The New Method / Substep / Concept Idea Thread

[Neuro]

Could be a nice method, only problem is the EO portion would be hard to recognize in inspection given so little time after planning Roux block.

As for Waterman/Roux, I would probably be more inclined to use PCMS, LMCF, or some other corners/columns first variant given the difficulty of the last few steps and the awkward first steps of Waterman. May add it/some variation into the doc because it does sound pretty nice. Right now I'm looking at how to improve inspection for what's current, but I am more than willing to add combinations.

 

[efattah]

Interesting document. I have long been a proponent of LMCF+[CFOP-ZBLL] as the ideal method neutral system. Inspection is extremely simple; rapidly rotate the cube; if you see three corners making a face (or even a full face), use LMCF, otherwise use CFOP. With three corners making a face, it is easy to one look the EG solve which gives a slick LMCF time. Best of all you can identify the presence of three corners making (almost) a face in probably 1-2 seconds. Giving you 13 seconds to inspect for CFOP if you choose that route.

 

[Neuro]

Interesting document. I have long been a proponent of LMCF+[CFOP-ZBLL] as the ideal method neutral system. Inspection is extremely simple; rapidly rotate the cube; if you see three corners making a face (or even a full face), use LMCF, otherwise use CFOP. With three corners making a face, it is easy to one look the EG solve which gives a slick LMCF time. Best of all you can identify the presence of three corners making (almost) a face in probably 1-2 seconds. Giving you 13 seconds to inspect for CFOP if you choose that route.

Looks nice, I'll be adding LMCF into the doc and make the variations as seen viable. May add other methods (Petrus. Even though I don't like it, other people do so I won't cut them out).

Others may include PCMS, SSC, MCELL, and ECE; but probably won't add, because of lack of lookahead/developmemt that the other methods have.

 

[Shiv3r]

Hey guys, why does no one use VH method? Because Im a rouxer, and I know COLL(mostly by conjugating CMLL's with M moves and such), If I dont use OBLBL(really only for 5x5), then I will solve 5x5+ with VH method. the odd-numbered cubes are easier because you don't get parity, but it seems to be pretty nice. I mean, I know 4 VHLS cases algorithmically and I can solve the other ones bu using the others inserts and stuff.
so if you don't know, VH method is a simplified ZB method(thats why ZZ with COLL and EPLL is sometimes called ZZ-VH). here are the steps:
-cross
-F2L-1
-insert the last pair(can be either pair or split pair), and orient edges
-COLL
-EPLL

I have an example solve here, which took as many moves as normal CFOP would.

now remember, I don't like using CFOP-based methods, but on 5x5+ where I solve with like either Hoya5 or just redux, the fact that not very much inspection is needed to start 3x3 stage is pretty beneficial. so I will do Vandenbergh-harris, because then I can just go into a COLL(i do know full PLL but not full OLL, and my corner orientation recognition is just as fast as my CxLL recognition), and then an EPLL, of which all the cases are pretty nice and 2-gen(except for H-perm, but its so low movecount its ok-- I use the 2-gen Z perm anyway).

 

[xyzzy]

Hey guys, why does no one use VH method?

But… people do? Some VHLS cases are bad and there's no point using them instead of just OLL, but I think most people know at least the easy cases.

I use it a bit more often on 4×4×4 than on 3×3×3 because some of the PLL algs I use are terrible on big cubes (and I haven't learnt alternatives), so forcing EPLL by doing VHLS and COLL is pretty useful. Forcing 1-flip cases when parity happens is also useful.

 

[Shiv3r]

But… people do? Some VHLS cases are bad and there's no point using them instead of just OLL, but I think most people know at least the easy cases.

I use it a bit more often on 4×4×4 than on 3×3×3 because some of the PLL algs I use are terrible on big cubes (and I haven't learnt alternatives), so forcing EPLL by doing VHLS and COLL is pretty useful. Forcing 1-flip cases when parity happens is also useful.

