The New Method / Substep / Concept Idea Thread

[Neuro]

Yeah that's what I thought, worth a shot though

 

[Shiv3r]

update on progress with HD method:
So we have been working on the HD method(what was called "EG VOP"), and we are done generating the algs. We will not fully publish the method, however, until one or two of the contributors(Neuro, Thermex, me, and Sue Doenim) have learned the whole method. so basically all we need to do is speed-optimize the algs.

 

[obelisk477]

I used to enjoy reading this thread

 

[crafto22]

@Shiv3r Move count is definitely lower, there are more edge pieces to choose from. Regular cross is ~7 moves whereas this uses about 5. F2L can't possibly be any less efficient. One looking is easily possible. Anything else?

 

[crafto22]

I used to enjoy reading this thread

"Used to"? What is that supposed to mean? The way I see it, methods are only progressing further and further. The older posts from this thread are generally bad ideas or nonsensical hybrids. We've finally cut things down to efficient and viable methods for speedsolving

 

[Thermex]

Yeah, sounds good! Do you think it's okay enough to pursue, or should I go back to square 1 (pun not intended)?

Nice pun lol. I think it might be worth pursuing, the seperation step probably wouldn't be >150 algs but I'm kinda skeptical about the recognition and ergonomics for the 2nd step. As for the alg count of the step, my estimate is that there are exactly 150 algs for the 2nd step since there should be 30 ELL-like algs for all 5 possible orientations of the midges. Still though, I have the feeling the algs suck for the second step and aren't worth pursuing. Maybe though.
@Shiv3r I'll get back to work on the NLL algs today, I also think I'll begin learning the full method starting this weekend.

 

[obelisk477]

"Used to"? What is that supposed to mean? The way I see it, methods are only progressing further and further. The older posts from this thread are generally bad ideas or nonsensical hybrids. We've finally cut things down to efficient and viable methods for speedsolving

Because it's been the same, re-hashed stuff over and over for years and years. "What if we don't solve an edge down here first, we solve it last instead" Who cares? The derivative method is not going to be any better than the one it was pulled from, and so is not novel or helpful. Taking a Picasso, and photoshopping the tree to the left of the picture and then the house over to the right, does not make the picture any better (it probably made it worse), and the only person who should get credit for it being good at all is Picasso.

IMO, the only decent methods that have come out of this thread for like literally 2 years are LMCF and 2GR. They're pretty novel in their own right, or at least in the case of LMCF are revisiting an idea that was discarded long ago, and bringing new life into it. Everything else seems to be to have just been rearranging fluff, to virtually no effect.

 

[Hazel]

Nice pun lol. I think it might be worth pursuing, the seperation step probably wouldn't be >150 algs but I'm kinda skeptical about the recognition and ergonomics for the 2nd step. As for the alg count of the step, my estimate is that there are exactly 150 algs for the 2nd step since there should be 30 ELL-like algs for all 5 possible orientations of the midges. Still though, I have the feeling the algs suck for the second step and aren't worth pursuing. Maybe though.
@Shiv3r I'll get back to work on the NLL algs today, I also think I'll begin learning the full method starting this weekend.

Maybe it would be better to just orient the LL edges while orienting the midges as opposed to fully solving them, then EPLL + midge permutation? We could separate that into intermediate (EPLL then z + midges), and advanced (EPLL + midges at the same time). I'm not sure how many cases the advanced set would have, but it's definitely in the hundreds if not more.

 

[Shiv3r]

Because it's been the same, re-hashed stuff over and over for years and years. "What if we don't solve an edge down here first, we solve it last instead" Who cares? The derivative method is not going to be any better than the one it was pulled from, and so is not novel or helpful. Taking a Picasso, and photoshopping the tree to the left of the picture and then the house over to the right, does not make the picture any better (it probably made it worse), and the only person who should get credit for it being good at all is Picasso.

IMO, the only decent methods that have come out of this thread for like literally 2 years are LMCF and 2GR. They're pretty novel in their own right, or at least in the case of LMCF are revisiting an idea that was discarded long ago, and bringing new life into it. Everything else seems to be to have just been rearranging fluff, to virtually no effect.

is this 3x3 only? I have mostly given up on 3x3 methods. However, I have found that there are a lot of things to be innovates in other puzzles. I have come up with several unique methods for 4x4+. the only problem is that usually no one answers those posts and they get flooded.

 

[obelisk477]

is this 3x3 only? I have mostly given up on 3x3 methods. However, I have found that there are a lot of things to be innovates in other puzzles. I have come up with several unique methods for 4x4+. the only problem is that usually no one answers those posts and they get flooded.

Fair point. I meant 3x3 only. I think we're only touching the surface of possibilities for big cubes.

 

[Shiv3r]

 

[FastCubeMaster]

Ok smart people show me how you can 1LLSLL that's quicker than the 5 or 6 steps shown in the other thread

 

[AlphaSheep]

There is CTLS where if you have any pair in the LL, you can always do an R or F' to setup a ZBLL case... It could in theory be a 1LLSLL, provided you set up the pair every time. Good luck with the recognition though.

I think any 1LLSLL that's actually possible will involve setting up to an alg "intuitively", rather than memorizing a unique alg for every case.

 

[LexCubing]

Ok smart people show me how you can 1LLSLL that's quicker than the 5 or 6 steps shown in the other thread

Well buddy there's CFOP out there with 4 steps and it's really fast! Its even in its name. C-F-O-P!

If you meant look I said let's not care about that anymore. Have as much looks as you like but LSLL must be done in 1-look. For ex:
Cross->1st Pair->2nd Pair->
3rd Pair->Jabari Last Slot and Last Layer

There is CTLS where if you have any pair in the LL, you can always do an R or F' to setup a ZBLL case... It could in theory be a 1LLSLL, provided you set up the pair every time. Good luck with the recognition though.

