The New Method / Substep / Concept Idea Thread

[Sue Doenim]

I was going to use WVCP recog which is like COLL almost.

Is it developed, and if so, can you give a resource? I would be quite interested in this.

 

[efattah]

Interesting. So in comparison to Roux the tradeoff is 4-5 moves saved + higher variance in solves in exchange for slightly worse ergonomics and a higher alg count.

What I like most about the idea is that it uses inspection very efficiently, planning ~13 moves of EG on average. If average movecounts are ~45ish, that's almost 30% of the whole solve, compared to ~10 moves planned in CFOP or Roux (approximately 16% and 20% respectively)

When I got into speed cubing a year ago I looked at CFOP and Roux and both suffer the same problem of executing a significant algorithm in the middle of a solve (42 for Roux CMLL and lots for CFOP LL). This meant that both the recognition and the mental algorithm 'fetch' are included as part of the solve time. It was for that reason that I created the LMCF method, whose ultimate purpose was to 'move' the large algorithm recognition & fetch time into the cube inspection time (recognizing and fetching EG1, EG2 and CLL during inspection like a 2x2 solver). If you finish the solve with pure LMCF and you don't use Waterman's complex final step, the rest of the LMCF solve does not require any significant algorithm recognition/fetch time (no more so than CFOP F2L). If, on the other hand you use the EGW variant and you introduce Waterman's complex final step, then indeed you do introduce one large-set algorithm into the middle of the solve, saving around 5 more moves at the expense of the recognition & fetch time. However EGW would have the lowest movecount of any speedsolving method, it would average almost 5 less than LMCF which is already 44 (so EGW would probably be 39-40).

The big learning curve for LMCF is the E2L phase, which is really similar to CFOP F2L in that there is so much work to do in on-the-fly optimization, lookahead, setting up cases to minimize rotations and so on. E2L requires an absolutely different way of thinking about the cube compared to F2L. Being good at F2L will not at all help you get good at E2L because they are so different. The E2L learning curve would be the major drawback for this method, in terms of the average cuber.

Another huge advantage of E2L vs. F2L, besides the previously explained statistical advantage of single pre-solved pieces, is that in E2L you can solve 3 edge pieces simultaneously (called an E2L triplet). For real time speedsolving I find that I can solve an E2L triplet once every second solve. The triplet does not take any longer than an E2L pair. Some of my fastest solves have solved 1 edge piece along with the corners, then solved two E2L triplets (6 edges pieces total), then all that is left is the final L5E step.

E2L triplets are actually 'available' in almost every solve, but for practical reasons setting them up isn't efficient. They sort of have to have a 'ready made setup' to be efficient to execute. Hence once every second solve they appear ready-made to execute.

The fastest Waterman solver I know of (somrandomkidmike) had some 12.XX Ao5's with Waterman and an 8 second single, and he also would solve 3 edge pieces at the same time when the opportunity showed itself.

 

[Teoidus]

I have to say that, having looked into CF and variants, I'm skeptical of your movecount estimates for EGW. Could you post the algorithms that you have so far, and maybe some example solves? As far as I understand it EGW solves are either either: EG + 1 edge (~14), 3x E2L (3 * 7? = 21?), L5E (14) which seems to yield 49 STM avg, or Waterman in it's advanced form: Layer (14), CLL (9), 2 Redges (7? i'm trying to use optimistic estimates, and from playing around 7 seems like a good minimum value), 2 Redges + Midge orientation (once again I will go with 7, that seems very optimistic), Midge permutation (5) which yields 42 STM.

 

[Shiv3r]

I have to say that, having looked into CF and variants, I'm skeptical of your movecount estimates for EGW. Could you post the algorithms that you have so far, and maybe some example solves? As far as I understand it EGW solves are either either: EG + 1 edge (~14), 3x E2L (3 * 7? = 21?), L5E (14) which seems to yield 49 STM avg, or Waterman in it's advanced form: Layer (14), CLL (9), 2 Redges (7? i'm trying to use optimistic estimates, and from playing around 7 seems like a good minimum value), 2 Redges + Midge orientation (once again I will go with 7, that seems very optimistic), Midge permutation (5) which yields 42 STM.

I would really like Alg sheets for E2L, efattah. If you make them, I will learn this method, because it means that any person good at 2x2 will be able to get pretty fast at LMCF(which I assume stands for Lazy Man's Corners First). I Mean, I like the fact that I can reuse my CMLL algs on 2x2(though they aren't too efficient Im still sub-5). It would be awesome to be able to do the reverse! Please, make an E2L alg sheet. If you want, I will make a tutorial for this method, I think it has serious potential.