I don't get parity because I use the Lewis method on 4x4, and I only do up to 5x5(OBLBL or Hoya5) rn. but the VHLS cases are all nice, and I can do most of them intuitively while inserting the last slot so I can do that, and it skips what would otherwise be a 3rd look for LL on 5x5.

 

[mDiPalma]

tbh I think petrus/zz neutrality can be very powerful

You are basically picking the easiest path to familiar territory. EO and blocks are about as independent as you can get.

It's no coincidence that many of the fast Petrus users are also fast ZZ users (and vice versa).

 

[Neuro]

I'm about to start work on the second version of my method neutrality doc. Any last minute suggestions?

Right now I have: Petrus, LMCF, maybe some high potential experimental methods (MCELL, SSC, ECE, PCMS, etc), and of course recog systems/viable combos for all

 

[xyzzy]

it skips what would otherwise be a 3rd look for LL on 5x5.

Or you could just use Petrus.

This is actually what I do on 5×5×5 since OBLBL means you already have one gigantic block sitting right there.

 

[Shiv3r]

Or you could just use Petrus.

This is actually what I do on 5×5×5 since OBLBL means you already have one gigantic block sitting right there.

I do OBLBL, and I actually do use Petrus with it.
Now I just need to find a way to combine AvG with EOpairing. Because my solves with OBLBL on 4x4 with EOpairing(the pairing used in Z4) have gotten me sub-1:15 solves(for comparison, I average 58-1:08ish with Lewis method on 4x4). This is because the hardest step in OBLBL is doing EO right after edgepairing, and EOpairing, even if slightly less efficient, skips that step, and there are few enough edges that it doesn't make that big of a difference.

 

[Neuro]

Here's the second rendition of the method neutrality doc. Do you think I should add subset neutrality in the doc or strictly limit to methods themselves? It'd basically be V2.5

**LARGE EDIT- Deleted tables, felt they were unnecessary and provided little in the way of understanding the concept

https://docs.google.com/document/d/1VtmdlFansxiboZMgIGGNKXppKDLsgiuVGlRz3e6z0fA/edit?usp=sharing

 

[Shiv3r]

Here's the second rendition of the method neutrality doc. Do you think I should add subset neutrality in the doc or strictly limit to methods themselves? It'd basically be V2.5

**LARGE EDIT- Deleted tables, felt they were unnecessary and provided little in the way of understanding the concept

https://docs.google.com/document/d/1VtmdlFansxiboZMgIGGNKXppKDLsgiuVGlRz3e6z0fA/edit?usp=sharing

A lot of the fastest petrus solvers use ZZ, I think that in fact they are Similar. so I disagree with petrus/zz being somewhat similar, they are similar. Petrus+roux is doo difficult to achieve I think though. Ill tlel you if theres anything else I disagree on.

 

[Neuro]

A lot of the fastest petrus solvers use ZZ, I think that in fact they are Similar. so I disagree with petrus/zz being somewhat similar, they are similar. Petrus+roux is doo difficult to achieve I think though. Ill tlel you if theres anything else I disagree on.

Ok I'll make sure to change some of the method comparisons. That was actually one I was on the fence about. I'll be thoroughly testing all the combinations and may delete some, but I think that they're all viable if even for just a small group.

Any thoughts on adding substep neutrality?

 

[Shiv3r]

Ok I'll make sure to change some of the method comparisons. That was actually one I was on the fence about. I'll be thoroughly testing all the combinations and may delete some, but I think that they're all viable if even for just a small group.

Any thoughts on adding substep neutrality?

Unless it's ZZ-CT.
I think that for ZZ OHers, I think that fast singles are possible with a method like ZZ-rainbow if you can see if the rainbow will be mostly solved after EOline, but that may be just a neutrality thing. like Mats valk is substep neutral between VLS-PLL and OLL-PLL. substep neutrality is nice if you know only nice cases for each one.

