The most common typing-test plateau looks like this. You sit at 60 WPM with 96% accuracy. You decide to push for 70. You speed up, your accuracy drops to 88%, your net WPM after corrections is now 58. You've gone backwards.
This isn't a failure of effort. It's arithmetic. Every typing test on the internet, including tempo, reports net WPM — gross output minus the cost of your errors. The minute you push past your accuracy threshold, the cost line crosses your speed line and you lose ground. The question is how to know where that threshold is and how to move it.
The backspace tax
Here's what a single uncorrected mistake actually costs you. Say you misspell a word and notice three characters later. To fix it you have to: stop, backspace four times (the wrong character plus the three you've typed since), retype the correct character, retype the three characters you just deleted. That's nine keystrokes spent recovering from one wrong press — for a net of negative one useful character.
per discovered error
If you catch the error immediately — backspace once, retype once — the overhead is "only" 200%. If you catch it at the end of a sentence, you might burn 15 keystrokes. The math is brutal because every key you've typed after the error has to come back out.
This is also why pasted typing-test runs are easy to detect statistically: a human typing at 120 WPM has a non-zero error rate and a backspace pattern that follows it. A pasted run has neither. We use exactly this signal in tempo's anti-cheat — but that's a different post.
Where the threshold sits
For most typists, the curve looks like this:
- Below 80% accuracy — your net WPM is roughly half your gross. You're typing for show.
- 80–92% accuracy — you're losing 15–30% of your speed to corrections. Net WPM is meaningfully lower than gross.
- 93–97% accuracy — small but persistent tax. Net WPM is ~95% of gross.
- 98%+ accuracy — corrections are rare enough that gross and net nearly match. This is the zone.
The 98% number isn't magic, it's just where the corrections-per-minute rate becomes small relative to total throughput. At 80 WPM with 98% accuracy, you're making one mistake every ~25 seconds — a barely-noticeable interruption. At 80 WPM with 92% accuracy, you're making one every ~7 seconds, and your nervous system is constantly switching between "type forward" and "recover."
Why pushing past it backfires
When you try to type faster than your current skill allows, two things happen: your finger motions become less precise (more glancing strikes, more wrong keys), and your eye-finger lookahead breaks down (you stop reading ahead and start reacting to each character). Both of these increase your error rate, often non-linearly.
So the speed you "gain" by pushing isn't free — it's borrowed against future corrections, and the interest rate is brutal. A typist who consistently runs at the edge of their accuracy threshold will plateau there forever, because the only way to go faster is to first get more accurate at the current speed, then let speed creep up underneath.
How to actually train accuracy
You can't simply will yourself to be more accurate; the motor errors are happening below conscious control. What you can do is set the conditions for accuracy to improve. Three approaches that work:
1. Slow down deliberately
Run a 60-second test at 70% of your current best. If you're a 70 WPM typist, aim for 50. You should feel like you're typing absurdly slowly. Hold this pace for 5 tests. You'll get to 98–99% accuracy easily, and your fingers will memorize the right motion for each key. After a week of this, run a normal-speed test. Your accuracy will be higher than before because the correct motion is now the default.
2. Use quote mode for accuracy work
Random word lists are forgiving — common words follow predictable patterns and your fingers can predict the next letter. Quotes, with their capital letters, punctuation, and unusual word combinations, expose every weak transition you have. If you're trying to push from 95% to 98%, switch your practice to quotes for two weeks and stop running word-mode tests entirely.
3. Turn on "stop on error" in tempo
Tempo has a setting in the gear menu: stop on error → letter. This refuses to accept any wrong character. You have to correct it before continuing. It feels punishing at first — your gross WPM drops 30%. But it forces your fingers to stop relying on backspace, and after two weeks it builds the kind of accuracy that survives at speed.
The aggressive version is stop on error → word: an error means restarting the whole word. This is for typists already at 95%+ who want to push to 99%.
What "freedom mode" is for
The opposite setting in tempo is freedom mode: you can move past errors without correcting them. This is useful for one specific case — diagnostic tests where you want to see your raw motor speed without the discipline overhead. It is not good for daily practice, because it trains your fingers to be okay with errors. Use it once a month to check your top-line speed; otherwise leave it off.
The single best signal to watch
If you want one number to track your real progress, it's this:
This number cuts through the noise of any single great run. A 110 WPM personal best at 89% accuracy is impressive; a 75 WPM average at 98% is more useful. Office work, code, transcription, email — they all reward the second typist, not the first.
Open tempo's stats → history tab. The recent-run sparkline next to your average is the number to watch. If it's going up week over week, you're getting better. If your best-run number is going up but your average isn't, you're getting more inconsistent — and that's a signal to drop your target speed and train accuracy until the average catches up.
One last thing
Accuracy work is boring. Slow, deliberate, methodical typing at 70% of your max feels like backwards progress. The PR runs that move your peak number are the fun part. But every fast typist hits a plateau where the only way forward is dropping speed and rebuilding the floor under it. The plateau lifts when you stop trying to climb directly through it.
Take a week. Drop to a comfortable pace. Aim for 99%. Then come back and try for a personal best. It'll be higher than it was before.