The short answer
Add Count, Track, Angle, Tempo and Split, divide by five, then round. S starts at 95, A at 85, B at 70, C at 55 and D at 40. A lower score is F. The original uses a smooth error curve for each event.
In this guide
Work out your total and grade
Total = round((Count + Track + Angle + Tempo + Split) ÷ 5).
80 + 90 + 70 + 85 + 95 = 420; 420 ÷ 5 = 84, B. Improve the 70 to 73 and the sum becomes 423: 84.6 rounds to 85, A. You need three extra points across the five events in this example, not five.
| Grade | Total |
|---|---|
| S | 95–100 |
| A | 85–94 |
| B | 70–84 |
| C | 55–69 |
| D | 40–54 |
| F | 0–39 |
Enter your five scores and plan a target grade. Lock scores you do not intend to change to see whether the remaining improvement is mathematically possible.
How a miss becomes an event score
The reviewed original uses round(100 × exp(−0.5 × (error ÷ tolerance)²)). At half a tolerance of error you score about 88; at one tolerance about 61; at two about 14.
Tolerance is a curve scale, not a pass/fail limit. Because points are rounded integers, a small range of errors can share the same score, including 100. You cannot recover one unique physical error from a score alone.
84 / B+3 event points →85 / A
Normal and Friday units
| Test and error | Normal | Friday |
|---|---|---|
| Count: count difference ÷ actual | 0.20 | 0.32 |
| Track: horizontal landing distance | 60 units | 90 units |
| Angle: angular error | 8°; shortest circle distance | 14°; full-turn difference |
| Tempo: timing error | 60 ms | 85 ms |
| Split: distance ÷ original line length | 0.05 | 0.10 |
Track uses an 800 × 500 original field. Tempo’s reference is calibrated from warm-up tapping; independent fixed-reference tools do not reproduce that calibration merely by using the same tolerance.
Two useful score examples
Count: 60 actual dots and a guess of 66 gives 10% relative error. That is half the Normal tolerance, producing about 88 points.
Angle: an 8° miss is one Normal tolerance, producing about 61. In Friday, the same full-angle error is only 8/14 of its tolerance and produces about 85. Use the error-to-score tool to change values and see the curve.
What counts as a good score?
A is at least 85 and S at least 95 under the game’s own labels. Those are concrete targets. They do not tell you the percentage of players you beat; only a comparison based on actual players of the same puzzle can do that.
One B can contain a very strong Count and a weak Tempo, or five broadly similar results. Check the event scores and any available errors before choosing a practice task. Match your next practice to the pattern.
IO imports: do not average twice
Original daily sets have five events. Independent IO sessions can contain selected skills and multiple sets, so the analyzer averages the actual scored rounds.
For six scores 84, 85, 85, 84, 84 and 84, the sum is 506 and the final average rounds to 84. Rounding the two skill averages to 85 and 84 first would incorrectly produce 85. Missing skills do not count as zero. Imported results remain independent practice, not authenticated original-game submissions.
Rules meet a real date
Try these dated puzzles
These are actual targets reconstructed from the original’s public generator captured October 17, 2026. Reveal a target when you want the example; each date page includes hints and the five answers.
#92 · Normal
October 17, 2026
Reveal sample targets
59 dots · 77° · 127 BPM · Split 11/15 from the left.
#91 · Friday variant
October 16, 2026
Reveal sample targets
76 dots · 579° · 111 BPM · Split 8/5 from the left.
The examples do not submit a game or recreate a player’s historical attempt. See how targets, timing and scores are derived.
Choose your next step
Related guides
Original rules: Perceptle, reviewed October 17, 2026. Explanations and teaching diagrams are independently written. Sources and corrections.