The New Method / Substep / Concept Idea Thread

[Neuro]

I don't like PCMS that much personally, for some reason I can't do Pairs fast to save my life. But I can see some merit to it. Maybe what we can do is this:

1: Pairs: For some people this can be done extremely fast
2: Corners: 42 algs, all pretty easy
3: 1 Dedge +center: A little awkward, but with rotations it should be passable. Reduces the cube to <RrUM>
4: L7E (need a good way to do this)

 

[Thom S.]

Yaknow, I feel that a lot of people don't form their own educated opinions because they are not naturally curious enough.

Yes, that happens literally everywhere. Noone believs me when I day this though

Everyone hates L4C, for example. Why? Because whoever put algs on the wiki put the optimal ones from some obscure orientation. THESE ARE ZBLLs PEOPLE. THERE ARE GOOD ALGS FOR THOSE CASES; THOSE JUST AREN'T IT. Because of this misconception, everyone immediately throws out any method idea that ends in L4C.

L4C is a bit like PLL. There are good and nice algorithms there as ZBLLs have relatively much research put into them, but as it is the ZBLL subset with the most solved, they have a higher tendency to have a higher movecount.

Everyone hates CP-variants like CPLS. Why? Because some 9-yr old, somewhere in a thread last necro'd 100 years before Christ, said it was hard to detect corner permutation, and scarcely anyone has tried it since. Because of this, people throw out decent variants like CPLS, which are phenomenally efficient in their alg/movecounts, compared to other methods.

With the amount of Methods that use CP I can't believe they are this rare but you have your reason to say this. I think the thing is that detecting six relative Corner Permutations during the solve requires about as much dedication to do pauseless as Petrus' EO step pauseless does(and I see good Petrus users still having just a little pause) to bring the time back in Last Layer but CPLS is something even I have to look into

Everyone hates PCMS. Why? Because the name includes "S", suggesting 1/4 of this method is awkward S moves. You can do d/y moves to convert S to M. The average CFOPper does 10 TRILLION rotations per solve. Why is just 1 off-limits? Because of this misconception, everyone immediately throws out any idea that passes through a columns state.

As I'm a Roux user(depends on mood and day) I wouldn't say automatically a Columns First method is bad but as I know E Slices aren't popular and I think nobody voluntarily uses S Slices, there is just so much to do when Edges are on every Slice and all of E, S, M aren't solved.


Why do I love to talk with strangers about twisty puzzle methods so much?

 

[Neuro]

I’m learning how to solve Squan right now, and I’m confused as to why we don’t teach Roux n’ Screw as the primary beginners method.

F2B is really easy, and all the algs you need to know are M2 (which basically isn’t an alg), inserting 1 D layer edge, Adj CP, Parity alg, and one EPLL. 4 (or 5 w/ M2) algs, and it’s possible you don’t even need the EPLL alg if you do it intuitively.

Also quite fast; Roux n’ Screw can get sub-20 and lead into the Lin Method which has proven speed, rivaling Vandenberg.

 

[Abram Lookadoo]

new method (BCE)

1) block
solve a 2x2x2 block in DBL

2) corners
solve all the corners using any 2x2 method (translating L,B,D to R,F,U)

3) edges
solve the edges intuitively using commutators

 

[Thom S.]

I’m learning how to solve Squan right now, and I’m confused as to why we don’t teach Roux n’ Screw as the primary beginners method.

F2B is really easy, and all the algs you need to know are M2 (which basically isn’t an alg), inserting 1 D layer edge, Adj CP, Parity alg, and one EPLL. 4 (or 5 w/ M2) algs, and it’s possible you don’t even need the EPLL alg if you do it intuitively.

Also quite fast; Roux n’ Screw can get sub-20 and lead into the Lin Method which has proven speed, rivaling Vandenberg.

Yes, Roux'n'Screw, Lin and Yau-1 are all similarly good and beginner-friendly methods but they all have a flaw, which kills them for me(why we still teach Vandenbergh) and this is that almost every solve has a seeable pause after CS because you need to look at Separation(Orientation) and Permutation at the same time. F2B on Roux is really good because you have inspection and are able to plan out good blocks. After 4-6 Slices, it's just not feasible.

That's kind of why

But I mean, objectively you are right, yet Roux solvers also say that we should teach Roux as a Beginner Method because the intuition and few algorithms, and still, very few learnt it instead of Layer by Layer

 

[Neuro]

Yes, Roux'n'Screw, Lin and Yau-1 are all similarly good and beginner-friendly methods but they all have a flaw, which kills them for me(why we still teach Vandenbergh) and this is that almost every solve has a seeable pause after CS because you need to look at Separation(Orientation) and Permutation at the same time. F2B on Roux is really good because you have inspection and are able to plan out good blocks. After 4-6 Slices, it's just not feasible.

