
CFOP Method: 2-Look OLL, PLL and Intuitive F2L
The CFOP (Fridrich) method explained: cross, F2L, OLL and PLL, its real history, and the 16 algorithms of the 4-look last layer with diagrams.
In short
The CFOP method solves the 3x3 Rubik's Cube in four stages: cross (the four edges of one face, usually on the bottom), F2L (the first two layers, placing each corner together with its edge), OLL (making the top face a single color) and PLL (moving the top pieces into their correct spots). It is the most widely used method in speedcubing and is also known as the Fridrich method.
- The full method has 41 F2L cases, 57 OLL cases and 21 PLL cases
- To get started you only need 16 algorithms: 10 for 2-look OLL and 6 for 2-look PLL
- The cross and F2L can be solved intuitively, with no memorization
- Several people developed it between 1979 and 1983; Jessica Fridrich spread it online in 1997
Almost everyone who learns to solve the cube with the layer-by-layer method ends up wondering how people get from over a minute down to ten or fifteen seconds. In the vast majority of cases, the answer is CFOP. It is not a fundamentally different method: it is the layer-by-layer method taken further, with steps grouped together to solve more pieces at once and with algorithms designed to be executed quickly.
This CFOP tutorial covers what each stage does and how many cases it has, where the method came from (its history involves more people than is usually told), how to move from the beginner method to CFOP without learning 78 algorithms at once, and the 16 algorithms of the 4-look last layer, each with its own diagram. We use the usual CFOP orientation: white on the bottom and yellow on top, with green in front. If any letter in the algorithms is unfamiliar, the notation article explains what each turn means.
What is the CFOP method
CFOP stands for its four stages: Cross, First two layers, Orientation of the last layer and Permutation of the last layer. You first solve a cross on one face, then complete the bottom two layers in pairs of pieces and, finally, solve the top layer in two parts: first you orient it (all the top stickers the same color) and then you permute it (each piece to its place).
Cross
4edgesThe four white edges, placed on the bottom and lined up with the side centers. It is solved intuitively and always takes 8 moves or fewer.
- No algorithms
- Planned during inspection
F2L
41casesFour pairs, each made of a white corner and a middle-layer edge, joined on top and inserted together into their slot. It is the longest stage and the one that saves the most time.
- Intuitive at first
- 4 pairs
OLL
57casesWith one algorithm, the top face becomes fully yellow, even though the sides do not match yet.
- 2-look: 10 algorithms
- 1-look: 57
PLL
21casesWith another algorithm, the top pieces swap places without twisting and the cube is solved.
- 2-look: 6 algorithms
- 1-look: 21
These case counts exclude the already-solved case and group together positions that are the same seen from another side. With everything learned, that makes 57 + 21 = 78 algorithms for the last layer, and 119 if you also memorize one algorithm for each of the 41 F2L cases. But nobody starts there, and you don't need to: the cross and F2L are solved by thinking, and the last layer can be done in four short steps with far fewer algorithms.
Where it came from: Fridrich and the others
For years, forums took it for granted that the whole method had been invented by Jessica Fridrich, a Czech cuber who won the first Czechoslovak championship in 1982. Today the community tells the story differently, because its building blocks appeared earlier and in several places at once:
- The cross already appears as the first step of a layer method in a 1978 work by Donald Taylor, and David Singmaster spread it in his notes on the cube from 1979 onward.
- The F2L idea, pairing a first-layer corner with its second-layer edge and inserting them together, was published in 1979 in two works by John Conway, David Benson and David Seal, with tables for every case. In 1981 it was also described by the Dutch newsletter Cubism For Fun, which credited it to René Schoof.
- Orienting and then permuting the last layer (OLL, then PLL) was developed along two independent paths from 1981: in the Netherlands, by Hans Dockhorn and Anneke Treep; in Czechoslovakia, by Jessica Fridrich with help from Mirek Goljan, her rival in that 1982 final.
Fridrich herself explains on her website that she created most of her last-layer algorithms between the summer of 1981 and the spring of 1983, with important contributions from Goljan, and that later, in 1982, she switched her first two layers to the pairing system, which took several seconds off her times. There are two versions of where that technique came from: English Wikipedia links it to the Dutch cuber Guus Razoux Schultz, second at the 1982 World Championship in Budapest (24.32 seconds, where Fridrich finished tenth with 29.11), while she herself said in 2003 that she learned it from university friends and did not know its exact origin.
Her decisive role came later. In January 1997 she published her system online, with every algorithm, based on her old notebook from the 1980s. It was one of the few complete sources available when speedcubing was revived in the late nineties, and many people learned from it: Dan Knights, who won the 2003 World Championship (the first in 20 years) with an average of 20.00 seconds, had learned the method on her website. That is why it became popular as the "Fridrich method".
