Every move is a letter naming a face of the cube. The cube stays visible behind this guide — press Try on anything to watch it happen.
The hidden sides are their opposites: L left, B back, D down. Faces are named from your point of view — whichever side faces you is F.
Clockwise means as if you were looking straight at that face. So B turns the opposite way from F when seen from the front — try both!
The same three forms work on every face. All 18 possible face turns — click any to watch:
Algorithms read left to right, one move at a time. The most famous one, the “sexy move”:
Repeat it six times from a solved cube and everything returns to where it started. Undo works on tried moves, so experiment freely.
Three more families of moves — and the ' and 2 modifiers work on all of them too (M', x2, r' …).
No stickers change place relative to each other — a rotated solved cube is still solved.
A lowercase face letter turns that face plus the slice next to it. Also written with a w: r = Rw.
On 4×4 and up, a number in front picks how deep. A bare number turns just that one inner layer; with a w it turns that many outer layers together. (Buttons grey out if the move doesn’t exist on the current cube size.)
Slice moves M E S exist only on odd sizes (3, 5, 7) — even cubes have no middle layer. Rotations x y z work on every size.
Every position of the Rubik’s Cube — all 43,252,003,274,489,856,000 of them — can be solved in 20 moves or fewer. That worst-case number is called God’s Number: the number of moves an all-knowing solver would ever need.
For any scramble there is some shortest possible solution. God’s Number is the longest of all those shortest solutions — the diameter of the cube’s universe. In the half-turn metric (where R and R2 each count as one move) it is exactly 20; in the quarter-turn metric it is 26.
The superflip — every piece in place, every edge flipped — was proven in 1995 to require the full 20 moves. It looks like this (works on the 3×3):
| 1981 | ≤ 52 | Thistlethwaite’s nested-group algorithm |
| 1995 | ≤ 29 | Reid — and proves the superflip needs 20 |
| 1992–2008 | ≤ 22 | Kociemba’s two-phase idea, refined by Rokicki |
| 2010 | = 20 | Rokicki, Kociemba, Davidson & Dethridge — about 35 CPU-years donated by Google checked every position |
Knowing 20 is enough doesn’t make it fast to find a 20-move solution — truly optimal solving still takes a long search per scramble. The Kociemba tab here uses his two-phase algorithm: it first maneuvers the cube into a friendly subgroup, then finishes it, giving ~19–23 move solutions in milliseconds. Short, but not guaranteed shortest.
The 2×2 is small enough to search exhaustively: its God’s Number is 11, and the Optimal tab really does return a shortest-possible solution.
CFOP — Cross, F2L, OLL, PLL — is the method behind almost every modern speedcubing record, used by the overwhelming majority of competitive solvers. You’ll also hear it called the Fridrich method.
The core ideas were worked out independently by several cubers during the original cube craze of 1981–83 — Jessica Fridrich in Czechoslovakia, with early F2L pairing developed by Guus Razoux Schultz and Anneke Treep in the Netherlands. When speedcubing revived around 2000, Fridrich’s 1997 website was where the world learned the system, which is why her name stuck to it. The neutral name CFOP, after its four stages, credits the many contributors.
| ~55–60 | moves | typical full solve at this level of F2L; advanced solvers trim toward 50 |
| 78 | algorithms | full OLL + PLL — or just 16 for the two-look beginner version |
| ~10+ | moves/sec | elite turn speed; CFOP’s ergonomic, right-hand-heavy algorithms are why it stays on top |
For two decades CFOP was the elite: Feliks Zemdegs, Max Park, Tymon Kolasiński, Leo Borromeo and Luke Garrett rode it through every world record era of the 2010s and early 2020s. The picture has since shifted: as of August 2026, only one of the top 9 by WCA 3×3 average — Yiheng Wang — still solves CFOP; the rest, including Xuanyi Geng and Zhaokun Li, have moved to its extension, the ZB method. Among alternative systems, Roux keeps a strong minority, with ZZ and Petrus further behind.
Top solvers also stopped fixing a single cross color. As of August 2026, most are white–yellow dual neutral — during inspection they pick whichever of the two opposite crosses opens better — some remain white-only, and a few (like Zhaokun Li) are fully color neutral across all six. Try it yourself with the cross-color swatches in the CFOP and ZB tabs.
Learners start with two-look OLL/PLL and grow the alg set over time. Experts bolt on extensions: x-cross (cross plus a first pair at once), COLL, Winter Variation — and the ZB system in the ZB tab here: ZBLS inserts the final F2L pair while orienting the last-layer edges, letting ZBLL finish the whole last layer in a single look. Full ZBLL is 493 algorithms — one of speedcubing’s great memorization feats.
This app solves each CFOP stage greedily with no lookahead, exactly as the stages are defined — switch cross colors in the CFOP tab to watch the same scramble unfold six different ways.
ZB — Zborowski–Bruchem — is CFOP’s heavyweight upgrade: the last F2L slot and the last layer are merged into two huge algorithm sets, finishing the cube with one look fewer. Long considered unlearnable, it is now the method of choice at the very top.