That's kind of why

But I mean, objectively you are right, yet Roux solvers also say that we should teach Roux as a Beginner Method because the intuition and few algorithms, and still, very few learnt it instead of Layer by Layer

Interesting points, I suppose the recog required for it can be hard. I still think I'll learn to solve using Roux n' Screw. I don't like algs and currently use Roux on 3x3+, so my Blocks recog is pretty good. From there, idk. Not looking to be world-class, but things could change.

As for beginner 3x3 methods, idk if we've really got a good one yet. IMO, LBL doesn't provide a true understanding of the cube. It's very easy to understand though, which is where most other methods fail (blockbuilding can be hard). Looking for a good beginners method may be something I end up doing over the next few weeks. I think that CriticalCubing's tutorial was a step in the right direction. But in terms of why LBL is used, a) it's easy to understand; b) tutorials are incredibly saturated towards LBL. Like, 95% of videos are LBL tutorials. Probably going to make a video series over the 4 major methods as beginner methods

 

[Thom S.]

Yeah, that's why LBL is so good as a beginner Method. You only need to have played around enough to intuitively understand how to build a cross. I mean, think about how hard it would be for non-cubers to build like a Roux Block or Petrus Block, but inserting one piece at a time is just right

Also, for Square-1, I personally also would like to have more freedom in Vandenbergh, do I made many tweaks and I'm working on Algorithm sets to 4(3 with luck)-look a solve, so there's that.
Keep solving Squan, it's really unique

 

[xyzzy]

L4C is a bit like PLL. There are good and nice algorithms there as ZBLLs have relatively much research put into them, but as it is the ZBLL subset with the most solved, they have a higher tendency to have a higher movecount.

This is bullsh— ahem, factually inaccurate.

Going from a somewhat outdated version of Jabari's ZBLL list I have saved, L4C takes 13.78 moves on average with speed-optimised algs. I don't recall what the move count is for speed-optimised ZBLL, but iirc it's around 14 moves? Not much of a difference, if any. And if you look at just FTM-optimal algs, ZBLL is 12.08 moves and L4C is 11.73 moves.

Everyone hates CP-variants like CPLS. […] Because of this, people throw out decent variants like CPLS, which are phenomenally efficient in their alg/movecounts, compared to other methods.

I agree with all of your post except this part. (Cargo cult is bad; people need to reevaluate ancient opinions more often.) 2GLL has a nice move count if you include weird algs that are like 5-gen or 6-gen, but once you restrict to just RU algs, the move count blows up and it's not really shorter than other ZBLLs. It's still faster because 2-gen is nice for ergonomics, but not by much, and I don't think this makes up for having to spend more moves up front to solve CP. (I can't comment much on other CP first methods due to lack of experience, but from brief testing, the RrU OLLCP algs aren't that great either.)

 

[BECubed]

I was brainstorming some method ideas and came up with a suggestion, here are the steps:
- Cross + 1 edge flipped
- F2L
- CMLL
- LSE

How does it sound? Currently I am averaging around 15 seconds with it.

 

[Sue Doenim]

I was brainstorming some method ideas and came up with a suggestion, here are the steps:
- Cross + 1 edge flipped
- F2L
- CMLL
- LSE

How does it sound? Currently I am averaging around 15 seconds with it.

Well, the cross + 1 edge flipped seems pretty arbitrary; I don't really see what it accomplishes over a regular cross. I would just leave out the cross totally (Roux), or leave out one edge completely (3CFCE). Actually, with one flipped cross edge, you can force an arrow EO case pretty easily, so this, but allowing the flipped edge to be swapped with it's opposite. Or you could put any non F2L edge in the DF and DB spots. Then you could just not solve the cross and just do the DR and DL edges. At that point, though, I think you're better off doing roux.

 

[BECubed]

Well, the cross + 1 edge flipped seems pretty arbitrary; I don't really see what it accomplishes over a regular cross. I would just leave out the cross totally (Roux), or leave out one edge completely (3CFCE). Actually, with one flipped cross edge, you can force an arrow EO case pretty easily, so this, but allowing the flipped edge to be swapped with it's opposite. Or you could put any non F2L edge in the DF and DB spots. Then you could just not solve the cross and just do the DR and DL edges. At that point, though, I think you're better off doing roux.

I agree, the sketch for this was originally meant to be like the TCLL of CFOP, where in 2x2's case you don't finish the layer, but in CFOP's case you just don't finish the cross. I also think it could be more efficient with ZZ-like blockbuilding, CMLL, and since you always get the same EO case, you could generate algorithms for the L5E and have it done more efficiently. Thanks though! - Erik

 

[Neuro]

I’ve been learning big cube stuff, and I think that past 4x4 using LEOR is pretty good for Roux users.

1: Centers
2: Freeslice pairing (influence FB)
3: First Block (1x2x3 obviously)
4: EO Line (I usually insert DB and do EO+DF) 3r, R, M rarely
5: F2L
6: COLL
7: EPLL

It’s also possible to do ZBLL/ZZLL, but 2step LL should be fine. Let me know your thoughts (credit to Pyjam for inventing LEOR on 3x3)

 

[Sue Doenim]

I’ve been learning big cube stuff, and I think that past 4x4 using LEOR is pretty good for Roux users.

