About Perceptle
An independent daily game about how well you see, track and time the world.
Perceptle is a small, independent browser game. It runs entirely in your browser: there is no account, no leaderboard server and no database of players. It is made for fun and for the satisfaction of a tidy two-minute ritual.
Not affiliated. Perceptle is an independent game. It is not affiliated with, endorsed by or connected to any other game with a similar name, including the one at perceptle.com. The tests, wording, artwork, sounds, scoring and share format here were created for this site. We make no claim to any trademark.
Not a diagnostic. Perceptle is a game, not a medical, psychological or vision test. Do not use your score to judge your eyes, hearing, attention or health.
Where the daily puzzles come from
Every day's dot count, ball path, target angle, beat tempo and fraction come from a seeded random number generator keyed by the day number (day #1 is 13 November 2026, and days change at 00:00 UTC). Parameter ranges ramp up a little through the week: Monday is the gentlest, Sunday the toughest. A script checks 400 days ahead to make sure every day is playable: dots never overlap, the ball stays on the stage and ends behind the screen, tempos are humanly tappable and fractions are never trivial.
Because the site is static, the puzzle is calculated in your browser from the date, so someone who reads the code could look ahead. That is one reason there is no global leaderboard. The archive shows the target values of every finished day.
Where the percentile label comes from
The results screen says things like "Ahead of about 60% of simulated players". The words matter. We have no statistics from real players and we do not invent any. The label is estimated from simulated players:
- We simulate 100,000 imaginary players. Each one plays the five tests with the real scoring formula from the how-to-play page.
- Each simulated player has a noise level: how far, on average, their answers wander from the truth. Overall skill varies between players (a log-normal spread, sigma 0.35), with an additional spread per test (sigma 0.2).
- The base noise levels are our own rough assumptions, loosely inspired by the order of magnitude found in psychology textbooks, not fitted to any dataset:
- Glimpse: typical error of about 20% (standard deviation of the log ratio, 0.2).
- Vanish: typical error of about 32 px on each axis of the 400 px stage.
- Tilt: typical error of about 8 degrees.
- Echo: typical gap error of about 7% of the beat, plus a slow drift of about 3%.
- Split: typical error of about 5 percentage points.
- We sort the simulated totals and keep the 101 percentile points (see
site/js/reference.js). Your total is placed on that curve.
The simulated average is 319 out of 500 (glimpse 66, vanish 53, tilt 59, echo 71, split 70). Real players will differ, perhaps a lot, and the label will never be updated from your results because your results never leave your device. Treat it as a friendly yardstick: "you did better than a typical imaginary player", not a ranking among people.
Credits and tools
Perceptle is plain HTML, CSS and JavaScript: canvas for the stage, pointer events so mouse, pen and touch behave the same, and the Web Audio API for the beat. There are no frameworks and no tracking pixels. When analytics are switched on, only anonymous visit and game-event counts are collected (see the privacy page).
Questions, bug reports or ideas for new tests? Use the contact page.