Proposed around 2002–03 by Zbigniew Zborowski and Ron van Bruchem, ZB spent two decades as a thought experiment: the ~800 total algorithms were considered beyond human memory budgets, and solvers adopted only fragments (COLL, ZBLL subsets, occasional ZBLS cases). That changed in the mid-2020s, when a new generation learned the full system — and the top of the world rankings flipped from CFOP to ZB.
| ~45–50 | moves | typical full solve — roughly ten fewer than CFOP |
| ~800 | algorithms | 300+ ZBLS + 493 ZBLL (CFOP: 78) |
| 1 | look | the whole last layer is recognized and solved in one go |
Of the current top 9 in the world by WCA 3×3 average, only one — Yiheng Wang — still solves CFOP. The rest, including Xuanyi Geng and Zhaokun Li, use ZB.
Elite solvers no longer fix one cross color: during inspection they pick whichever color gives the best opening. As of August 2026, most top solvers are white–yellow dual neutral (choosing between the two opposite crosses), a few still solve white-only, and Zhaokun Li goes fully color neutral — any of the six. The cross-color swatches in the CFOP and ZB tabs let you compare exactly the way a color-neutral solver does.
The ZB tab shares its inspection, cross, and first three pairs with the CFOP tab, then switches to a searched ZBLS (edge orientation tracked inside the search) and a one-look ZBLL drawn from a precomputed table covering all 7,776 last-layer states with community algorithms.
Yau is the dominant method for the 4×4 (and larger cubes), developed by Robert Yau around 2009–2010. It is a smarter ordering of reduction — turning the 4×4 into a virtual 3×3 — that gets the cross finished almost for free.
Basic reduction solves centers, pairs all twelve edges, and only then starts the 3×3 stage — including a cross it has to dig for. Yau front-loads the cross while the cube is still free, so the 3×3 stage starts at F2L with perfect lookahead, and pairing happens in a comfortable, consistent grip. As of August 2026 it remains the method of choice for most of the world’s fastest big-cube solvers.
A reduced 4×4 can reach states impossible on a real 3×3, because the two wings of an edge are distinct pieces and center pieces are interchangeable. Two cases exist: OLL parity (a single edge appears flipped — an odd wing permutation) and PLL parity (two edge pairs appear swapped). Each has a dedicated algorithm that shuffles wings without breaking the reduction.
The Yau tab follows the stages above: greedy searched centers (built bar by bar), cross edges as exact wing pairs, and 3-2-3 edge pairing the way humans do it — one Uw slice per group, the pairs lined up around the cube with outer moves and created together when the slice closes, with only y rotations between groups. Both parities are detected on the virtual 3×3 and fixed up front, and a 3×3 stage is delegated to the same CFOP solver as the 3×3 tab — with its moves translated back to 4×4 notation. All six cross colors are computed, arriving in the background.
Roux is the main alternative to CFOP at the top level — a block-building method published by French cuber Gilles Roux in 2003. Instead of a cross and four slots, it builds two 1×2×3 blocks, solves the corners, and finishes everything else with only two kinds of moves.
Roux averages noticeably fewer moves than CFOP (≈45–50 STM vs ≈55–60) and needs no cube rotations at all after the first block. The price is heavy M-slice turning — which is also why Roux solvers count in STM (slice turn metric), where an M move counts as one turn. The M/U-only ending is famously fast and fingertrick-friendly, and makes Roux especially popular for one-handed solving.
As of August 2026, the fastest 3×3 Roux solver is Alexey Tsvetkov — the first person to record an official sub-4-second single with the method. But Roux shines brightest in 3×3 one-handed, where it is the main rival to CFOP: the low move count and the way M moves suit one-handed turning make it a natural fit. Prominent OH Roux solvers include Nicholas Archer, Crimson Arradaza, Sean Patrick Villanueva and Dwayne Ramos.
The Roux tab on this site follows the same stages: searched blocks (FB free, SB in <R,r,U,M>), one-look CMLL where the pool allows, and an <M,U>-only LSE — all counted in STM.
A Pyraminx move turns everything around one corner of the tetrahedron. Hold it with a face toward you and an edge at the bottom: the four corners are named U (up), L (front-left), R (front-right) and B (back).
There is no 2 suffix: corners have three positions, so two turns one way equal one turn the other way. All eight main moves:
Scrambles list the main moves first and finish with the tip twists, e.g. R U' B L R' U l' b. Only the six edge pieces need real solving — the four axial pieces just rotate in place and the tips are trivial — which is why the Pyraminx has just 933 120 meaningful states.
The Skewb is a corner-turning cube: every move spins half the puzzle 120° around a corner. Scrambles use fixed-corner notation — hold the puzzle with a corner pointing at you (white on top, green on the front-left). That corner never turns; the four moves twist the corners around it: U (top-back), R (bottom-right), L (bottom-left) and B (bottom-back).
The dot is the corner facing you — the one you hold. The B corner is directly behind it, at the bottom-back. Every corner of the puzzle can still reach every position: four axes are all a Skewb has.
There is no 2 suffix: corners have three positions, so two turns one way equal one turn the other way. All eight moves:
Official scrambles are 11 of these moves, e.g. R U' L B' U L' R B U' R' L. Despite the wild cuts the Skewb has only 3 149 280 states and can always be solved in 11 moves or fewer.
Pick a puzzle and set on the left, then choose exactly which cases to drill — whole subsets or single cases. Each scramble sets up a random case on the puzzle (cubes with yellow on top). Like any cubing timer: hold space until the timer turns green, release to start, and press any key to stop — the case and its algorithm are then revealed. Space deals the next case.
Cases and algorithms come from the site’s algorithm database (sources credited per set). Every set keeps its own recorded session in this browser — switch sets and back, your stats are waiting.