How we measure your typing

A word is 5 characters, including spaces. This is the standard convention and it is what makes typing speeds comparable between people typing different text.

gross WPM = (all characters you typed ÷ 5) ÷ minutes

net WPM = (characters you got right ÷ 5) ÷ minutes

The headline number everywhere in Typing Coach is net WPM. Gross WPM is shown beside it so you can see what your speed would be if accuracy were not costing you anything.

We deliberately avoid the other common definition of net WPM (“gross minus errors per minute”). Mixing the two conventions is the main reason typing scores from different sites do not line up, and using both would mean showing you two different numbers for the same run.

accuracy = correct keystrokes ÷ all keystrokes × 100

error rate = 100 − accuracy

Accuracy is measured on first attempt. If you mistype a character and fix it with backspace, the mistake still counts. Hiding it would make your accuracy look better and would make the weakness analysis wrong - the whole point is to find the keys your fingers get wrong, not the keys you failed to rescue.

Error rate is always exactly the complement of accuracy. It is never a separate count, so the two can never disagree.

Consistency measures how even your rhythm is - not how fast you are. A typist who holds one steady pace scores higher than one who sprints through easy words and stalls on hard ones, even at the same average speed.

CV = standard deviation of keystroke gaps ÷ mean gap
consistency = 100 × (1 − CV), clamped to 0-100

Gaps longer than 1000ms are excluded and counted separately as pauses. Without that, stopping to read the next line would look like bad rhythm, which would penalise careful learners rather than uneven ones.

Fewer than five usable gaps is not enough signal, so consistency is reported as unavailable rather than as a bad score.

A key can be weak in two different ways, and they need different practice:

• Inaccurate - you hit the wrong key. Fixed with slow, deliberate precision work.
• Slow - you hit the right key, but hesitate. Fixed with rhythm and repetition.

“Slow” is always measured against your own median keystroke, never an absolute threshold. A key counts as slow when it takes more than 1.35× your typical keystroke time. A 90 WPM typist’s worst key may still be faster than a beginner’s best.

Accuracy at or above 97% counts as mastered. Below 8 attempts we do not judge a key at all, and estimates with little evidence are pulled toward 100% so that one unlucky miss on a rare key cannot dominate your training.

Severity also weights each key by how often it really appears in English. Fixing a weak e is worth far more of your time than fixing a weak ~.

Transitions are pairs of consecutive characters. Moving between two keys is a separate motor skill from either key alone, which is why you can be accurate on P and on O and still lose time on PO. This is where most speed plateaus actually live.

The recommendation engine is a fixed set of rules evaluated in priority order. It is deterministic - the same profile always produces the same recommendation - and every recommendation shows the evidence behind it. There is no model and no external service involved.

1. 1. No data yet

   Take the placement assessment. We do not guess your level, and we do not ask you to.

2. 2. Accuracy below 93%

   Everything else waits. Speed built on errors has to be unlearned later.

3. 3. A weakness above severity 18

   Drill that key, transition or finger, from the isolated key up to real prose.

4. 4. Consistency below 55%

   Long passages at one steady pace, to even out the rhythm.

5. 5. Curriculum unfinished

   Continue to the next lesson.

6. 6. Everything clean

   Push the ceiling with short speed bursts and recovery runs.

When the engine finds a weakness it builds a ladder around it, and every rung uses the same target keys:

key → combination → word → sentence → paragraph → real-world text

So a weak P and a weak PO transition produce something like:

The motor pattern you learn on the isolated key is immediately rehearsed in context, which is what makes it stick outside the drill.

A one-minute test measures your peak; five minutes measures what you can hold. The endurance track splits the run into whole minutes and reports each one on its own, because a decline is invisible in a running average.

drop-off = (first-minute WPM − final-minute WPM) ÷ first-minute WPM × 100

Below 4% counts as holding steady; 8% or more is a real decline rather than noise. A negative drop-off means you finished faster than you started.

There is deliberately no single “endurance score”. Drop-off, final-minute speed and per-minute accuracy each mean something concrete; blending them into one number would hide which of them is the problem.

Passing a lesson once does not mean you still own those keys three months later. Mastery is never stored as a flag - it is recalculated from what you have actually been typing since, from two independent factors:

mastery = accuracy score × retention score

• Accuracy - how correct you are right now on the keys that lesson taught. Full marks at 98%, zero at 80%.
• Retention - how recently you last used them, halving every 30 days and resetting the moment you type them again.

Both are needed. Accuracy alone would never flag the symbol lesson you passed in March and have not touched since. Retention alone would nag someone who is demonstrably still perfect at the material.

The result lands in one of four bands, and every lesson carries a sentence saying which factor moved it:

If your typing on a lesson’s keys is still strong and you use them daily, it never decays. Only genuinely unused or deteriorating material comes back.

Typing ability changes. A key you fixed three months ago should not keep being recommended because of how it used to look, so every stored measurement decays with a half-life of 45 days. Recent sessions dominate; old ones fade rather than being deleted outright.

Sessions shorter than 12 seconds do not move your headline speed and accuracy. They are drills, not measurements, and letting a five-second burst set your profile would make the numbers meaningless.

Individual keystrokes are never saved. Your browser aggregates a session as you type and sends only the totals: how many attempts and how many correct for each key, letter pair and finger, plus the session summary.

Without an account, none of it leaves your browser at all. It lives in local storage on your device and you can clear it at any time.