How the leaderboard is verified
A ranking is only worth reading if you know what was checked. This page states exactly what is screened, what is not, and why the thresholds are what they are.
Updated 13 August 2026
The honest headline first. This leaderboard is not cheat-proof and cannot be. TypeCrt runs in your browser, so a determined person can construct a submission by hand. Everything below is about raising the cost of doing that and making it detectable afterwards — not about claiming it is impossible.
What gets submitted
A ranked run sends its keystroke timing record, not just a final score. That single decision does most of the work: faking a result stops being a matter of editing one integer and becomes a matter of generating several hundred inter-key intervals that survive the checks below.
The timings are stored with the entry. That matters for a reason beyond screening: if the checks improve later, every past entry can be re-examined against the better test. A board that discards its evidence can never correct itself.
What is checked
| Check | Rejects |
|---|---|
| Rhythm — coefficient of variation of intervals below 0.06 | Automated typing on a timer |
| Interval floor — any gap under 25 ms | Speeds no hand produces |
| Speed agreement — reported raw speed within 20% of what the timings imply | Scores edited after the fact |
| Accuracy agreement — net speed consistent with raw × accuracy | Internally contradictory results |
| Length — at least 60 keystrokes | Runs too short to screen meaningfully |
Why those numbers
The rhythm threshold is the one that does the heavy lifting, and it is not a guess. Dhakal, Feit, Kristensson and Oulasvirta measured 136,857,600 keystrokes from 168,960 people (CHI 2018). Their reported distributions give a usable floor for what human typing looks like:
| Group | Mean interval | Std. dev. | Implied variation |
|---|---|---|---|
| All participants | 238.66 ms | 111.60 | ~0.47 |
| Fastest 10% | 121.70 ms | 11.96 | ~0.10 |
| Slowest 10% | 481.03 ms | 123.36 | ~0.26 |
The fastest decile is the demanding case. These are the most rhythmically consistent typists in the dataset — the people whose defining trait is metronomic evenness — and they still vary by about 12 ms around a 122 ms interval. A script on a fixed timer lands an order of magnitude below that. Setting the floor at 0.06 leaves real record-holders clear headroom while catching mechanical input.
Human typing is irregular even at its most practised. That irregularity is the signature.
What is not checked
Stating these plainly is the point of the page.
- Who is typing. Nothing verifies that the person signed in is the person at the keyboard.
- Whether input came from a keyboard at all. A sufficiently good simulation of human timing variance passes, and the rationale for what "good" means is published on this page.
- Whether the text was visible. Copy-paste is handled elsewhere in the engine, but the board does not independently reconfirm it.
- Hardware advantage. Keyboard, switch type and polling rate are not accounted for.
What is ranked
English time tests at 15, 30 and 60 seconds. Everything else is measured and saved to your own history, but is not comparable between people and so is not ranked:
- KeyForge generates text weighted toward your weak keys, so it is deliberately harder for you than for anyone else. Ranking it would reward having easy weaknesses.
- Quote, custom and zen use text that differs between runs.
- Word-count modes could be ranked later; they are not yet, to keep the first version of the board small enough to watch.
Entries cannot be edited
Once written, an entry is immutable — the database rules refuse updates and deletions from any client. A leader cannot quietly revise their own row, and the evidence attached to it stays available for re-screening. Removing an entry is an administrative action taken outside the app.
Found a way through? That is a security report, and it is welcome: [email protected]. Result forgery is explicitly in scope on the security page, and good-faith research is covered by the safe harbour there.