Make the scoring curve visible
Perceptle error & score calculator
How much does a near miss cost? Enter the target and your estimate to turn a physical error into a score out of 100.
The Perceptle scoring formula
Each test uses round(100 × exp(−0.5 × (error / tolerance)²)). Zero error gives 100. One tolerance of error gives about 61; twice the tolerance gives about 14. The same scoring curve applies to five different units.
| Test | Error unit | Normal | Friday |
|---|---|---|---|
| Count | Difference ÷ actual count | 20% | 32% |
| Track | Distance on 800-unit field | 60 | 90 |
| Angle | Degrees | 8° | 14° |
| Tempo | Timing error | 60 ms | 85 ms |
| Split | Fraction of original line length | 5% | 10% |
Count example: 60 dots, estimate 66
The difference is 6 dots. Divide by the actual 60 to get 10% error. Normal tolerance is 20%, so the ratio is 0.5 and the result rounds to 88. The same miss receives 95 under the more generous Friday count tolerance.
Angles change their error definition on Friday
Normal compares directions using the shorter route around the circle: 359° and 1° differ by 2°. Friday compares accumulated rotation: 90° and 450° differ by a full 360°. Use the angle helper to see both on the same dial.
Tempo needs a timing error, not a BPM difference
BPM sets the period, while the score uses timing deviation in milliseconds. Independent Unlimited practice combines 60% of the absolute mean offset with 40% of jitter. The original game calibrates from warm-up taps, so an uncalibrated timing measurement is not automatically its exact error. The Tempo tool shows this fixed-reference calculation explicitly.
Can I reverse a score into an exact miss?
No. Rounding means several different errors can produce the same integer score. Even 100 does not always mean a perfect match. This calculator starts with your known target and estimate rather than inventing a unique error from a score.
To combine the five test scores into a grade, use the overall score calculator. To inspect an actual day, choose a puzzle from the answer archive.