How is WPM calculated on a typing test?

The standard definition treats one word as five characters. Net WPM equals the number of correctly typed characters divided by five, divided by the elapsed time in minutes. TypeCrt implements this as ((correct word characters + correct spaces) * (60 / elapsed seconds)) / 5, where a character only counts if the entire word containing it was typed correctly. This is the same word-level counting Monkeytype uses, so scores are comparable between the two.

What is the difference between net WPM and raw WPM?

Raw WPM counts every character you typed, correct or not. Net WPM counts only characters inside words you completed correctly. Raw WPM shows the speed your fingers moved; net WPM shows the speed you actually produced usable text. The gap between them is the cost of your errors, so a large gap means accuracy is limiting you rather than finger speed.

How is typing accuracy calculated?

TypeCrt counts accuracy at the keypress level: correct keypresses divided by total keypresses. Every keystroke is recorded when it happens, and pressing backspace does not remove a recorded error. This is deliberate, because erasing a mistake does not mean it was never made. Accuracy measured this way is stable and cannot be inflated by correcting errors after the fact.

What does consistency mean on a typing test?

Consistency measures how steady your typing speed was during the test, not how fast it was. TypeCrt samples raw speed once per second, computes the coefficient of variation (standard deviation divided by the mean), and maps it to a 0-100 score using the kogasa function: 100 * (1 - tanh(x + x^3/3 + x^5/5)). A perfectly even pace approaches 100. Bursts followed by pauses lower the score even if the average speed is high.

Are TypeCrt scores comparable to Monkeytype scores?

Yes for WPM and consistency. TypeCrt uses word-level character counting for net WPM and the same kogasa consistency function that Monkeytype uses, so results on equivalent test settings are directly comparable. Accuracy may read slightly lower on TypeCrt because errors are recorded per keypress and are not erased by backspacing.

How TypeCrt Calculates WPM and Accuracy

Every formula, published in full.

Most typing tests report a number without saying how they got it. Two sites can give you a 20 WPM difference on identical typing purely from counting rules. This page documents exactly what TypeCrt computes, so any result can be checked by hand.

Net WPM

The standard definition of a "word" in typing measurement is five characters, including spaces. This convention exists so that scores are comparable across languages and text of different word lengths.

It is also the convention used in academic text-entry research — including the study of 168,960 people that produced the population figures on our research page. That matters practically: a WPM score from this site can be placed directly against those published percentiles, because both count a word the same way.

netWPM = ((correctWordChars + correctSpaces) × (60 / elapsedSeconds)) / 5

The important detail is word-level counting: a character only contributes if the entire word containing it was typed correctly. Type hello as helol and you score zero for it — not four out of five.

Why word-level? Character-level counting rewards partially-correct words, which lets a fast, sloppy typist outscore an accurate one. Word-level counting is what Monkeytype uses, so TypeCrt scores are directly comparable to Monkeytype scores on the same settings.

Raw WPM

rawWPM = ((allCorrectChars + spaces + incorrectChars + extraChars) × (60 / elapsedSeconds)) / 5

Raw counts every character you typed, right or wrong. It measures pure finger speed with no penalty.

The gap between raw and net is the price of your errors. If raw is 90 and net is 62, mistakes are costing you 28 WPM, and accuracy work will gain you far more than speed drills. If the two are close, you can safely push for speed.

Accuracy

accuracy = correctKeypresses / (correctKeypresses + incorrectKeypresses) × 100

Accuracy is counted at the keypress level, and every keystroke is recorded at the moment it happens.

Backspace does not erase an error. This is deliberate. Correcting a mistake does not mean you did not make it, and an accuracy figure that can be repaired by backspacing measures your editing, not your typing. It also means TypeCrt accuracy may read slightly lower than sites that let corrections wipe the record.

Consistency

Consistency measures how steady your pace was, independent of how fast it was. TypeCrt samples raw speed once per second, then computes the coefficient of variation:

cov = standardDeviation(samples) / mean(samples)

That value is mapped onto a 0–100 scale using the kogasa function, the same one Monkeytype uses:

kogasa(x) = 100 × (1 − tanh(x + x³/3 + x⁵/5))

A perfectly even pace gives cov ≈ 0, which returns ≈ 100. Typing in bursts separated by pauses produces a high cov and a low score, even when the average speed is high.

Consistency	What it means
90–100	Metronomic. Rare, and usually a sign of well-consolidated muscle memory
75–89	Smooth with minor hesitations — typical of a strong typist
60–74	Noticeable stalling on specific words or keys
Below 60	Burst-and-pause typing. Usually points at unresolved weak keys

Character Breakdown

Results show four counts in correct / incorrect / extra / missed order:

Correct — characters matching the target text
Incorrect — wrong character typed in a valid position
Extra — characters typed beyond a word's length
Missed — characters in the target word you never typed before moving on

These are derived by comparing your recorded input against the target words directly, not by reading the page — so the numbers stay correct regardless of what the display is doing.

Verifying This Yourself

All of the above is checkable without trusting us:

Check the WPM — finish a test, take the correct character count from the breakdown, divide by 5, divide by minutes elapsed. It will match the reported net WPM
Check the raw gap — raw minus net should track your error count
Check consistency — type deliberately evenly for one test, then in hard bursts for another at the same average speed. The consistency figures should diverge sharply while WPM stays similar
Check comparability — run the same settings on Monkeytype. Net WPM and consistency should land close; accuracy may read a little lower here for the backspace reason above