The New Method / Substep / Concept Idea Thread

[Teoidus]

Based on watching others have to explain 2GR on Discord, I've made significant simplifications to the CPLine page of the tutorial: http://teoidus.github.io/dump/2GR/CPLine/CPLine.html

I hope this version is more understandable! Let me know what you find confusing--CP is (despite not being that hard once you get it) annoyingly hard to explain, and I'd love to get feedback from you all.

 

[Neuro]

Hmm, they're very different methods, which could make neutrality difficult: if the difference is so small, the times saved changing methods may be less than the time taken deciding to change methods!

Also my method doesn't use TCMLL so I don't know where you got that from

Yeah I know it isn't TCMLL, I just call it that bc I can't think of a better/proper name

Interesting point, could be true. The only other thing I can think of would be OLS+Pinkie Pie but that would be a pain to set up in solves. @Teoidus that's great news! At the moment I'm debating between Roux and LLOB+ZBLL, but CP Roux may just be a contender now

 

[Arc]

Yeah I know it isn't TCMLL, I just call it that bc I can't think of a better/proper name

Interesting point, could be true. The only other thing I can think of would be OLS+Pinkie Pie but that would be a pain to set up in solves. @Teoidus that's great news! At the moment I'm debating between Roux and LLOB+ZBLL, but CP Roux may just be a contender now

As far as CP Roux goes, myself and Shadowslice have both tried to find a way to abuse CPFB, but nothing notably good so far. Not to mention that idk if anyone can actually do an efficient CPFB in a speed solve yet.

 

[Hazel]

I've heard once that the only major downside of ZZ-CT is that you have to use ZZ. So I've come up with a Roux variant (a CFOP variant has already been made, Ribbon, but it has many many algs).
Step 1: Standard Roux FB
Step 2: 2x2x1 block in BRD
Step 3: TSLE
Step 4: EO+D layer (14 super easy M/U algs)
Step 5: TTLL

Example: L' R2 B U2 F U2 B2 F' R2 U2 F D R' D U L' D2 U2 B'
z2 L F' D' L2 F D' B' // FB (7/7)
U2 R2 U' R U' R // SB (6/13)
R U R' U' R U R' U' R U2 R' // TSLE (11/24)
U' M U2 M2 U M U M' U M U2 M' U2 M // EO+D (14/38 - pretty bad case I think)
R U' R U' R' U R' U R2 D' R U R' U' D R U2 // TTLL (17/55)

Not the best solve, but could anybody figure out what the average movecount would be?

 

[TDM]

I've heard once that the only major downside of ZZ-CT is that you have to use ZZ.

Not at all. The worst part of ZZ-CT is the CT part: see Matt's graph here.

 

[Arc]

I've heard once that the only major downside of ZZ-CT is that you have to use ZZ. So I've come up with a Roux variant (a CFOP variant has already been made, Ribbon, but it has many many algs).
Step 1: Standard Roux FB
Step 2: 2x2x1 block in BRD
Step 3: TSLE
Step 4: EO+D layer (14 super easy M/U algs)
Step 5: TTLL

Example: L' R2 B U2 F U2 B2 F' R2 U2 F D R' D U L' D2 U2 B'
z2 L F' D' L2 F D' B' // FB (7/7)
U2 R2 U' R U' R // SB (6/13)
R U R' U' R U R' U' R U2 R' // TSLE (11/24)
U' M U2 M2 U M U M' U M U2 M' U2 M // EO+D (14/38 - pretty bad case I think)
R U' R U' R' U R' U R2 D' R U R' U' D R U2 // TTLL (17/55)

Not the best solve, but could anybody figure out what the average movecount would be?

The major downside of ZZ-CT is that it has high move count and high alg count compared to similar LSLL approaches. Roux-CT was suggested twice in the past few pages. Though I would like to see your 14 super easy <M,U> algs for EODFDB.

Recently on vacation, I thought of a pretty cool LLOB-ZZ hybrid that went like this (I call it "cantelope" because I was eating a cantelope when I thought of it).

1. FB (1×2×3 block on left, ~7 moves)
2. SB (1×2×2 on back right, ~9 moves)
2.9. Use an M/ M' to orient the FR edge (~0.5 moves)
3. Solve the FR edge into its slot while orienting the corners (this is just TSLE in ZZ-CT but you don't need to preserve EO, so probably ~7 moves)
4. EOFB (Orient the edges while solving the DB and DF edge, same step as most LLOB methods, ~8 moves)
5. TTLL (~12 moves)

This method seems pretty decent, it could very good for someone switching from ZZ to Roux (or any casual solver) and probably averages in the ballpark of 44-46 moves (with AUFs). Anybody think this is worth genning algs for?