Before you start: from beginner method to CFOP
CFOP builds on what you already know if you solve the cube layer by layer. If you can't solve it yet, start with our guide on how to solve the Rubik's Cube step by step: the cross is the same, and several algorithms used there come back here. What changes is that the steps are grouped together.
| Layer-by-layer method | CFOP | What you gain |
|---|---|---|
| White cross on top, then flipped to the bottom | Cross directly on the bottom | No need to turn the cube over, and you can see the pieces for the next step |
| First-layer corners, then second-layer edges | F2L: corner and edge together | Far fewer moves (J Perm estimates about 20 per solve) |
| Last layer in 4 or more steps | OLL and PLL: 4 steps at first, 2 at the end | Fewer "looks" and shorter algorithms |
The usual path is to do it in stages. First you learn OLL and PLL in two steps each (2-look OLL and 2-look PLL, together known as the 4-look last layer, or 4LLL): 16 algorithms that already let you solve the last layer the same way every time. Then you move on to intuitive F2L, which is where you gain the most time but also what takes the most practice. Improving your cross is left for later. J Perm, author of one of the most followed CFOP tutorials, recommends exactly that order and places the jump to full OLL and PLL at around a 30-second average.
| Stage | What you learn | New algorithms |
|---|---|---|
| 1 | 2-look OLL (edges, then corners) | 10 Start here |
| 2 | 2-look PLL (corners, then edges) | 6 |
| 3 | Cross on the bottom and intuitive F2L | 0 (worked out by reasoning) |
| 4 | Full PLL | 21 (includes the previous 6) |
| 5 | Full OLL | 57 (includes the previous 10) |
Sources don't agree on a fixed order for stages 4 and 5; many cubers learn full PLL first because it has fewer cases and comes up in every solve.
Step 1: the cross, on the bottom and planned
The goal is the same as in the beginner method: the four white edges on the white face, each with its other color lined up with the side center. The difference is where and when.
Where: on the bottom face. Making the cross on top and then flipping the cube is convenient for learning, but when you turn it over you lose sight of the pieces you need right afterwards. Practically every CFOP solver builds it directly on the bottom. It is hard at first because you can't see the white face; you learn by watching the sides and checking underneath now and then.
When: during inspection. In competition, the WCA regulations give you less than 15 seconds to look at the cube before starting, without turning any face. That time is used to find the full cross solution and then execute it without stopping. What makes this possible is why the cross matters so much: it can always be solved in 8 moves or fewer, and the optimal solution averages about 5.8 moves if you always use the same color.
Two improvements come later. The first is color neutrality: being able to start with any color instead of only white. According to Lars Vandenbergh's cross study, cited by Feliks Zemdegs on CubeSkills, the average optimal cross drops from 5.81 to 4.81 moves, and the chance of a cross of 4 moves or fewer rises from 6% to 29%. The second is the X-cross: solving the cross and the first F2L pair in one go. Both are for when everything else already works.
Step 2: intuitive F2L
With the cross on the bottom, there are four slots left between the side centers, one at each corner of the white face. Each slot is filled with two pieces: a white corner and the second-layer edge that shares its other two colors (the white, green and orange corner goes with the green and orange edge). In the beginner method they are placed separately; in F2L they are joined on top and inserted into their slot in one go.
F2L has 41 different cases for a pair, and there are algorithms for all of them, but the general advice is to learn it intuitively: understand why each move works instead of memorizing. Almost everything comes down to three ideas:
- Find the pair. Look for a white corner in the top layer and its edge, also on top.
- Join them or set them up. Move them with top-layer turns (and by lifting and lowering an empty slot) until they sit together as a 1x1x2 block, or in one of the positions that insert directly.
- Insert them. With three or four moves that drop the pair into the slot without disturbing the others.
These are the four basic insertion cases, all with the slot you are filling at the front right. Every other case is reduced to one of these.
| Where the pieces are | Algorithm |
|---|---|
| Corner directly above the slot with white facing right; edge on top at the back center | R U R' |
| Corner directly above the slot with white facing you; edge on top at the left center | F' U' F |
| Corner and edge already joined along the right side of the top layer, with the corner's white facing you | U R U' R' |
| Corner and edge already joined along the front side of the top layer, with the corner's white facing right | U' F' U F |
Notice that the first two are the same move seen from two sides, and so are the third and fourth. If the slot is somewhere else, rotate the whole cube to bring it to the front right, or use the left-hand equivalents (L' U' L instead of R U R', for example).
For reference, Jessica Fridrich estimated an average of about 7 moves per pair, which puts F2L at around 28 moves. It is by far the longest stage of CFOP.
Cubes for every level
From a first magnetic 3x3 to competition cubes and less common shapes such as the Megaminx or the Pyraminx. Take a look at our catalogue.

