A rating that actually means something.
Speedcubing with a real ladder. The server hands you a scramble nobody has seen, replays your solve to prove it happened, and only then does it count. Plus the analysis that tells you which part of your solve to fix.
Solving needs no account. Free, and it works offline.
God's number is 20, so a solution within a move or two of that is about as short as a cube can be solved. This engine is in this repository — no solver library.
A solve, in seven steps
Four edges. The foundation, and the only part you plan in advance.
0 / 40 moves
Ranked
One number, and you can always read it in seconds.
Chess needs Elo because it has no absolute scale — you only ever learn that one player beat another. Cubing has seconds. So the rating is fixed to two landmarks the sport already uses, and every rating converts straight back into the average that earned it.
The server picks the puzzle
Every ranked scramble is generated when you ask for it, so there is nothing to cherry-pick and nothing to solve in advance.
Every solve is replayed
Your moves are applied to that exact scramble on the server. If the cube does not end solved, the result does not exist.
Failure costs certainty
Abandon a bad solve and your rating cannot rise — a failed average widens your margin instead of inventing a time for you.
A stopwatch tells you that you took 22 seconds.
It does not tell you that F2L pair three costs you 2.1 seconds every single solve, or that your OLL time swings between one and four seconds depending on which case you get. Those are different problems with different fixes, and the total hides both.
Phase splits
Tap through the stages as you solve, or connect a smart cube and let it read every turn. Either way you get cross, F2L, OLL and PLL separately.
The case, not the stage
It works out which OLL and PLL you actually faced, then ranks them by how much time you would get back — how often it comes up, times how slow it is.
Honest about itself
Nothing is claimed below a stated sample size, and a change inside normal variation is reported as no change instead of as progress.
Under pressure
A target that keeps tightening.
Every timer tells you your average afterwards. None of them ever put you in the position of needing this solve to be fast — which is what a competition round actually is. In Rush the number is on screen before you turn a face, and it shrinks every time you beat it. Three misses ends the run.
The target comes from your own pace, so it is the same difficulty whether you average eight seconds or forty. A fixed number would be a lazy solve for one and unreachable for the other; this finds the edge of what you can do today, which is the only place anybody improves.
Every day
One scramble. One attempt.
The same cube for everyone, resetting at midnight UTC. Your time is final the moment the clock starts — so the result you share is the one you actually got.
No cube in your hand?
Solve one with your keyboard. The clock starts on your first turn and stops the instant the cube is actually solved — no spacebar, no reaction delay at either end. It is the most accurate timing in the app, and it is what ranked runs on.
Try keyboard cubing →