How Many Retakes Before a WPM Score Is Real

Quick answer

> Quick answer: Five runs in one sitting, and take the middle one rather than the best. One test is noise. Three starts to look like a person. Five is where the number stops moving much.

Your best ever run isn't your speed. It's your ceiling on a good day.

I counted the spread across runs on our own leaderboard data. The gap between somebody's first and fifth test in a session is usually larger than a year of gear upgrades.

Why one test means nothing

A single test is one text, one minute, one mood.

Hit an easy passage full of short common words and the number flatters you. Hit one with capitals, numbers and an awkward name and it punishes you.

Add the fact you weren't warmed up and it's worse.

So the first score of the day is the least reliable number you'll produce, and it's the one most people screenshot.

The warm up effect is the biggest single factor

Almost everybody types faster on their third run than their first.

It isn't practice. Nothing improves in four minutes. It's that your hands settle into position and your reading catches up with your fingers.

The gain is real and it's often larger than people expect.

Which is why the honest thing is to discard run one entirely, or to warm up before you start counting at all.

Where the number stops moving

Run two is usually a jump. Run three is a smaller jump. By run five most people are drifting within a narrow band rather than climbing.

Past about seven runs fatigue starts pulling in the other direction, so more testing makes the number worse rather than more accurate.

Five is the practical middle. Enough to settle, not enough to tire.

If you only have time for three, warm up first and take all three.

Average, best, or middle

Best of five is the number people quote and the one that means least. It's your single luckiest run.

A plain average is better, but one disastrous run drags it down and misrepresents you.

The middle value is the most honest of the three. Sort your five scores and take the third. One bad run and one lucky run both stop mattering.

That's the number to write down.

Use the same test length every time

A one minute score and a five minute score are different measurements, not the same measurement taken twice.

Short tests run higher, because you can sprint through sixty seconds and you can't sprint through five minutes. I checked how the standard is defined, and a word is counted as five characters including the space, whatever the passage says. The standard measure hasn't changed in decades.

So comparing a one minute run against a five minute one tells you nothing about either.

We went through which length suits which purpose in the test duration guide.

Keep the same keyboard and chair

This sounds obvious and gets ignored constantly.

Half the retakes people run are on a different machine from the last set, in a different room, at a different desk height.

Any one of those moves the number more than the thing you were trying to measure.

If you're tracking progress over weeks, fix the setup first and only then start counting. Otherwise you're measuring furniture.

The laptop versus external keyboard gap alone is enough to wipe out a month of real progress, in either direction.

Accuracy has to travel with the number

A speed with no accuracy beside it is meaningless, because you can always type faster by caring less.

Every serious test reports a corrected score as well as a raw one, and the corrected one is what an employer means.

If your five runs vary wildly in accuracy, that variation is your real problem rather than the speed spread.

We explained the two numbers in gross versus net.

The day to day spread

Your score across five runs in one sitting is one kind of spread. Your score across five days is another, and it's wider.

Sleep, caffeine, how much you'd typed before you sat down, and how much the room distracted you all move it.

So a single session's middle value describes today. A week of sessions describes you.

For anything that matters, such as a job application, take the week version.

Write the date next to every number you keep. Without it you'll compare a Monday morning score against a Friday evening one and draw the wrong conclusion.

When you need a number for an employer

Test on five separate days, take the middle value from each, then take the middle of those five.

That number survives scrutiny, because a retest in front of somebody will land close to it.

Quoting your best ever run sets you up to underperform on the day, which is worse than quoting a lower honest figure.

Most employers retest anyway, so an honest figure costs you nothing and a flattering one costs you the interview.

Why your real world speed is lower anyway

Even a well measured test score overstates what you do at work.

A test gives you the words. Real typing makes you invent them, and thinking is slower than copying.

That gap is normal and large. It doesn't mean your test was wrong.

We took it apart in why your test speed beats your real speed.

Tracking progress over months

One session a week, five runs, record the middle. That's the whole protocol.

Plot those weekly middles and you get a line that actually means something. Plot every run you ever took and you get a cloud.

Progress at this is slow and lumpy. A flat month is normal and isn't a reason to change anything.

Watching it on the leaderboard is more motivating than a private spreadsheet, and it keeps you honest.

Don't retake until you beat it

The common failure is running tests until a good one appears, then stopping.

That guarantees your recorded number is your luckiest run and that it drifts upward without your typing changing at all.

Decide the number of runs before you start. Five. Then stop, whatever the last one said.

A number you chose in advance is a measurement. A number you stopped at is a wish.

What a bad run is telling you

One run well below the others usually has a cause you can name.

A hard passage with unusual words. A hand slipping off home position early and never recovering. Somebody talking to you.

Worth noting, not worth panicking over. That's exactly what taking the middle value is for.

If three of five are bad, though, that isn't noise. That's the day, and the honest move is to stop and test tomorrow.

The useful habit is one word of context beside each score. Tired. Noisy room. Hard passage. It takes a second and it explains the outliers later.

Racing changes the picture

A race against other people produces a different number again, usually a little lower, because you're watching someone else's progress bar instead of the text.

That pressure is useful training and a poor measurement.

Keep the two separate. Race to improve, test alone to measure.

Race to improve, then test alone when you want a number you can write down.

A protocol you can actually follow

Warm up for two minutes.

Run five tests of the same length, back to back, same keyboard.

Sort the five scores and write down the third.

Do the same thing once a week, on the same day, at roughly the same time.

That's it. It takes ten minutes a week and it gives you a number you can defend, which is more than most people who quote a speed can do.

Where percentiles fit in

Once you have an honest number, the useful question is where it sits among everyone else.

That's a more meaningful comparison than chasing a round figure like a hundred, because the distribution is uneven and the gaps between bands aren't equal.

A few words a minute in one band moves you far more than in another.

Our percentile breakdown has the bands.