What 136 Million Keystrokes Reveal About Fast Typing

In 2018, researchers at Aalto University published an analysis of 136 million keystrokes collected from 168,000 people taking an online typing test. It remains one of the largest datasets on how people actually type, and several of its findings sit awkwardly with how typing is usually taught.

Three are worth knowing.

The average is 52 words per minute

Across the full sample, average typing speed was around 52 WPM. That is higher than the 40 WPM figure that circulates widely, and the gap is worth understanding: the participants were people who chose to take an online typing test, which selects for computer users rather than the general population.

Treat 52 as the average for people who type regularly, and treat the lower figures as the average across everyone. Neither is wrong; they are measuring different groups.

Fast typists overlap their keystrokes

This is the finding with the most practical weight.

Fast typists do not press one key, release it, then press the next. They begin the next keystroke while the previous one is still being pressed — an overlap called rollover. In the study, fast typists used rollover on somewhere between 40% and 70% of keystrokes.

The striking part is that this was true regardless of formal training. Self-taught typists who had never been shown a technique were rolling over at similar rates to those who had. The pattern emerges from practice itself, not from instruction.

That reframes what “typing faster” means. It is not primarily about moving each finger more quickly. It is about the degree to which movements overlap.

Coordination beats per-key speed

The third finding follows from the second, and it is the one that should change how you practise.

The researchers found that inter-hand and inter-finger parallelism predicted overall typing speed better than how quickly any individual key was struck. How well your hands work simultaneously matters more than how fast either one moves alone.

This is a direct argument against a very common approach to typing practice: drilling one letter at a time until it is fast, then adding the next. That method optimises per-key speed, which the data says is the weaker predictor. It also isolates each key from the transitions around it — and transitions are precisely where parallelism lives.

The implication is not that letter drills are useless. It is that they train the wrong variable if used alone. A key you can strike quickly in isolation but cannot flow into from the previous key has not been learned in the way that produces speed.

What to practise instead

If coordination is the target, a few things follow.

Practise sequences, not keys. Common bigrams and trigrams — th, er, ing, tion — are where hand alternation and rollover actually happen. They are also unavoidable in real text, so the practice transfers immediately.

Favour text with natural letter distribution. Words drawn from real frequency are dense in the transitions you need. Artificial sequences built to feature one target letter are dense in one thing and sparse in everything else.

Do not force uniform speed. Skilled typists are not uniformly fast. They accelerate through familiar patterns and slow at awkward ones. Trying to hold one constant pace across everything suppresses exactly the variation that rollover produces.

Let alternation do the work. Sequences that alternate hands are naturally faster than same-hand sequences, because the two hands can overlap freely. Same-hand sequences, especially same-finger ones, cannot overlap at all — which is why they feel slow no matter how well you know the keys.

Where this leaves single-letter drilling

Adaptive systems that introduce one letter at a time and hold it until it is fast are optimising per-key speed on the assumption that the whole is the sum of its parts. This dataset suggests the whole is substantially about the joins.

That does not make such systems worthless — they are effective at building initial key location knowledge, which has to come from somewhere. But it does suggest a limit. Once you know where the keys are, continuing to drill them individually stops addressing what separates a 50 WPM typist from an 80 WPM one.

Measuring your own coordination

You cannot see rollover directly in a typing score, but consistency is a reasonable proxy. A consistency figure compares your speed second by second across a test: a high score means an even rhythm, a low one means you alternate between bursts and stalls. Stalls are usually where a sequence failed to overlap.

If your peak speed is decent but your consistency is poor, the problem is unlikely to be individual keys. It is far more likely to be specific transitions that break your flow — and those are findable by paying attention to where you hesitate.

Where to start

A longer test surfaces coordination problems that a short burst hides. The 5 minute typing test reports consistency alongside speed, and the decline between your first and last minute is itself informative.

Track it over time on the statistics page — the trend across many tests tells you far more than any single result. And if you want a quick baseline before any of that, the typing test takes a minute.

---

Source: Dhakal, Feit, Kristensson and Oulasvirta, Observations on Typing from 136 Million Keystrokes, CHI 2018. Project page