Perceptle scoring

Score calculator

How the total and grade work

Add the five event scores, divide by five, then round to the nearest whole number. Every event carries the same weight. The grade comes from that rounded total.

For example, 80 + 90 + 70 + 85 + 95 = 420. Divide 420 by five to get 84, which is a B. A raw average of 84.6 rounds to 85 and earns an A.

How each round becomes points

The original uses a Gaussian curve: an exact answer earns 100, and the score falls smoothly as error increases. The normal-mode formula is:

The tolerance puts different skills on a comparable scale. At one tolerance of error, the score is about 61. At two tolerances it is about 14. Tolerance is a scale for the curve, rather than a pass-or-fail cutoff.

Track’s units describe the game’s internal drawing field. They are not the number of physical screen pixels on your phone. Count and Split use proportions, so the same absolute error can mean different things at different target sizes.

Three concrete examples

• Count: The actual count is 60 and you guess 66. Relative error is 6 ÷ 60 = 0.10; dividing by the 0.20 tolerance gives 0.5. The round score is 88.
• Angle: An 8° miss is one tolerance and earns 61. A 4° miss earns 88. In normal mode, 359° and 1° are 2° apart.
• Split: A mark 5% of the line away from the correct point earns 61; a 10% miss earns 14.

What does Tempo measure?

The original can also report whole-beat early or late entry separately from its remaining timing error. Our fixed-reference drill keeps the starting phase, so the two timing scores can differ even with a similar tapping pattern.

Tempo compares your silent continuation with the expected pulse. The original calibrates that pulse to your audible warm-up taps, then combines 60% absolute average drift and 40% timing jitter across the continuation. The tolerance is measured in milliseconds; it is not a BPM difference or simply a count of missed taps.

Our standalone practice uses the same drift-and-jitter weighting against a fixed reference of four scheduled beats, rather than calibrating the reference to your warm-up taps. Listen to those four beats, then continue with eight silent taps. Its results are training feedback, not an official timing replay.

Do Friday scores work the same way?

The original keeps the same five-event, 100-point structure and rounded-average total. It uses more generous Friday tolerances for the harder tasks:

Friday Angle measures how far you wound the handle, so 450° and 90° are different answers even though they point in the same direction. Friday Split still measures error against the original line length when the requested point extends beyond an endpoint.

Our standalone exercises are labeled normal-mode practice. Their scores use the listed normal tolerances, but independently generated targets and device timing mean they are not an exact replay of an official puzzle.

What is a good Perceptle score?

85 earns an A; 95 earns an S. Those are defined grade thresholds. A percentile such as “top 10%” needs actual player comparison data for the relevant puzzle; the score alone cannot establish it.

For practice, compare your own repeat results in the same exercise. A useful target is to reduce the kind of error the result shows: fewer dots, less distance, fewer degrees or steadier timing. One weak round affects the average just as much as one strong round.

Work on your weakest round

Pick one skill, make an estimate, then use the revealed error to adjust your next attempt. Each exercise can be repeated without replaying all five rounds.

Diagnose your repeated mistakes and choose a practice method →