I think any 1LLSLL that's actually possible will involve setting up to an alg "intuitively", rather than memorizing a unique alg for every case.

Correct me if I'm wrong but I know it's still 2 looks. Except F2L is "easier".

Lucky pair ->ZBLL
Pseudo pair->CTLS

When I said easier I meant if the F2L pair sucks there's a chance the other "pairs" will be easier and just do those. ZBLL is easier because of recog but CTLS still requires setup like CT.

Edit:
If everyone is confused when I said let's not care about looks I meant the looks before LSLL. For example:
Weird step 1 + WS 2 + W3 + ... + WS n + 1LLSLL
LSLL must be 1 - look. That's it.​

 

[AlphaSheep]

Correct me if I'm wrong but I know it's still 2 looks. Except F2L is "easier".

Lucky pair ->ZBLL
Pseudo pair->CTLS

When I said easier I meant if the F2L pair sucks there's a chance the other "pairs" will be easier and just do those. ZBLL is easier because of recog but CTLS still requires setup like CT.

Edit:
If everyone is confused when I said let's not care about looks I meant the looks before LSLL. For example:
Weird step 1 + WS 2 + W3 + ... + WS n + 1LLSLL
LSLL must be 1 - look. That's it.

I use ZZ, so I'll stick with that as an example. Hyperthetically the method could look something like this for someone with super human recognition.

1. EOLine (1 look)
2. Left block (2 looks)
3. 3rd square + LL pair (1 look, something like phasing - use alternate inserts to force a pair, which doesn't require an extra look)
4. CTLS (1 look) - you recognise the setup the same time that you recognise the alg.

It's not two looks. It's definitely a 1LLSLL. Recognition is just really really really hard.

 

[cuber314159]

Let me explain: as far as I know, no one has learned all 3668 1lll cases or algs so why do you then ask for a 1llsll which is 817344, that cannot be reduced further than 20000 algs from mirrors and inverse and auf so the only way of achieving anything with any relation to 1llsll is complex setup so even if it could be done it is not a viable speedsolving method I do not believe that recognition could ever be sub1 second also algs would be long and the execution would be hard so it is not viable for speedsolving as it either requires complex setup or a rediculous amount of very long algs

 

[Thermex]

Fair point. I meant 3x3 only. I think we're only touching the surface of possibilities for big cubes.

I see your point, I sometimes do tire of seeing the people who randomly pop up with variants of Petrus and Roux that are basically the exact same thing, and I think this thread is too focused on 3×3 as other puzzles could use some major advancements too (megaminx, skewb, etc.). But I think you're incorrect in stating that people "used to" come up with better ideas, most of the old ideas in this thread were just hybrids of Roux and CFOP. And I don't think that just because people recommend these kind of meh methods that aren't special in any way we can criticize them, part of this thread is getting people to understand that you have to be a lot more creative than just coming up with Roux/Petrus variants, and getting them into true method designing. After all this thread isn't supposed to be something you watch for entertainment, it's where people brainstorm their ideas, good or bad, and we make improvements on them.
Although I'm kinda done coming up with 3×3 methods at this point for the very reason I wrote in that last paragraph (they're hard to come up with), I still think there are ways of solving the LS+LL in CFOP and variants of WaterRoux that have potential and should still be pursued. Nonetheless, I still would like to see more ideas for non-3×3 puzzles come out of this thread.

 

[Shiv3r]

onetheless, I still would like to see more ideas for non-3×3 puzzles come out of this thread.

I have posted several times about methods for 4x4+, however no one else suggests any further developments. Also I think that 2x2 is actually somewhat unexplored, due to the mindset that "EG is love, EG is Life, EG is the only way to survive"(poetry intended). Also, seeing some intersting Square-1 methods or Skewb methods would be nice as well.

so guys, with all the 3x3 talk, there is still room for lolmethods on 3x3.
here's one I came up with a while ago, K3(basically a corners-first method based off K4)
1.solve D layer
2.CLL(can be done after step 3 if you want)
3.F2L with K4 F3L algs
4.ELL

 

[Neuro]

I think that from this point we need to do some major statistics and weight the advantges and disadvantages on 3x3 methods and substeps. From this point on 3x3, you'll probably get methods with similar movecounts but with different approaches that will all end up about the same. Doesn't mean we shouldn't pursue new methods (as we'll have more variety and hopefully advancements), but I honestly believe that for a fast human method the lowest possible average movecount while still retaining the ergonomics and lookahead necessary to be fast is ~43. Non-3x3 still has a good ways to go, especually cubes beyond 4x4 and non-cubic puzzles (minus pyraminx).

Speaking of 3x3 stats, I'll be running some major tests on what I believe are the best ZZ variants that are under 200 algs to see what tends to come up the most often, what is the most efficient, and what can give you the "luckiest" solves and averages. I'll be using ZZ-CT, ZZ-B (ZZLL), CLS, ZZ-WV, ZZ-VH (COLL+EPLL), and ZZ-Tripod. I'll probably use 50 random scrambles and try all of them out using Cube Explorer and HARCS to ensure all solves and methods are on an equal playing field in terms of efficiency up to the LS/LL variant. Any other ZZ variants that should be tested? Maybe someone can help me out for methods other than ZZ?

 

[Shiv3r]

(minus pyraminx)

I think there are still things to be discovered with Pyraminx, which was proven by my recent genning of all the FP method algorithms(On pyra I'm currently method neutral between LBL and FP-1, which gets you nice faces pretty often and only 9 algs, 4 for FP-1 set and 5 LBL)