The New Method / Substep / Concept Idea Thread

[shadowslice e]

Have you seen Speed-Heise? I think it would be a good supplement to your methods.

which ones? ;-p

But no I had not seen speed-heise before, thanks for showing me .

Also, I'm definately putting thus on the wiki

 

[shadowslice e]

4x4x4+ method (orient first(ish)) completely experimental (so pls dont expect anything amazing)

1) orient the centres (so white/yellow, red/orange and blue/green centres)
2) orient edges either using slices or in the same way as as for zz 3x3x3 by pairing misoriented edges
3) pair edges using double slice or wide moves
4) solve centres cage style
5) ZZ 3x3x3?

NOTE: this is a highly experimental method which will need to be fully checked so don't expect it to be fast especially as some of the concepts are almost completely new as far as I am aware.

ok, building on this idea I have completely oriented a 5x5x5 cube to the <double turn> permutation group.

Basically I was seeing if I could do a thistlethwaite-like solve and I'm wondering what order would probably be most efficient.

To get to this stage I did opposite centre (ie. Green/blue, white/yellow, red/orange centres)

Next, I did the tredges by using freeslice/insertion. The orienting dictated that any edges could go with an edge with either the same colours, 1 same and one opposite colour or 2opposite colours (so a white/red edge could be matched with either another white/red, a white/orange, a yellow/red or yellow/ orange edge).

Afterwards, I placed the edges in their slots so each face (ignoring corners) was only opposite colours.

After this I oriented the corners using the same technique as human thistlethwaite.

So, after some experimenting I decided that the best way to do this is to pair the edges first then solve the centres like in a cage solve then do the 3x3x3 stage like the final stage of kociemba or orient first etc.

General evalutation of the solve:

Evaluation of the solve: centres: really easy
Centres (first phase) easy enough, no harder than normal reduction
Edges (first phase) not too bad but needs work for proper efficiency could add some algs for the last few edges. Bit of a hassle to permute them but it could be done like a 3 colour cube so is not too bad.
Parity: potential for really short alg and very easy to spot because the misoriented edges must always be on the last edges you finish.
Edge pairing: very fast, very easy just double moves and set ups
Centres: learn cage. Not too bad but may be slightly inefficient
3x3x3: very fast, fairly easy, very intuitive.

Pros:
1) Easy to perform once learned.
2) very intuitive.
3) Very easy edge pairing
4) if a way for the edges to be done efficiently, it could have quite a low movecount.
5) Few algs.
6) Potential for a shorter, more ergonomic parity alg

Cons:
1) Lots of small steps, could be good for computers
2) Doubling of standard reduction steps

Final thoughts:

This method might be mostly for fun although it would probably get better and better for the bigger big cubes as well as some applications for computer solving (although people have probably found much better ways to do it).

 

[mDiPalma]

ok

Have you tried expanding outwards from 3-color 1x1x1 > 2x2x2 > 3x3x3 > ... > nxnxn instead of using reduction to a 3x3x3 cube (and then solving to 3-color cube)? My gut tells me it would be more efficient

 

[shadowslice e]

Have you tried expanding outwards from 3-color 1x1x1 > 2x2x2 > 3x3x3 > ... > nxnxn instead of using reduction to a 3x3x3 cube (and then solving to 3-color cube)? My gut tells me it would be more efficient

do you mean build from a corner or using big block corners?

 

[IAmAPerson]

Even numbered big cubes suggestion (based on CFCE):

Redux to 3x3
Cross+F2L
CLL
ELL (fix parity, orient, and permute edges in one alg)

Pros:
With CFOP, you can still have up to a 4LLL even with full OLL/PLL because of parity. This method has 2LLL with any solve.

Cons:
Lots of algs
oh god no the recognition

 

[mDiPalma]

do you mean build from a corner or using big block corners?

I mean, building a 3-color cube expanding out from a corner, like Petrus on a 3x3x3. The tricky part is going from n*n-1*n-1 to n*n*n. OBLBL is probably the best way for that, in my opinion.

 

[shadowslice e]

I mean, building a 3-color cube expanding out from a corner, like Petrus on a 3x3x3. The tricky part is going from n*n-1*n-1 to n*n*n. OBLBL is probably the best way for that, in my opinion.

I've done some playing around and I've decided that it is easier and more efficient to do the OBLBL with a 3 colour cube. Perhaps the acronym for this method could be mismatched opposite colour MDPSSOBLBL (you can work out what this stands for) Lol name+ Lol method= awesomeness ;-P

In seriousness though, I'm going to look into this method a bit more to see if I can find a more efficient way to do centres as they will always be opposite each other (maybe make algs for each of the 5 corner centre cases and 5 edge centre cases- I'm not counting skips- maybe even the 35 one-look cases (that would occur on a 5x5x5)) and see if there are any special edge tricks that may work only for the specific cases that occur in the method.

 

[IAmAPerson]

Even numbered big cubes suggestion (based on CFCE):

Redux to 3x3
Cross+F2L
CLL
ELL (fix parity, orient, and permute edges in one alg)

Pros:
With CFOP, you can still have up to a 4LLL even with full OLL/PLL because of parity. This method has 2LLL with any solve.

Cons:
Lots of algs
oh god no the recognition

If I did math right, the total number of cases for ELL exceeds 300. The minimum number of algs needed is something which I don't know how to calculate.

 

[shadowslice e]

If I did math right, the total number of cases for ELL exceeds 300. The minimum number of algs needed is something which I don't know how to calculate.

