How Typometer measures typing

Words per minute

WPM = (correct characters ÷ 5) ÷ minutes elapsed

A word is defined as five characters, including the space that follows it. Only characters you typed correctly are counted, so an error costs you the character rather than merely failing to add to the total.

Raw WPM applies the same formula to every character you typed, right or wrong. The gap between raw and net WPM is the price your errors cost you, which is usually more informative than either number alone.

Accuracy

accuracy = correct keystrokes ÷ total keystrokes

Measured on keystrokes rather than on the finished text. Pressing backspace repairs what is on screen but does not un-press the wrong key, so corrections do not restore accuracy. This is a deliberate choice and it is the one people query most: an accuracy score you could rescue by correcting would reward heavy backspacing and conceal exactly the weakness the measurement exists to find.

Consistency

consistency = 100 × e^(−cv), where cv is the coefficient of variation
of the intervals between keystrokes

An even rhythm scores near 100; alternating bursts and stalls score low. It is reported alongside speed because a steady typist is usually a fast typist, and because consistency tends to predict your speed a month from now better than your speed today does.

How a level or game is scored

Speed is not the score. Every level, game and boss uses the same weighting, so a fast and careless run cannot beat a slower accurate one:

Accuracy

40%

Speed

35%

Consistency

15%

Completion

10%

Speed is scaled against a ceiling of 100 WPM, so beyond that point extra speed stops earning more. There is a test in the codebase asserting that 55 WPM at 99% accuracy scores higher than 70 WPM at 80%, because that rule is easy to state and easy to break by accident.

The percentages above are read from the scoring module at build time rather than typed into this page, so this description cannot fall out of step with the behaviour.

When the clock starts

On your first keystroke, never on page load. A slow connection, or a moment spent reading the instructions, does not count against you. The engine holds no clock of its own: the caller supplies each timestamp, which is what makes a session reproducible and allows the server to recompute a score rather than trust one.

What is not measured

Being explicit about the limits matters more than the formulas, because this is where typing sites tend to overclaim.

• Which finger you actually used. A browser reports the character and the physical key, never the finger. The site teaches the correct finger and blocks on the wrong key; it does not claim to detect your hand.
• Your posture, your keyboard, or how hard you press. None of it reaches the page.
• Anything you type as a guest. Without an account, results are calculated in your browser, shown to you, and discarded.
• Comparison against other users. There is no leaderboard, so no score here is relative to anyone else's.

Why the scores cannot be faked

For signed-in players, the browser reports what was typed and never what it was worth. Database rules permit a session to be inserted only with every scoring column at zero, and there is no write permission of any kind on the progress table. Every derived figure — speed, accuracy, experience — is recalculated on the server from the raw counts, which also flags submissions that are not physically plausible.