The New Method / Substep / Concept Idea Thread

[FredTheCuber]

Has anyone proposed a version of Winter Variation or Summer Variation that simultaneously permutes the edges yet? I'm currently working on this. The ergonomics side of this is really good and you will only get H perm A perm and E perm after this, with a much greater chance of a complete LL skip. Alg count: 27*6=162. Slightly annoying, but the algs that I have generated are pretty short so far and fast to execute. Is there anything that I should look into?

 

[Cale S]

Has anyone proposed a version of Winter Variation or Summer Variation that simultaneously permutes the edges yet? I'm currently working on this. The ergonomics side of this is really good and you will only get H perm A perm and E perm after this, with a much greater chance of a complete LL skip. Alg count: 27*6=142. Slightly annoying, but the algs that I have generated are pretty short so far and fast to execute. Is there anything that I should look into?

27*6 is 162

And you could do WVCP with the same number of algs, which doesn't give you E perms

 

[FredTheCuber]

27*6 is 162

And you could do WVCP with the same number of algs, which doesn't give you E perms

But recognition for CP is not that great. You only need to look at 3 more stickers to determine an EP case.

 

[Cale S]

But recognition for CP is not that great. You only need to look at 3 more stickers to determine an EP case.

Isn't CP also 3 stickers?

 

[FredTheCuber]

Isn't CP also 3 stickers?

It is, but edges are already oriented so you kinda know where to find the 3 stickers, otherwise ZZ-d would have been much easier. Anyway this might not be a great idea given that amount of algs. Better off EPLS+L4C or full ZZ-b.

 

[Shiv3r]

And if youre a roux or ZZ solver your CLL recognition will be as fast as just recognizing OCLL, and so just recoging the corners is better.
If youre going to do something like this, why not just have 2 sets of WV(54 algs total) that avoids diag swap PLL's? you only need to recog when you have a diag swap and then execute the alg, if you arent going to get diag swap you dont need to care about recognition, and can do a different alg. about 110 less algs, and you don't get Y perm, V perm, E-perm, or N-perms. seems good to me.
EDIT: if you're going to avoid diag swaps, it may be better to use the cases that will solve the diag swap, so you get a U-perm afterwards, but if its not a diag swap then do a different alg(I'm thinking 2-gen mostly).

anyway, here's another example solve of step 5 of Lewis Method.

 

[Neuro]

38 move ZZ-Tri (NLS) speedsolve, had a ZBLL skip
https://alg.cubing.net/?setup=B2_D-... R2_U-_R-_U_R_U_R-_U- U-_R2_D_R-_U_R_D-_R2 U-

Just something off the top of my head: why don't Meyer users do EO pairing and skip part of LSE?

 

[Shiv3r]

EOpairing isn't all that good compared to 3-2-3.

 

[shadowslice e]

Just something off the top of my head: why don't Meyer users do EO pairing and skip part of LSE?

Because Eopairing sucks.

 

[efattah]

Okay, I have received a number of requests for more information on the LMCF method so I posted a huge elaborate document in another thread:

Full LMCF 3x3 method now available

 

[Neuro]

Anyone know of any good programs to generate 4x4 algs? Looking at ways to simplify K4 ELL

 

[Sue Doenim]

In CFOP, would it be useful to orient E-slice edges while building the cross? It would force
<R,U,L> F2L, and isn't too hard to do. This exists for ZZ (ZZ-top) but does not build a cross.

 

[Neuro]

Just realized how easy it is to transfer LMCF into big cubes. Here's the first edition for a 4x4 method:

1: Solve all corners
2: Build preferred LR centers and place
3: Solve E2L using direct solving/commutators
4: Build L4C
5: Pair LSE (similar to Meyer)
6: Solve using L6E and fix parity

May not be good for anything over 4x4 due to reliance on M slices, but this could be good. @efattah what are your thoughts?

 

[Cale S]

Just realized how easy it is to transfer LMCF into big cubes. Here's the first edition for a 4x4 method:

1: Solve all corners
2: Build preferred LR centers and place
3: Solve E2L using direct solving/commutators
4: Build L4C
5: Pair LSE (similar to Meyer)
6: Solve using L6E and fix parity

May not be good for anything over 4x4 due to reliance on M slices, but this could be good. @efattah what are your thoughts?

Doing first two centers before solving corners is objectively better than the other way around

don't know about the rest

 

[Neuro]

37 move ZZ-Tri solve, NLS skip
https://alg.cubing.net/?setup=D-_L2...U_R- U2_y_L-_R_U2_R2_U-_R2_U-_R2_U-_L_U-_R_U-

 

[efattah]

Just realized how easy it is to transfer LMCF into big cubes. Here's the first edition for a 4x4 method:

1: Solve all corners
2: Build preferred LR centers and place
3: Solve E2L using direct solving/commutators
4: Build L4C
5: Pair LSE (similar to Meyer)
6: Solve using L6E and fix parity

May not be good for anything over 4x4 due to reliance on M slices, but this could be good. @efattah what are your thoughts?

At first I agreed with Cale to build LR centers before the corners. Then you lose the 1-look possibility of the corners though. You would need to see almost the entire LR corners solve in the inspection. Either could work.

E2L might actually work really well for 4x4. I had never considered it. This method might be worth exploring. I'm not a 4x4 specialist though. Any 4x4 method that can efficiently direct-solve the pieces instead of first pair them then solve seems to have potential.

In my case solving blue-green on L and R in LMCF I know the colors of the E2L edges backwards and forwards and recognition should be a breeze for someone who is good at 3x3 LMCF. The E2L algorithms would need to be adjusted a bit though. Because there are more edges to solve (12 instead of 6), the chance of E2L triplets also goes up to the point where there would be at least one triplet every solve.

It does seem remarkably suited to 4x4 since the LR centers are never disturbed by any of the E2L algorithms. When solving the last E2L edges it might be possible to begin solving the last four centers at the same time.

 

[Shiv3r]

I already basically use E2L on 4x4, with the Lewis method. the L5E uses the same technique for recognizing as E2L, and is just a bunch of comms. same Idea, I could probably sub-1:30 this with some practice.

 

[Shiv3r]

Hey guys, was looking at squan stuff and stumbled across this. it actually looks really viable, and since I use R&S (started doing R&S then solving D layer than LL, so RBL), this looks really nice.

the steps are basically this:
-cubeshape (or CSP)
-roux F2B
-insert one of the D edges with an "M2"
-insert the last D edge and permute corners at the same time(6 algs)
-EPLL(and parity if you didnt do CSP)(4-5 algs)

this is much less algs than vandenbergh and I think that it can have some serious potential.

 

[Teoidus]

I think if we up the algs closer to the number that vandenbergh has:

CSP
F2B
DB + CLL
L5E (120 algorithms, 60 with parity)

This seems pretty strong.

 

[Shiv3r]

but still with 12 algs you can get fast