run on the same machine says something. Comparing your laptop to someone else's mechanical keyboard says almost nothing about either of you.
Questions people ask
Why is my result so much worse than my typing test speed?
Because a typing test lets you read ahead. In a paragraph, your eyes are several words in front of your fingers and the recognition cost is hidden inside the typing. Here every word arrives alone and cold, so you pay the reading and planning cost separately on each one. The letters-per-minute equivalent shown at the bottom of the results makes the comparison explicit, and it will be well below your usual figure for exactly this reason.
Does capitalisation matter?
No. Submissions are lowercased and trimmed before comparison, so caps lock or a stray leading space will not reject a word. Only the letters have to match.
Can I use this on a phone?
It works, but touch keyboards add their own latency and autocorrect will fight you, so the numbers are not comparable to a physical keyboard. If you do run it on a phone, turn autocorrect off first or a correct word may be rewritten into a wrong one after you submit.
Is the run saved?
No. Times exist in the page while it is open and are gone on reload. There is no history, no account and nothing sent to a server. The share button copies the summary if you want to keep it.