QiYi Crazy Gear Cube 3x3 - Puzzle con Engranajes
QiYi 2x2 Mini Pillowed Llavero Stickerless Removible
QiYi QiHeng S2 Megaminx Stickerless Dodecaedro
QiYi Void 3x3 Magnético UV Coated StickerlessStep 3: 2-look OLL
With the first two layers solved, the goal is to make the top face entirely yellow. Full OLL does this with a single algorithm chosen from 57; the 2-look version splits it like this:
- Orient the edges to form the yellow cross (3 cases).
- Orient the corners to complete the face (7 cases).
In the diagrams, the top face is seen from above, with the side facing you at the bottom. Yellow stickers are already oriented; gray ones are not; the strips around the edges are the side stickers of the top layer. In the first step the corners are shown in light gray because they don't matter yet.
First, the yellow cross
Look only at the top edges. You can have the cross already done, a line of two opposite edges, an L of two adjacent edges, or just the center (a dot). If you have the cross, move on to the corners.
Then, the corners
With the yellow cross done, count how many corners have yellow on top (none, one or two) and notice which way the yellow stickers of the others are facing. Each combination matches one of these seven cases.
If you have two corners on top and don't recognize the position, turn the top layer with U until it matches the diagram: it is the same case, just rotated. And if memorizing all seven is hard, the Sune and Antisune are enough to get out of any of them by repeating them, at the cost of more steps.
Step 4: 2-look PLL
Now the top face is yellow, but the side colors don't match. PLL moves the top pieces without twisting them. In two looks it works like this:
- Permute the corners (2 cases).
- Permute the edges (4 cases).
In these diagrams the side strips show their real colors and the arrows show which pieces swap. After each algorithm you may need to turn the top layer (U, U' or U2) to line it up with the rest.
First, the corners
Look at the four sides of the top layer for headlights: a side where both corners have the same color. If all four sides have them, the corners are done; move on to the edges. If only one side has them, it is the first case. If none do, it is the second.
Then, the edges
With the corners correct, look at the edges. If there is a solved side (all three pieces the same color), it is one of the two U perms; if there is none, it is the H or the Z.
In the H and Z perms, M is the middle layer between left and right, which turns in the same direction as L; if it feels strange, the notation article explains it with diagrams. With these 16 algorithms you can already solve any cube with CFOP using a 4-look last layer.
For CFOP, the cube matters too
OLL and PLL algorithms rely on fast finger tricks, and a speed cube that turns smoothly and doesn't lock up helps much more than it does with the beginner method.
See speedcubing cubesReference move counts and times
There is little reliable data on how long each level takes, because it depends so much on practice. What is documented is Jessica Fridrich's own breakdown of her system, with which she averaged about 17 seconds in 1983:
| Stage | Average moves | Time |
|---|---|---|
| Cross | 7 | 2 s |
| F2L (4 pairs) | 4 × 7 | 8 s |
| OLL | 9 | 3 s |
| PLL | 12 | 4 s |
| Total | 56 | 17 s |
Today's numbers are similar: a CFOP solve with full OLL and PLL takes around 55 to 60 moves. Some practical benchmarks from J Perm: with training, the cross takes under 5 seconds, and most cubers consider learning full OLL and PLL when their average is close to 30 seconds.
A curious fact you can calculate: the top layer has 216 orientation combinations and, once the final layer turn is discounted, 72 permutation combinations, so without forcing it you will find the yellow face already done once every 216 solves and the pieces already in place once every 72. These are the well-known OLL skip and PLL skip.
Alternatives: Roux and ZZ
CFOP is not the only competition method. The other two used at the highest level are Roux and ZZ, and they are worth knowing even if you stick with CFOP.
Roux
Created by Frenchman Gilles Roux in the early 2000s. It builds a 1x2x3 block on each side, solves the top corners with one algorithm and finishes the six remaining edges almost entirely with M and U turns.
- About 45-50 moves
- No cube rotations
- Very popular for one-handed
More detail
Its corner step, CMLL, has 42 cases, and can be done in two looks with fewer algorithms. The rest is intuitive, so it needs less memorization than CFOP but more understanding of how the pieces move. Middle-layer turns are its weak point for very fast turning.
ZZ
Associated with Polish cuber Zbigniew Zborowski, who developed it in 2006. It starts by orienting all the edges while making the cross (EOCross) or a line (EOLine), then solves the rest using only R, U and L turns.
- About 45-54 moves
- No rotations after the first step
- Last layer with edges already oriented
More detail
The idea of orienting the edges at the start had been proposed by Ryan Heise in 2003, but the method became popular through Zborowski, who named it after his initials. Since the edges reach the last layer already oriented, there are fewer possible cases, although the first step is hard to plan during inspection.
| CFOP | Roux | |
|---|---|---|
| Average moves | 55-60 | 45-50 |
| Algorithms to master it | 78 for the last layer (16 to start) | 42 for corners (fewer in two looks), the rest intuitive |
| Cube rotations | Some | None |
| Style | Recognize cases and execute them fast | Build blocks and reason |
| Learning resources | By far the most plentiful | Fewer, but enough |
| Moving on from the beginner method | Direct: they share the cross and the layers | You start almost from scratch |
For anyone coming from the layer-by-layer method, CFOP is the natural next step and the method with the most material, tutorials and studied algorithms. Roux suits people who enjoy thinking through the solution and want less memorization; many of the best one-handed times of recent years have been set with it.