 

[Neuro]

Well if I were to add it it'd look something like this:

Roux: All LSE

• Pinkie Pie: Easy inserts or 1/both UF/UB or UR/UL oriented on bottom
• EO LR/FB- Good case after CMLL
• UFUB- Good case after CMLL
• L5E- 1 top edge presolved or very fast insert after CMLL

ZZ: LS+LL

• ZBLL/OCLL+PLL/COLL+EPLL- LS presolved or 3/4 move insert
• L5EP- If LS corner presolved
• ZZ-CT- No LS, hard insert

CFOP:LS+LL

• OLS/OLL+PLL/VH- Edges disoriented
• Any ZZ variant-Edges oriented, follow same ZZ criterion

Also, what about Opp cross ELL+EPLL for Roux? With all edges oriented of course. Might be helpful if their are no good options to solve UF/UB or UR/UL. Would be used in conjunction with OLLCP

So Pinkie Pie with DB/DF on bottom (and centers solved, but I may make y2 reflection algs), and direct solving from there

*EDIT- Could be used with OLLCP, but would also work after CMLL/Roux EO if easy.

 

[Shiv3r]

Stuff about a method Neutrality Document

I was thinking. Since both LMCF and Roux use a lot of M-slices and wierd fingertricks and both can get low movecounts, and you should already know C(M)LL if you do Roux, I think that LMCF/Roux neutrality could be powerful.
how it would work:
scan around the cube quickly. If you can find a face you can do easily, then work on onelooking that, and then do LMCF. If you do not find one within the first few seconds of inspection, then plan a Roux first block and continue with Roux.
I think that it would mean you would normally do Roux more often than LMCF, but since many fast Rouxers know at least EG-1 along with C(m)LL, They can easily do it. And besides that, there are not very many algs in LMCF. And, since the movecount is low, even if your TPS is slower than your Roux TPS, then you can get times probably roughly as fast as your roux times.
Just an Idea. Also I am considering learning LMCF and doing that, maybe you should add that Neuro?

 

[Neuro]

It's in the newest version along with other LMCF combos, and that's basically the same as how I made the recognition system. Any other methods that should be in the doc? Currently I have CFOP, Petrus, Roux, ZZ, and LMCF along with neutrality systems for all combos

 

[Shiv3r]

Update on the Lewis method:

 

[ShyGuy1265]

Yesterday I made a face first 3x3 method and I posted it on r/Cubers. It got a surprisingly good response considering how much they hate seeing new method posts, so now i'm posting it on here.

Step 1: Make a face with one corner missing

This is pretty simple. Just blockbuild a face on the bottom. This is just like making a layer but you do not need to permute the pieces.

Step 2: Solve 3 middle layer pieces using keyhole

Use the slot without a corner to insert middle layer pieces one by one. If you see an unsolved middle layer piece, put the spot where it goes above the empty corner slot and insert it. Repeat until 3 middle layer edges are solved.

Step 3: Permute 3 corners on the first layer

Find the corner that is not in the first layer and insert it in the spot where it belong relative to the other corners. Make sure you insert using the empty edge slot so that you don't mess up what you have already done. When you insert the corner, don't worry about taking another one out. That is exactly what you are supposed to do. Just repeat with the corner you take out. If all of the corners are inserted but you don't have 3 permuted, take an unsolved corner out and put it in the right spot. When you finish this step, you should have 3 solved corners on the first layer and the 4th permuted corner in the last layer. If you have 4 solved or permuted corners on the first layer, that is okay too.

Step 4: Solve the final first layer corner and final middle layer edge

For this step, just use non-matching F2L. If you know F2L, the only challenge to this is making sure the edge is oriented. You can use last slot tricks if you want, but they are not required.

Step 5: Solve the last layer and permute the first layer edges

I know of 2 2-look version under 400 algs: CLL and ELL+EPFL, and OLLCP and EPBL. I see the most potential in the second one. If you are afraid of algs, then you can intuitively permute the first layer edges with M, U, and D moves and solve the last layer. Unfortunately, I don't have algs for both of the 2-look versions.

 

[Neuro]

Other steps? What I see now is a blockbuilder's approach to CFOP with keyhole F2L