1: Centers
2: Freeslice pairing (influence FB)
3: First Block (1x2x3 obviously)
4: EO Line (I usually insert DB and do EO+DF) 3r, R, M rarely
5: F2L
6: COLL
7: EPLL

It’s also possible to do ZBLL/ZZLL, but 2step LL should be fine. Let me know your thoughts (credit to Pyjam for inventing LEOR on 3x3)

Pyjam actually already does this, but he does first block like in Meyer.

 

[Thom S.]

I’ve been learning big cube stuff, and I think that past 4x4 using LEOR is pretty good for Roux users.

1: Centers
2: Freeslice pairing (influence FB)
3: First Block (1x2x3 obviously)
4: EO Line (I usually insert DB and do EO+DF) 3r, R, M rarely
5: F2L
6: COLL
7: EPLL

It’s also possible to do ZBLL/ZZLL, but 2step LL should be fine. Let me know your thoughts (credit to Pyjam for inventing LEOR on 3x3)

I'd say that doing FB is faster before Centres and Edges,but duch a statement would need long testing. What my suggestion just for your LEOR case would be is to build Opposite Centres, FB and a SB square(could be partial, doesn't even need a correct corner) to reduce the Edge pairing pieces, which could help a lot at higher order cubes

While COLL and EPLL is a great Big Last layer way, I'd say ZBLL is definitely something to put into practise if you're going for 4x4+ as going up from 6x6, saving some moves can really make a difference + you won't need M Slices, what more could you wish for.

 

[Hazel]

Here's something I thought of. It isn't the most efficient thing but lookahead is really easy because after the first step, all the pieces are on the F and U layers.
1) 2x2x3 in DB
2) Finish E layer
3) EODF
4) Insert last 2 F2L corners
5) ZBLL (or other LL method)

 

[Pyjam]

I did not invent Leor. It would be like saying I've invented the warm water. I did invent the name, however. And Arc suggested that we use "Leor" and not "LEOR", and I agreed.

I believe Leor is pretty good for 4x4:
1) 2 opposite centers
2) Left 4x3x1 block
3) Last 4 centers (Yau)
4) Edge pairing
5) EO-line
etc.

For 5x5 (and above)... hum... I'm not convinced by this order of steps.

I've tried what neuro has suggested and I think it's better.
I understand steps 2 and 3 like this:
2: Freeslice pairing (influence FB) = placing the edges at their correct place.
3: First Block (1x2x3 obviously) = solving the 2 left corners.

However, my cube isn't great and I'm not good at 5x5. So, I don't think I'm the best judge.

 

[BECubed]

Another thing that crossed my mind is VERY efficient corners first methods, like solving all the corners with advanced 2x2 methods, and then using 1 algorithm to solve all of the remaining edges. While not great in speedsolving, it would be good for potentially low movecount.

Reconstruction Solve:
F' L2 R2 D2 U' F2 D L2 U R2 D' L2 R' D L' U' B L R' F' L // Scramble

y L U R U R' U2 R U' R2 F R F' R U R' U' R U R' // Corners (19/34)
F' M' B' E2 F2 L F' E' L2 F M2 F' E2 L2 F // Edges (15/34)

34 HTM

See reconstruction on alg.cubing:
https://alg.cubing.net/?setup=F-_L2_R2_D2_U-_F2_D_L2_U_R2_D-_L2_R-_D_L-_U-_B_L_R-_F-_L_&alg=y_L_U_R_U_R-_U2_R_U-_R2_F_R_F-_R_U_R-_U-_R_U_R-_ F-_M-_B-_E2_F2_L_F-_E-_L2_F_M2_F-_E2_L2_F&view=playback

 

[Sue Doenim]

Well, the thing is that there doesn't seem to be much application for this, as solving all edges is going to be really huge (like, REALLY HUGE) on algs, and doing sll edges is going to be really heavy on slice moves. Slice moves are also counted as 2 moves in HTM, and HTM is what's used for FMC. I would look at LMCF though. It's definitely a super cool method.
https://www.speedsolving.com/forum/threads/full-lmcf-3x3-method-now-available.63998/

 

[Hazel]

I thought of a LSLL subset that may actually be useful. The idea is that if there's a 1x3 block on the last layer of the LL color and the LS is paired or joinable, then you can use an algorithm to solve the LS and OLL at the same time. I've generated a few algs so far, and they aren't bad.

 

[Sue Doenim]

I thought of a LSLL subset that may actually be useful. The idea is that if there's a 1x3 block on the last layer of the LL color and the LS is paired or joinable, then you can use an algorithm to solve the LS and OLL at the same time. I've generated a few algs so far, and they aren't bad.

So, a subset of VLS? I think it would only have 17 cases for the made pair, and 9 for the split pair.