World records and CFOP
The WCA records results, but not the method each competitor uses; that has to be found in other sources. These are some 3x3 world records set by cubers who are documented as using CFOP:
| Year | Cuber | Record | Competition |
|---|---|---|---|
| 2003 | Dan Knights (USA) | 20.00 s average | World Championship 2003, Toronto |
| 2011 | Feliks Zemdegs (Australia) | 5.66 s single | Melbourne Winter Open 2011 |
| 2018 | Feliks Zemdegs (Australia) | 4.22 s single | Cube for Cambodia 2018 |
| 2019 | Feliks Zemdegs (Australia) | 5.53 s average | Odd Day in Sydney 2019 |
| 2023 | Max Park (USA) | 3.13 s single Record until 2025 | Pride in Long Beach 2023 |
As of October 2026, according to the WCA, the 3x3 single record is 2.51 seconds, set by Xuanyi Geng (China) at Guangzhou Grand Open 2026, and the average record is 3.51 seconds, set by Yiheng Wang (China) in June 2026. Among today's elite it is common to extend the CFOP base with larger algorithm sets for the last pair and the last layer, such as COLL, Winter Variation or ZBLL, which more often let them solve the last layer with a single algorithm.
Frequently asked questions
How many algorithms does the CFOP method have?
The full method has 78 for the last layer (57 OLL and 21 PLL), and 119 if you also memorize one algorithm for each of the 41 F2L cases. To get started, 16 are enough: 10 for 2-look OLL and 6 for 2-look PLL, with the cross and F2L done intuitively.
Who invented the CFOP method?
Several people between 1979 and 1983. The cross and the idea of pairing pieces in F2L had already been published in 1979; the OLL-then-PLL order was developed at the same time by Hans Dockhorn and Anneke Treep in the Netherlands and by Jessica Fridrich with Mirek Goljan in Czechoslovakia. Fridrich spread it online in 1997, which is why it is also called the Fridrich method.
What is the difference between the beginner method and CFOP?
CFOP groups the steps together: the first and second layers are solved at the same time in pairs (F2L) and the last layer is solved with two algorithms (OLL and PLL) instead of four or more. The cross is the same, but it is built directly on the bottom face.
What are 2-look OLL and 2-look PLL?
They are the two-step versions of the last two stages. 2-look OLL first orients the edges (3 algorithms) and then the corners (7); 2-look PLL first permutes the corners (2) and then the edges (4). They are the recommended starting point before full OLL and PLL.
Why is the cross done on the bottom?
Because that way you don't have to flip the cube after finishing it, and the F2L pieces stay in view in the top layer. Also, the cross always fits in 8 moves or fewer, so it can be fully planned during the 15 seconds of inspection.
CFOP or Roux: which is better to start with?
If you already solve the cube layer by layer, CFOP is the natural next step because it builds on what you know and has more learning material. Roux uses fewer moves and fewer algorithms, but you have to start almost from scratch and think more in blocks.
How long does it take to learn CFOP?
The 16 algorithms of the 2-look version can be learned in days, without straining your memory so you don't forget them. Intuitive F2L makes sense quickly, but making it fluid takes weeks or months of practice, and that is where the difference in times shows.
If you are making the jump to CFOP, a speed cube set up for your hand makes it easier to execute algorithms without lockups. In our Rubik's cubes and speedcubing section you will find magnetic models across different price ranges, and in our guide to the best cubes for speedcubing we compare the most common ones in competition.
Want more guides like this one? Sign up for the newsletter.
Keep exploring
- How to solve the Rubik's Cube step by step (beginner method)
- Rubik's Cube: complete guide
- Best Rubik's cubes for speedcubing
- Types of Rubik's cubes: 3x3, 4x4, megaminx, pyraminx and more
- Which Rubik's Cube to give as a gift
Rubik's Cubes and Speedcubing
YuXin 2x2 Baseball Shape Mod
YuXin 2x2 Basketball Shape Mod
YuXin 2x2 Ball Shape Mod Purple
Skewb Mirror DianSheng (Pink)
QiYi Magic Snake 60 Blocks Multicolour
QiYi Sail W 3x3 reinforced mechanism 56 mm
QiYi Sail 6x6 black classic beginner 60.5 mm
QiYi Sail 2x2 black compact for beginners
Contact our team for any clarification about this document. Contact us