Variant Roux "Collums Roux"

[Deleted member 27249]

I think this:
F2L
Solve four Last corners
Oriented eight Last Edges
Solve eight Last Edges

 

[Deleted member 27249]

there will be only 4 edges to always orient.
In the next Better solve Two opposite Edges of cross White.
In the Last step final Roux's step.

 

[artless1der]

I think this:
F2L
Solve four Last corners
Oriented eight Last Edges
Solve eight Last Edges

if you do F2L, arent there only 4 edges left?

Isn't that pretty much the Friedrich Method?

 

[Deleted member 27249]

if you do F2L, arent there only 4 edges left?

Isn't that pretty much the Friedrich Method?

No because i dont create cross.
After F2L i solved Last 4 corners.
I have 4 columns and 8 Edges dont solved.

 

[Deleted member 27249]

if you do F2L, arent there only 4 edges left?

Isn't that pretty much the Friedrich Method?

 

[Deleted member 27249]

if you do F2L, arent there only 4 edges left?

Isn't that pretty much the Friedrich Method?

Solved First 4 corners and belt.
After solve Last 4 corners.
There are 8 Edges dont solve.

 

[EthanMCuber]

Solved First 4 corners and belt.
After solve Last 4 corners.
There are 8 Edges dont solve.

To me that seems pretty inefficient (U, D, M, S at the end of the solve). Are there any benefits for speedsolving?

 

[Deleted member 27249]

To me that seems pretty inefficient (U, D, M, S at the end of the solve). Are there any benefits for speedsolving?

It was an idea that suddenly formed.
I don't know if it is beneficial for speedsolving.

 

[ruffleduck]

Not a new idea

 

[Deleted member 27249]

Not a new idea

Ah ok.

 

[efattah]

LMCF-H also finishes from the columns state but reaches columns in a different way. I think people have underestimated the speed at which the last 8 edges can be solved from columns onwards. The solve is completed with U-M-D which is a much faster set than I originally anticipated; while the TPS of UMD is slower than RUF in CFOP, when finishing the solve with UMD almost every algorithm uses a U & D move at the same time that are done simultaneously. For example LMCF-H uses about 42 moves on average to solve the cube of which the last third are U/M/D, but, if you count U & D simultaneously as one move, then it actually averages around 39 moves. Yes, the TPS is significantly slower than RUF, but if you compare a 39-42 move solve to a 65 move solve, the speed averages out once TPS is taken into account.

 

[nirmod06]

I have this idea for a long time too.

Some slight optimisation that I thought of:
1. End the solve with <M, U, D> gens only, which is reasonably ergonomic;
2. In order to achieve that, use wide move to bring the LR centers on top while solving the last pair, then do a S move while recognizing LL corners;
3. F2L pairs could be solved with a D offset (pseudo pairs) and there is really no point to fix it before L8E.

Also you should read this article, although I think it is rather suboptimal way to do this method: https://www.speedsolving.com/wiki/index.php?title=PCMS

 

[efattah]

LMCF-H solves all the corners and the FL and BL edges (an 'H' or columns on the left half of the cube), all in 1-look. What remains is a single opposite edge pair in the R slice and you have reached full columns in 2 looks. After which one more edge pair is solved, and you finish with L6E, which is a somewhat more expanded L6E set than Roux because it allows for the two remaining unsolved L/R edges to be UR/DR or UL/DL.
Columns has promise, the problem is the speed & number of looks to reach full columns. Once you are there the rest is fast.

 

[nirmod06]

LMCF-H solves all the corners and the FL and BL edges (an 'H' or columns on the left half of the cube), all in 1-look. What remains is a single opposite edge pair in the R slice and you have reached full columns in 2 looks. After which one more edge pair is solved, and you finish with L6E, which is a somewhat more expanded L6E set than Roux because it allows for the two remaining unsolved L/R edges to be UR/DR or UL/DL.
Columns has promise, the problem is the speed & number of looks to reach full columns. Once you are there the rest is fast.

My question would be how feasible is it to one-look 10 pieces. That seems too wild to be possible.

I would suggest an approach closer to PCMS, in that you can take advantage of the D offset and form 3 pseudo pairs in the first step, and track what the last belt edge is. Then you do the last pseudo-pair and finish the columns with CMLL or alike.

I totally agree that MUD 3 gen finish is underestimated. If we count by etm, as you said, this approach could linearly reach sub40.

 

[efattah]

Top 2x2 solvers one look 3-4 solutions within the inspection period. With LMCF-H, you reduce that from 3-4 solutions to just one solution; the two edge pieces can be solve by changing the face-solve to wide move variants, and injecting M/E/S moves that do not affect the corner solution or the corner case. It works out fairly well, although it is challenging-- the big difference is between LMCF-H and WaterRoux, in WaterRoux you one look that one extra piece (BL), but that is the 'straw that breaks the camel's back', and makes it impossible for most people to 1-look. Ignoring that one piece makes all the difference. Being able to consistently solve 10 pieces in the inspection is like seeing an XXXcross in CFOP on every inspection.