Roux, seen from CFOP

Roux is not a variation on CFOP. It solves a different set of pieces in a different order and finishes with a step CFOP has no equivalent of. Gilles Roux published it in 2003, and it is the only serious rival to CFOP at the top of competition results.

The four steps

1. First block. Build a 1×2×3 block on the left of the bottom layer: the down-left edge, the two corners either side of it, the two edges above them, and the left centre. No algorithms — it is worked out, the way you work out a cross, but there is more of it.
2. Second block. Build the same thing on the right. It has to leave the first block standing, which restricts you to turns of the right-hand face, the top and the middle slice. The two blocks together are the first two layers, minus the middle slice.
3. CMLL. Solve all four top corners in a single algorithm — orientation and position at once — while ignoring the edges completely. Forty-two cases.
   RUR^'URU²R^'
4. LSE — the last six edges. Everything unsolved is now in the middle slice and the top layer: six edges and two centres. Orient them, place the two side edges, and finish with half turns of the slice.
   M^'U²M^'U²

What a CFOP solver notices first

• There is no cross. There is no step whose only purpose is to make the next step possible.
• Nothing is a layer. You never protect a finished layer, because until the very end nothing is a finished layer.
• The corners come before the edges in the last part, which is the opposite way round from OLL and PLL.
• The solve barely rotates. Roux is mostly R, U and M with a single y between blocks — and the M slice is turned with the left hand rather than gripped.
• Fewer algorithms. Forty-two for CMLL, a handful of patterns for the last six edges, and nothing else. That is the entire method.

CMLL

CMLL is the same forty-two corner cases as COLL. The difference is the rule an algorithm has to obey. A COLL algorithm must leave the last-layer edges oriented, because CFOP still needs them that way. A CMLL algorithm only has to leave the two blocks standing, and the six loose edges may be thrown anywhere, because throwing them anywhere is what the next step expects.

The last six edges

This is the step with no CFOP counterpart, and the one people find alien. Six edges remain: the four in the middle slice and the two on the sides of the top layer. It goes in three passes, and only the third involves anything you would call an algorithm.

1. Orient them. Using only slice turns and turns of the top, flip every one of the six the right way up. Recognition is by counting which edges show the top colour, and it is worked out rather than memorised.
2. Place the two side edges. Bring the up-left and up-right edges home, again with slice and top turns only. Six or seven turns typically does it.
3. Finish the slice. Four edges are left in the middle slice, in one of the twelve arrangements they can reach. A handful of short patterns of slice and top half turns covers every one.
   M²U²M²U²

Move counts, honestly

Roux is usually quoted at forty-five to fifty moves against CFOP's fifty-five to sixty, and the gap is real but smaller than it looks. Roux counts are given in slice-turn metric, where an M counts as one move. Count the slices as two, which is what the half-turn metric does, and most of the advantage disappears.

Should you switch

Probably not, and certainly not as a remedy. Switching methods costs months, your times go backwards for several weeks, and the fastest Roux solvers sit a shade behind the fastest CFOP solvers and comfortably ahead of nearly everybody else — which tells you plainly that the method is not what is keeping you at twenty seconds.

There is a better reason to learn it: block building is a different kind of thinking from slot filling, and it makes you better at seeing the cube whatever you solve with. Plenty of people keep CFOP for speed and Roux for pleasure, and the pleasure improves the speed by a route nobody expected.