COLL Algorithms: All 40 COLL Cases for CFOP and ZZ

What is COLL?

COLL stands for Corners of the Last Layer. According to the SpeedSolving Wiki, COLL was proposed by Lars Petrus. This step solves the last-layer corners while keeping the edges facing the same way, so after COLL only an edge swap is left, called EPLL (SpeedSolving Wiki).

Compared with the usual CFOP way:

• Normal CFOP: OLL makes the yellow face, then PLL swaps both corners and edges.
• With COLL: COLL makes the yellow face and puts the corners in place at the same time, then EPLL only swaps the edges.

EPLL has just four cases: Ua, Ub, H and Z. SpeedCubeDB notes that after COLL you get one of those four cases, or with a 1/12 chance nothing more to do at all (SpeedCubeDB). These four algorithms are in the 2-look PLL post and the U perm algorithm post.

How many COLL cases are there?

The SpeedSolving Wiki counts 42 cases if you include mirror cases, and 24 if you don't. On average a case takes 9.78 moves when solved optimally (SpeedSolving Wiki).

Two of those 42 cases already have the yellow face done and only need the corners swapped. That is really a PLL job, so SpeedCubeDB lists only 40 cases, split into 7 sets (SpeedCubeDB). This post follows that split:

The set is simply the yellow face shape you already know from 2-look OLL. Inside each set, the cases share the same yellow sticker pattern but differ in how the corners are swapped, so you also need to look at the colours on the sides.

When is COLL really worth it?

COLL only works when the yellow cross is already there right after F2L. In CFOP that doesn't happen very often: the four top-layer edges have 8 equally likely flip patterns, and in only 1 of the 8 are all four already yellow on top. That means on average just 1 solve in 8.

That is why the SpeedSolving Wiki says COLL is mainly used as the last layer for the ZZ method, where the edges are always facing the right way before the last layer. In CFOP, COLL is an extra for the times the yellow cross shows up by itself (SpeedSolving Wiki).

Cuboss, the site that collects Daniel Wallin's algorithm sets, gives two pieces of advice (Cuboss):

• Take on COLL after you know full OLL and PLL, although you can start right after 2-look OLL and 2-look PLL.
• You can skip the Sune and Antisune sets, because Sune and Antisune are very fast, so COLL then EPLL is usually not faster than OLL then PLL for these two sets.

In short:

• You use ZZ: COLL is worth it, because you use it in every solve.
• You use CFOP and aren't under 20 seconds yet: you don't need it yet. Get full OLL and full PLL down first.
• You use CFOP and already know OLL and PLL: add the H, Pi, U, T and L sets bit by bit. These are the sets where OLL tends to be slower.

How to hold the cube and read the pictures

• Yellow face on top, F2L already done. Green face at the front, orange on the right.
• Each picture has two cubes: on the left is the case to solve, on the right is the result after the algorithm. Only the top layer is coloured; the two bottom layers are grey.
• The pictures only show the top, front and right faces. Turn the top face (U, U' or U2) until the yellow stickers and the colours on the front and right match the picture exactly, then do the algorithm.
• In the pictures, the top-layer edges are already placed so the cube ends up fully solved. On a real cube, there is usually still an EPLL step after COLL.

Notation: Rw means turning the two right layers together (many sites write it as a lowercase r). D is the bottom face and L is the left face. Algorithms with D pull F2L apart and then put it back, so don't worry when the two bottom layers look scrambled halfway through. New to notation? See Rubik's Cube notation for beginners.

The algorithms in this post come from the SpeedCubeDB list, usually the most voted one. If the original algorithm starts with a y (turning the whole cube), this post leaves that move out and draws the picture to match how you hold the cube, so you don't need to turn the whole cube.

H set: no yellow corners, two pairs of yellow stickers facing each other

Four cases. H4 is the H algorithm from 2-look OLL, now used for the one case where it makes the yellow face and puts the corners in place too.

H1: R U R' U R U' R' U R U2 R'

H2: F R U' R' U R U2 R' U' R U R' U' F'

H3: R U R' U R U L' U R' U' L

H4: F R U R' U' R U R' U' R U R' U' F'

Pi set: no yellow corners, one pair of yellow stickers

Six cases. Pi1 is the same Pi algorithm you already know from 2-look OLL.

Pi1: R U2 R2 U' R2 U' R2 U2 R

Pi2: R' F2 R U2 R U2 R' F2 U' R U' R'

Pi3: R' U' F' R U R' U' R' F R2 U2 R' U2 R

Pi4: R U R' U' R' F R2 U R' U' R U R' U' F'

Pi5: R U' L' U R' U L U L' U L

Pi6: Rw U R' U R' F R F' R U' R' U R U2 Rw'

For Pi2 and Pi6, this post picks the second algorithm in the SpeedCubeDB list: for Pi2 because it is shorter, and for Pi6 because the first one uses an S move (turning the middle slice that sits parallel to the front face), which is hard for beginners.

U set: two yellow corners next to each other, two yellow stickers on the same side

Six cases. U3 is the U (headlights) algorithm from 2-look OLL.

U1: R' U' R U' R' U2 R2 U R' U R U2 R'

U2: R' F R U' R' U' R U R' F' R U R' U' R' F R F' R

U3: R2 D R' U2 R D' R' U2 R'

U4: F R U' R' U R U R' U R U' R' F'

U5: R2 D' R U2 R' D R U2 R

U6: R2 D' R U R' D R U R U' R' U' R

U3 and U5 are a pair: take U3, flip the direction of every R and D move (R becomes R', D becomes D' and the other way round), and you get U5. Pick up these two on the same day.

T set: two yellow corners next to each other, two yellow stickers pointing opposite ways

Six cases. T5 is the T algorithm from 2-look OLL.

T1: R U2 R' U' R U' R2 U2 R U R' U R

T2: R' U R U2 R' L' U R U' L

T3: R' F' Rw U R U' Rw' F

T4: F R U R' U' R U' R' U' R U R' F'

T5: Rw U R' U' Rw' F R F'

T6: R' U R2 D Rw' U2 Rw D' R2 U' R

L set: two yellow corners diagonal from each other

Six cases. L5 is the L (bowtie) algorithm from 2-look OLL.

L1: R U R' U R U' R' U R U' R' U R U2 R'