 

[Hazel]

The major downside of ZZ-CT is that it has high move count and high alg count compared to similar LSLL approaches. Roux-CT was suggested twice in the past few pages. Though I would like to see your 14 super easy <M,U> algs for EODFDB.

The way I do it is solve DB while correcting centers, then insert DF with EODF: https://docs.google.com/spreadsheets/d/1cCDVpuPzUXq_c2lItT3dBg0r2HscVXtT8ptqkD19c6I/edit?usp=sharing

 

[Neuro]

EO+LE can be a good approach but it can be extremely inefficient. What may be good is to generate a few sets of algs for-say-when the edge is swapped or have opposite centers (like dots LSE)

Doing EODFDB intuitively is pretty easy for me though so I don't see a whole lot of use for it other than a few cases here and there. May make seeing your corner case easier if you use algs though.

 

[Sion]

Since I'm twisting my main method around, I decided I'd go ahead and post up what I'm trying to do.

1: PCMS style CE pairing. PCMS has a wonderful system for finding small movrecount proper columns and pairing them effectively. I'm keeping that.
2: orient 'LL corners. This could end in an a case, an e case, or a solved set of corners. This is to improve recognition because I'm not looking for too much. Because there are technically only three last layer corner cases, recognition will be faster for the final step.
3: Down edges and L5eop. This is very fast and quickly done, not much to be said.
4: pll. While this would seem less efficient, from a recognition point of view, it is easier since you familiarize yourself with the corners during the prior step.

 

[Thermex]

@Sion I believe this idea is mentioned in the original columns/PCMS wiki page, check that out.

 

[Sion]

@Sion I believe this idea is mentioned in the original columns/PCMS wiki page, check that out.

It's similar, but this has l5eop instead

 

[Thermex]

It's similar, but this has l5eop instead

Whether it's L5EP, L5EOP, L5E or whatever for the edges step, Columns first really isn't worth it. The ergonomics are bad and in general PCMS variants are pretty inefficient. For newer method designers looking to make variants of things I would recommend LMCF, CFOP LL+LS and even Petrus/HF variants much more than columns first and Roux variants.


Lately I've been experimenting with Megaminx LS+LL subsets and I've ran across a couple of interesting ideas, I'll throw them out here for you guys to judge:

Idea 1 (megiminx MGLS):
This could be good, I see no major flaws in it. It's a fairly efficient 3-look LL+LS idea without a rediculous number of algs. The main problem is that Minx PLL kinda sucks :/

(In cases you've never heard of MGLS, it goes as follows):

1. FR edge+EO
2. DFR corner+CO
3. PLL

Idea 2: Phasing?
Maybe phasing could be utilized on megaminx? I have no idea how it would be used or how it would work since there are 5 edges on a Minx, but who knows, phasing could be useful on Mega.

Idea 3:
This is a cool concept I came up with today, basically you:

1. Orient edges
2. Permute edges
3. Oreint corners
4. Permute corners

Because of how well EO+EP flow this is essentially a 3-look last layer with ~100 algs, seems great but I haven't really done much experimenting with this yet.

Idea 4:
Probably worse than the last idea, but still pretty decent with even less algs (maybe):

1. 2×2 block (Oriented and Permuted pieces)
2. Expand the block to a 2×2×3 (The pieces you add on don't need to be permuted)
3. Orient the 7 remaining pieces while inserting the FR edge (this can't be too many algorithms)
4. TTLL (greatly reduced alg set since there's a 2×2 solve in the back)

Opinions?

 

[Neuro]

What I've long thought would be good but suspected it'd been done before is doing EOLS. And then you get easy 3LLL; CO, CP, EP.

Method design takes time and dedication, not to mention a little bit of math skills & being able to discern what's good and what's not. I agree start with basic things/ideas and learn some basic skills (cube explorer and harcs are essential IMO) and have a good understanding of several pre-existing variants/methods.

Officially in university now, so probably won't be posting very often or working on my projects. If/when I have time to do that stuff I'll update. But in the meantime, good luck to everyone on here!