 

[Metallic Silver]

Algs have been generated for it, 162 algs that aren't all very good isn't worth it

i actually want to see the algs

 

[Cale S]

i actually want to see the algs

https://docs.google.com/spreadsheets/d/1w_ZdrQP_IRK8p9MirpP0xKXu1j4k2pn24UUFPuFFtHw/edit#gid=0

 

[efattah]

Okay I created two example solves for LMCF. I am only semi-color neutral. I solve either the green or blue face (corners) first. If I cannot see ANY easy green or blue face corners, I will solve any other color and then do the appropriate EG algorithm, but since I am not good enough to predict the AUF I waste time doing the AUF if I start on any other color than blue or green (since starting with my 'normal' colors I do not need to AUF).

To be totally fair and prevent me from selecting 'favorable' scrambles I went into this week's race to sub-20 and picked the first two scrambles and solved them naturally, then reconstructed them in alg.cubing.net

Scramble: B2 D R2 D' U2 F2 U B2 F2 U2 L' B' L' D F' R F2 R2 B' U' L
y2 x' F2 L2 U L2 U' // face [5]
R U2 R' F R' F' R U' R U' R' // CLL [11]
U M U M' // 1 edge piece [4]
z R2 M d' M U M' // E2L triplet [6]
y' x M U M U2 M' U // E2L pair [6]
l2 M U M U M U M' U2 M' U' // solve last edge piece while orienting the midges, this is worst case scenario! [11]
M' U2 M' U2 M2 // permute midges [6]
Total 49 STM
Total for CLL 16
Total for E2L 16
Total for L5E 17

Scramble: D' F2 U' L2 U' R2 U F2 U R2 D2 B' U2 F U' R' D L' D B' U R
x' U r F r U' R' y2 // face [6]
R U R' F2 U F R U R' // EG1 Sune [9]
z R2 U M' U' // E2L pair [4]
R2 L2 U M' U2 M2 U // E2L pair [7]
x M2 U' M U2 M' U' // E2L pair [6]
R2 U M U2 M U R' // solve last edge while orienting midges [7]
U2 M' U2 M' // permute midges [4]
Total 43 STM
Total for EG 15
Total for E2L 17
Total for L5E 11

To comment on the movecount estimates. You'll see here that in the first solve we got a free triplet, and in the second solve we did not get a free triplet but we did get a free pre-solved edge piece on the blue face. This is typical; either you get a free edge piece pre-solved, or you get a free triplet, as both have a 50/50 chance of happening.

My EG movecounts are higher than 2x2 experts mostly because I am not color neutral and not really that good at making a face in 4 moves or less. However you'll see the extreme low movecount of the E2L phase. When counting the moves for an E2L pair, you need to include set up moves to be fair. These two examples solved 14 edges pieces in the E2L phase with 33 moves, for an average movecount of 2.35 per edge piece or 4.70 moves per pair which is WAY LOWER than an F2L pair.

I post the E2L algs asap.

 

[Teoidus]

I see why the E2L pairs are so moveoptimal now--they're almost all reducible to R/L setups of EO + Roux 4c 3-cycles! 1 move for EO, 4 to solve the edges, and at most 2 setup moves to place the two open slots at UL and UR -> pretty much 7 moves maximum. Very interesting.

I'm not sure how L2E + MidgeEO is as efficient as that seems like a bit of an awkward step, but I'll wait to see what the algs look like.

 

[VenomCubing]

Quick idea: Would it be possible to have a zz variant where instead of two 2x3x1 blocks on the sides of the line, you build two 3x3x1 blocks? if so, it would force a last layer skip every time because of edge orientation. If not zz, there would be only the 2-flip (opposite) ELL case after the blocks, because of permutation rules.

 

[xyzzy]

Quick idea: Would it be possible to have a zz variant where instead of two 2x3x1 blocks on the sides of the line, you build two 3x3x1 blocks?

After you build your first block, you have only one face that can turn freely, which makes it basically impossible to do the second block intuitively.

 

[VenomCubing]

After you build your first block, you have only one face that can turn freely, which makes it basically impossible to do the second block intuitively.

nevermind then.

 

[Metallic Silver]

https://docs.google.com/spreadsheets/d/1w_ZdrQP_IRK8p9MirpP0xKXu1j4k2pn24UUFPuFFtHw/edit#gid=0

Thank you!! This is gonna be useful > : )

 

[Shiv3r]

nevermind then.

I tried this once, we called it 1LLLLL, where we solved a solved line of pieces and EO solved. We had some algs to set it up from EOF2L-1, and it had only 15 cases for LL. I can send you the algs if you want.

