How to Type Faster: What Actually Works

Fix the errors before you chase the pace

The fastest available gain is not extra finger speed. It is keeping the speed you are already producing instead of throwing part of it away. On any test that reports net words per minute, a mistake is not a small stylistic blemish — it is a subtraction, and on Keyboardo it is a subtraction that happens twice.

Here is the arithmetic, using the formulas this site actually runs. Take a 60-second English test on which you press 400 printable keys: the letters, plus the space after each word. Raw WPM counts every one of them, right or wrong, so your raw score is 400 / 5 / 1 minute = 80.00.

Now suppose 20 of those keys were wrong letters, landing in 20 different words, and you left all 20 standing. Three things happen at once:

• The 20 wrong letters are not in your final text, so they do not count toward net WPM.
• Each of those 20 words now fails the exact-match check, so it forfeits the single space that a completed correct word contributes to the correct-character count. That is another 20 characters gone.
• The space you pressed after each broken word was itself recorded as an incorrect keystroke, so 40 of your 400 keystrokes are wrong rather than 20.

Net WPM is therefore (400 − 40) / 5 / 1 = 72.00, and accuracy is 100 × 360 / 400 = 90.00. Twenty mistyped letters cost you eight words per minute — a tenth of your score — without your hands moving any slower.

What a correction costs instead

Fixing the error is not free either, and it is worth knowing exactly which way the trade runs. A repair costs three keypresses' worth of time — the wrong key, the backspace, the right key — and backspaces are excluded from all three of Keyboardo's numbers. They never appear in your raw WPM, they never appear in your accuracy, and they never appear in your net WPM. The only place their cost shows up is the elapsed clock that divides every one of those scores.

So a correction buys back the characters and loses the seconds, while the original slip stays on your accuracy permanently: a character typed wrong and then fixed counts once as an incorrect keystroke and once among your final correct characters. Neither choice is free. Not making the mistake is the only move that is.

Learn to overlap your keystrokes

The most useful finding in the largest published typing study is not about how quickly fast typists press keys. It is about how much their keypresses overlap. In the study of 168,000 volunteers behind the University of Cambridge's report on what makes a faster typist, the fastest participants performed between 40 and 70 percent of their keystrokes using rollover — the next key going down before the previous one has come back up.

That reframes the target. You are not trying to make each individual press quicker; below a certain point a key simply cannot travel any faster. You are trying to stop waiting for one letter to finish before beginning the next, so that common pairs and short common words leave your hands as a single motion rather than a sequence of decisions.

Overlap cannot be forced by intent. It appears when your hands know the next letter before your eyes have finished with the current one, which is a consequence of two other habits: reading ahead, and repeating common sequences until they stop needing thought.

Read ahead of the caret

Most people who have never deliberately practised typing read exactly the character the caret is sitting on. That guarantees a stop-start rhythm, because every letter is a fresh instruction that has to arrive, be recognised, and be executed before the next one is even requested.

Train your eyes to sit one or two words ahead of where your fingers are. It feels wrong at first, in the specific way that reading music while playing it feels wrong, and it will briefly cost you accuracy. What it buys is a queue: your hands always have somewhere to go next, which is the precondition for the overlap above.

Looking down at the keyboard destroys this outright. Every glance at your hands is a round trip your eyes have to make, and while they are away the queue is empty. This is the honest, mechanical case for typing without looking, and it does not depend on any promised speed multiple.

Break the backspace reflex

There is a habit worth breaking that is distinct from making fewer errors: reflexively rewinding. Some typists hit backspace at the first flicker of doubt, before they have established that anything is actually wrong, and some rewind several words to repair a typo they have already passed.

Keyboardo will not let you rewind indefinitely, and the boundary is a useful discipline. You can backspace freely inside the word you are typing, and you can step back into the previous word only if that word is still wrong. Once a word matches its target exactly, it is locked in and the caret will not go back past it. That means the only corrections available to you are the ones that actually change your score, which is a good approximation of the rule to follow in real writing too.

Train rhythm, not bursts

Two typists can produce identical average speed and be in completely different shape. Keyboardo separates them with the consistency score, which is the coefficient of variation of your per-second raw WPM: it samples how many characters you produced in each completed second, then scores 100 × (1 − min(1, standard deviation / mean)).

Run four seconds through it. A steady typist producing 5 characters a second scores 60, 60, 60, 60 in per-second WPM: the standard deviation is 0, so consistency is 100.00. A bursty typist alternating 8 characters and 2 characters a second scores 96, 24, 96, 24 — the same 60 WPM average — with a standard deviation of 36, giving 100 × (1 − 36/60) = 40.00.

The bursty typist is not slower today, but almost all of their output is arriving in the good seconds, and the bad seconds are where the errors and hesitations live. Because the score is a ratio, it does not flatter fast typists: a 40 WPM typist and a 140 WPM typist with equally even rhythm score exactly the same. Chasing consistency rather than peak speed pushes you toward the pace you can actually hold, which is the pace that will still be there next week.

Practise the keys that are actually costing you

Generic practice improves generic weaknesses. After a test, Keyboardo's results screen shows a per-word breakdown of exactly where you slipped and a chart of your speed second by second with your errors marked on the timeline. Both are worth thirty seconds of attention, because they tell you two different things: which characters you get wrong, and when in a test you start getting them wrong.

If the errors cluster on particular reaches, practise those reaches specifically. Turning on the punctuation and numbers toggles adds symbols and digits to word and time tests, which is the closest this site gets to the awkward, off-home-row keys that everyday writing and code actually demand. If instead the errors cluster in the last twenty seconds of every 60-second run, the problem is stamina and attention rather than fingers, and shorter tests will not fix it.

What will not reliably make you faster

A great deal of typing advice promises gains nobody has measured. Three claims are worth deflating before you spend money or months on them.

• Switching keyboard layouts. No controlled study establishes a speed advantage for Dvorak or Colemak over a fluent QWERTY typist, and the best independent test found against retraining. The evidence is worth reading in full — see QWERTY, Dvorak and Colemak: what the evidence says.
• Buying a different keyboard. Comfort is real and subjective, and it is a perfectly good reason to prefer one board over another. A percentage speed gain from a keyboard shape is not a sourced figure, and any site quoting one is inventing it.
• Taking a formal course as such. In the same 168,000-participant study, people who had previously taken a typing course typed at similar speeds, used their hands similarly and made similar errors to people who never had. That finding is about credentials rather than practice — deliberate repetition still works — but it does undercut the idea that a course is the missing ingredient.

How to tell whether any of this is working

A single test result is noise. Your speed moves several WPM between runs on the same material for reasons that have nothing to do with skill, so improving by three WPM once tells you nothing at all.

Watch a trend instead. A free Keyboardo account records the average WPM of your last ten tests and charts your daily average over the last 90 days on your profile, alongside a heatmap of which days you actually practised. Those two views answer the two questions that matter — is the curve moving, and are you turning up — and neither can be faked by one lucky run. Without an account, your most recent 200 results are still kept in your browser.

For the practice loop those numbers are meant to measure, see building a typing practice routine. To take a measurement right now, run a 60-second test.