 

[Thermex]

What I've long thought would be good but suspected it'd been done before is doing EOLS. And then you get easy 3LLL; CO, CP, EP.

Method design takes time and dedication, not to mention a little bit of math skills & being able to discern what's good and what's not. I agree start with basic things/ideas and learn some basic skills (cube explorer and harcs are essential IMO) and have a good understanding of several pre-existing variants/methods.

Officially in university now, so probably won't be posting very often or working on my projects. If/when I have time to do that stuff I'll update. But in the meantime, good luck to everyone on here!

EOLS+3LLL is pretty good for only like 50 algs but not really what I'm shooting for. I've been looking for a pure 3 look LL+LS that's more efficient than the current best approach (pair, OLL, PLL) and preferably 30 moves. I'll try to gen some of the algs for these methods myself, but does anyone know of any other megaminx solving programs other than ksolve? That would be helpful. @Neuro GL at university

 

[Arc]

Am I missing something here? EOLS is just LS + EO right? Like ZBLS? That's 302 algs on 3x3, and orders more on mega. If what you're doing is making a pair first and then EO while inserting it, that's VHLS and would be 5look LSLL. If you want something more efficient than OLL PLL, perhaps you could look into making fish and chips on mega.

 

[Thermex]

Am I missing something here? EOLS is just LS + EO right? Like ZBLS? That's 302 algs on 3x3, and orders more on mega. If what you're doing is making a pair first and then EO while inserting it, that's VHLS and would be 5look LSLL. If you want something more efficient than OLL PLL, perhaps you could look into making fish and chips on mega.

Yes you are missing something, you first connect the pair THEN insert it. And as you said it's a pretty decent 4/5LLL approach for 50 algs but there could be better. A couple posts back I threw out some of my best ideas, and I think I'm gonna try to start generating algs for the third one. Problem is I don't have anything that can gen megaminx algs :/

 

[Arc]

Yes you are missing something, you first connect the pair THEN insert it. And as you said it's a pretty decent 4/5LLL approach for 50 algs but there could be better. A couple posts back I threw out some of my best ideas, and I think I'm gonna try to start generating algs for the third one. Problem is I don't have anything that can gen megaminx algs :/

I don't know if VHLS is really worth it. It's the same looks, just in slightly different order and with some more algs. Probably not going to save more than a handful of moves. Certainly better than just insert and EO but definitely nowhere close to OLL/PLL.

 

[Thermex]

I don't know if VHLS is really worth it. It's the same looks, just in slightly different order and with some more algs. Probably not going to save more than a handful of moves. Certainly better than just insert and EO but definitely nowhere close to OLL/PLL.

I didn't just propose VHLS for megaminx, that was Neuro. I think you're right that it's better than regular insert+4LLL but not that much better. I'm still trying to (intuitively) generate algs for the method I proposed a couple of pages above (EO, EP, CO, CP, all without the last pair solved), but it's difficult. It might turn out really good (sub 30 moves?) and if it is I'll post it, but I seriously need a megaminx algorithm generator...

Some other ideas I had recently:

Square-1 BOPE: I'm still working on 3×3 BOPE (OSC is done but PSC algs still need to be generated, I also have thought of some ideas for L8E that I'll post today on the BOPE thread), but I think I've thought of a pretty good way to transfer BOPE to square-1, it goes as follows:

1. FB on d-layer (DFL+DBL corners, DL+DB edge, basically just a regular 1×1×3 plus the DB edge)
2. PSC (same step as BOPE PSC, similar algs too)
3. Solve the remaining 6 edges (DF, DR, and all of the U-layer edges) in one algorithm.

This seems pretty efficient and would probably be around 120 algs. Opinions?

My second idea is just skewb TCLL, I know I've proposed this before and others have said skewb TCLL algs have been made before, but I've never actually seen a doc with the algs. I'd be willing to generate them if they don't already exist, but if they do, where can I find them??

 

[ananonymouscuber]

New method for 2x2
1. 3/4 of a face
2. Winter Variation
3. PBL

It would be good if you already had 3/4 of a face.

 

[Thermex]

New method for 2x2
1. 3/4 of a face
2. Winter Variation
3. PBL

It would be good if you already had 3/4 of a face.

This has already been created. It's called "SS" and I believe there's a wiki page on it. It's pretty much irrelevant considering HD is 3-4 moves less on average and has about 15 fewer algorithms.