# Beginners Roux - why orient corners before permuting them?

#### rcgldr

##### Member
Most beginner's roux examples use Sune (once or twice as needed) to orient the corners (which doesn't permute them), then do J perm (13 turns) twice or Y perm (17 turns) to permute corners. If corners are permuted first, a 9 turn sequence where only the 5th turn varies can be used to swap front 2 corners, back 2 corners or front+back 2 corners (diagonal swap with U or U' done first). This order makes sense if doing layers, but with Roux, the last 6 edges are handled last, so for beginner Roux, it seems that permuting corners first is simpler. To simplify looking for corner permutation, I look for adjacent orange or green corners put them at the front top. If orange on front top, green should be on left, if green on front top, red should be on left.

swap front - L' U' L F U F' L' U L
swap rear - L' U' L F U' F' L' U L
swap front+rear - L' U' L F U2 F' L' U L
diagonal - U then swap front+rear

Sune sequence doesn't permute corners, although you often have to do it twice. I do Sune in reverse order so that one hand does L or R turns, the other U turns:

L U2 L' U L U L
R U2 R' U' R U' R'

For reverse order Sune, the starting position needs to be adjusted. View from the top, start at the +. If only 2 yellow corners facing up, start with yellow facing out on back side. If 0 corners facing up, start with side with single corner facing out (or any headlight if pair of headlights).

Code:
      ---         ---            ---         ---           ---        ---
* |   |       |   | *      * |   | *   * |   | *     * |   |      |   | *
* |   |       |   | *        |   |       |   |         |   |      |   |
---         ---            ---         ---           ---        ---
+             +            +             +           + *        * +

*             *            * *         * *
---         ---            ---         ---
|   | *   * |   |          |   |       |   |
|   | *   * |   |          |   |       |   |
---         ---            ---         ---
*             *            * *         * *
+             +            +             +`

Last edited:

#### CarterK

##### Member
Recognition.

Alternatively, learn full CMLL

Edit: btw Sune does preserve corners, do a U2 after

#### rcgldr

##### Member
Recognition
Ok, so if corners are oriented first, it's easier to spot them, but you need to memorize a longer turn sequence, (J-perm (maybe twice), or learn J-perm and Y-Perm).

btw Sune does preserve corners, do a U2 after
I had the impression that some of the non-Sune corner orientation algorithms shown in some beginner Roux methods do permute corners (in which case you would need to orient first).

Alternatively, learn full CMLL
That's 24 algorithms plus the mirrored versions of them? Currently that's more than i want to learn. I'm an old guy that bought a cube in 1980 and made my own algorithms, doing top, middle, bottom. I used corner swaps twice (second one mirrored) to flip bottom edges, and Sune twice (second one mirrored) to rotate 3 bottom edges. I switched to top, bottom, middle, after finding middle edge algorithms while working on patterns. Now I'm using beginner's Roux. I'm not fast, as I take around 1 1/2 minutes to 2 minutes to solve, but I'm not using any look ahead.

Last edited:

#### CarterK

##### Member
Ok, so if corners are oriented first, it's easier to spot them, but you need to memorize a longer turn sequence, (J-perm (maybe twice), or learn J-perm and Y-Perm).

I had the impression that some of the non-Sune corner orientation algorithms shown in some beginner Roux methods do permute corners (in which case you would need to orient first).

That's 24 algorithms plus the mirrored versions of them? Currently that's more than i want to learn. I'm an old guy that bought a cube in 1980 and made my own algorithms, doing top, middle, bottom. I used corner swaps twice (second one mirrored) to flip bottom edges, and Sune twice (second one mirrored) to rotate 3 bottom edges. I switched to top, bottom, middle, after finding middle edge algorithms while working on patterns. Now I'm using beginner's Roux. I'm not fast, as I take around 1 1/2 minutes to 2 minutes to solve, but I'm not using any look ahead.
Longer turn sequence yeah, but it’s not a very hard one and can be executed very fast.

Oh yeah I see what you’re saying here. This just gives more of a reason to do Orientation first (better algs)

And yeah it’s 40 algs. Very small compared to other alg sets, and most are pretty easy

No point in optimizing 2-look cmll

#### xyzzy

##### Member
I used to do two-look CLL (on 2×2×2 and on big cubes) with permutation first, then orientation, just because that's how I originally learnt it. For what it's worth, I don't think this is a good method, for the already-mentioned reason of recognition difficulty.

With permutation-orientation order, for an adjacent corner swap, you can use Niklas (L' U R U' L U R'; swaps the right corners) and for diagonal corner swap, you can use F R U R' U' F'. These are both short and fast algs, but this comes at the cost of making the orientation step worse (the U, T and L cases, specifically; you're switching out 6-8 moves for those cases with 10-15 moves).

With the more common orientation-permutation order, note that there are significantly shorter alternatives to the standard J perm (13 moves) and Y perm (17 moves) algs used in CFOP, if learning long algs is one of your concerns. For example, there's R U' L U2 R' U R U2 R2 x (9 moves) for adjacent swap and R U' R' U' F2 u' r U r' u F2 (11 moves) for diagonal swap. (These algs aren't faster than the standard J/Y perm algs. They're just shorter. I don't recommend learning these.) It's arguable whether the algs here are better (if there's a move count difference between orient-permute and permute-orient, it's pretty small), but recognition-wise this is so much better that you shouldn't use permute-orient.

#### WoowyBaby

##### Member
For a beginner method, the number of moves doesn't matter as much as other things. Sure, long algorithms can be hard to memorize, but for a beginner method I teach, I use R' D' R D to orient, and I use the "Star Wars" algorithm for permuting corners which is only 9 moves. R2 D2 (R2-D2 is the cute droid character lol) R U R' D2 R U' R (A-perm). If you have a diag swap you apply it twice.

Understand-ability and recognition and simplicity is what makes a beginner method good.

#### rcgldr

##### Member
No point in optimizing 2-look cmll
Especially in my case, as I spend about 2/3rds of my solve time doing first two blocks, 1/3rd on CMLL and last six edges. I'm not doing any look ahead for first two blocks, and that is where I could make the most improvement.

One point in favor of permuting corners first is that it's the same 9 turn sequence for all cases, except for the 5th turn (U, U', U2), and only involves two F turns. This means corner orientation is restricted to non permuting algorithms. I'm only using reversed Sune, but that only involves L+U or R+U turns, which I can do a bit faster than my other turns. The non permuting algorithms that avoid doing Sune twice are 11 turn patterns, so they only save 3 or 4 turns. Since I learned permuting first and got used to it, I can't judge how difficult recognition would be for beginners

Last edited:

##### Member
I also use a 2-Look step for corners during Roux.

When I was first learning Roux, I compared Orient-First vs. Permute First.
I was finding that the combination of algs I was using were about 1 move fewer for P-first over O-first.

F (R U R' U') F’ swap diagonal corner
(R U R' U') (R' F R F') swap front corners
...then after 1 or 2 Sunes was usually about 20 moves total

vs

Orient-First: 8 to orient, then 13 to permute with these (14 and 10 htm):
R U’ L U2 R’ U R... L’ U’ L U2 R’ U L’ swap diagonal corner
R U’ L U2 R’ U R... U2 R’ L’ swap front corners

or a little shorter still... (which then makes O>P vs P>O basically the same move counts)
Diag: R’ (U r' F2 R F')2 R
Adj: R2' F2 (R U r' F2 R F' R)

But I'm faster recognizing the Perm cases after they're oriented, so 1 move didn't help me.