 

[crafto22]

Hey guys, I started practicing with ECE again and came up with a few ways to optimize it. Here's how it goes:

1. Solve the FL and BL edges whilst ensuring the other E-slice edges are oriented

2. Use the experimental corner orientation method outlined on the SSC wiki page to orient the corners and solve the E-slice

I'd like to add that when I solve CO and the E-slice, I am almost always able to do this step in around 12 moves, and I've even had solutions as low as 7 moves. I believe this may be due to the fact that I actually do things more intuitively, rather than following every step.

3. Separate the corners into their respective layers

4. Permute all corners

5. Solve two or three D-layer edges, leaving L5E or L6E, whichever you prefer solving

6. Solve the remaining edges

Example solve:

Scramble: R2 U2 L2 B2 L2 R2 B' L2 B' D2 B' D F2 R' B2 L U R2 B' F2 U2

x2 D L' // FL + BL (2/2)
U R2 (D2 U') R U' R (U' D2) R' // CO + E-slice (8/10)
y' R2 (U2 D') R2 D' R2 D R2 D' R2 D' R2// corner separation + cancels into CP (11/23)
M' U' M' U2 M2 u y // 3 D-layer edges (6/29)
M' U M U M' U' M' U' M U2 M' U' M2 U'// L5E (14/43)

Total moves: 43 ATM

Although many parts of this solve may seem lucky, they are actually surprisingly common. Let me know what you guys think!

 

[genericcuber666]

whats the difference between cpls and wvcp? does cpls not orient corners? if it dosent isn't that not many algs and therefore viable?

 

[Neuro]

whats the difference between cpls and wvcp? does cpls not orient corners? if it dosent isn't that not many algs and therefore viable?

WVCP would orient and permutes the corners and CPLS only does permutation. CPLS is only 6 algs, but recognition is really difficult. If someone came up with a good recog system it could be good for OH or as a direct lead into 1L2GLL though

 

[shadowslice e]

Hey guys, I started practicing with ECE again and came up with a few ways to optimize it. Here's how it goes:

1. Solve the FL and BL edges whilst ensuring the other E-slice edges are oriented

2. Use the experimental corner orientation method outlined on the SSC wiki page to orient the corners and solve the E-slice

I'd like to add that when I solve CO and the E-slice, I am almost always able to do this step in around 12 moves, and I've even had solutions as low as 7 moves. I believe this may be due to the fact that I actually do things more intuitively, rather than following every step.

3. Separate the corners into their respective layers

4. Permute all corners

5. Solve two or three D-layer edges, leaving L5E or L6E, whichever you prefer solving

6. Solve the remaining edges

Example solve:

Scramble: R2 U2 L2 B2 L2 R2 B' L2 B' D2 B' D F2 R' B2 L U R2 B' F2 U2

x2 D L' // FL + BL (2/2)
U R2 (D2 U') R U' R (U' D2) R' // CO + E-slice (8/10)
y' R2 (U2 D') R2 D' R2 D R2 D' R2 D' R2// corner separation + cancels into CP (11/23)
M' U' M' U2 M2 u y // 3 D-layer edges (6/29)
M' U M U M' U' M' U' M U2 M' U' M2 U'// L5E (14/43)

Total moves: 43 ATM

Although many parts of this solve may seem lucky, they are actually surprisingly common. Let me know what you guys think!

So what times are you getting with it atm?

 

[TDM]

whats the difference between cpls and wvcp? does cpls not orient corners? if it dosent isn't that not many algs and therefore viable?

Something having not many algs doesn't make it viable. Solving the cube entirely with setup moves and T perms isn't viable for speedsolving. And many larger alg sets like OLLCP and ZBLL are viable.

I don't think CPLS is viable unless you know full 2GLL (except sunes are possibly optional). It's a lot of time to add on to a solve and won't necessarily save you much time during LL.

 

[genericcuber666]

Something having not many algs doesn't make it viable. Solving the cube entirely with setup moves and T perms isn't viable for speedsolving. And many larger alg sets like OLLCP and ZBLL are viable.

I don't think CPLS is viable unless you know full 2GLL (except sunes are possibly optional). It's a lot of time to add on to a solve and won't necessarily save you much time during LL.

If there's only 6 cases (like in a coll set) why can't you learn 27*6 recognition cases/angles. It couldn't be that hard, and you get the potentially amazing advantage of 2gll.

I didn't check what cpls really was since everyone said it was bad anyway.

 

[shadowslice e]

If there's only 6 cases (like in a coll set) why can't you learn 27*6 recognition cases/angles. It couldn't be that hard, and you get the potentially amazing advantage of 2gll.

I didn't check what cpls really was since everyone said it was bad anyway.

2GLL really isn't as much of an advantage as you think.