There would be (for a 4x4x4) 10*(4+6+2+1)= 130 cases wouldn't there?

The minimum number of algs would be 29 (standard ELL) + 20 (parity ELL)= 49 including mirrors I think.

 

[IAmAPerson]

There would be (for a 4x4x4) 10*(4+6+2+1)= 130 cases wouldn't there?

The minimum number of algs would be 27 (standard ELL) + 20 (parity ELL)= 47 including mirrors I think.

Here was my math:

4 edge pieces with 2 orientations each = 4!*2⁴=384

Anyway, if my math is wrong and there are much fewer algs, I wouldn't be opposed to learning this myself. It would be after I finish normal ELL and learn COLL/WV (for ZZ OH, of course!)

 

[shadowslice e]

Here was my math:

4 edge pieces with 2 orientations each = 4!*2⁴=384

Anyway, if my math is wrong and there are much fewer algs, I wouldn't be opposed to learning this myself. It would be after I finish normal ELL and learn COLL/WV (for ZZ OH, of course!)

actually your case count is probably correct as I accidentally removed some Z-perms from my calculation although I think the alg count I got was slightly off as well- 29 standard ell (from the wiki) and for the parity there are

1 pure flip
4 U-perm 1 flip cases
4 anti-U-perm 1 flip cases,
1 H-perm 1 flip
2 Z-perm 1 flips

1 pure 3 flip
4 U-perm 3 flips
4 anti-U-perm 3 flips
2 Z-perm 3 flips
1 H-perm 3 flip

Total: 24 algs (53 algs)
You can also reduce further by excluding mirrors which I believe would give

1 U-perm
1 Z-perm
1 H-perm

1 pure flip parity
4 U-perm 1 flips
1 Z-perm 1 flip
1 H-perm 1 flip

2 pure flip
6 U-perm 2 flips
2 Z-perm 2 flips
2 H-perm 2 flips

1 pure 3 flip
4 U-perm 3 flips
1 Z-perm 3 flip
1 H-perm 3 flip

1 pure 4 flip
1 U-perm 4 flip
1 Z-perm 4 flip
1 H-perm 4 flip

Total: 14 algs for parity + 19 algs for normal = 35 algs in total excluding mirrors

 

[IAmAPerson]

35 algs in total excluding mirrors

You forgot the permutation parity. You would have stuff like W and O perms, too.

EDIT: ANOTHER NEW (maybe) METHOD

This one probably isn't as practical.

3x3 speedsolving method:
Solve corners
1-look edges
M2 BLD

Pros:
If one knows EG, this could be a 2-look solve (it would be a 3-look solve for me).
NO RECOGNITION PAUSES AFTER MEMO IS COMPLETE!
Even if one does not know BLD solving, they could still use this method and just 12-look the edges.

Cons:
HUGE gap for pure memo of edges.
M2 is not as efficient as normal speedsolving methods (though it is a very good method).
Memorize edges incorrectly? DNF unless you pause to make sure the puzzle is solved.

 

[shadowslice e]

Mega ZZ
1) Solve a 2x2x2 (3 centres, 4 edges, 1 corner)
2) solve a 2x2x3 (4 centres, 5 edges, 2 corners on the opposite side
3) solve F2L while orienting remaining edges (extend the 2x2x2 to two adjacent centres and merge with the 2x2x3 somewhat so there are only 3 unsolved faces left)*- (solved section now has 9 faces in it)
4) 3-gen S2L
5) LL

*you could orient then do the F2L but this would be slow

EDIT: NOTES: Edge orientation: while doing the f2L you can use partial edge control in the same way as on a 3x3x3 and then fixing any remaining edges using the F face or the Sexy moves.

For the LL, my method could provide a 3-look LL with only about 50 algs and the algs have fairly good recognition as the solved block could be places on the back.

Alternative (probably better) order: inverse tripod/ZZ/triangle/blockbuilding path thingy

Spoiler

Spoiler

Spoiler

Method instructions

Spoiler

1) find 3 adjacent centres and solve the edges and corner between them.
2) Pick two of the centres and solve the F2L such that there is now one extra block (as would be created by the Westlund method S2L on the back formed by the previous tripodesc block.
3) Using the centre opposite the block, solve a second F2L that would resemble the second step of the Yu Da Hyun style. However, you should orient the remaining edges on the last 3 faces so that you would get a 3-gen rest of the solve (I would do this by solving one edge between the F2L and the new centre such that there would be a slot for two pairs on each side of the centre/edge pair and then doing VHLS to insert the pair (there are 3 slots to insert for each face so I would recommend solving the one pair on each face then inserting them so that all the edges would be oriented - I think thus works out pretty much 100% of the time (apart from one case where there woild be 2 left over but this could easily be corrected by one sexy) as there would be 16 edges left and each VHLS orients 5 edges))
4) 3-gen ZZish style F2L
5) LL: My minimega tripod would give a 3 look LL with 42 algs using the intermidiate method and 2-look with 128. For standard OLL+PLL, this would reduce the OLL count. It may also be feasible to do a mega WV to give a 1-look LL for mega.

Pros:

Spoiler

1) 3-gen L3F (last 3 faces)
2) reduced LL alg count
3) Both hands can be used for the L3F with the mega oh the table as many of the moves that would be more awkward to perform are removed.
4) potential LS+1LLL
5) few rotations.
6) after the F2L there are no very hidden pieces (all the mega remaining can be seen with very few/no rotations.
7) fewer pieces need to be tracked for the first 2 steps compared to other methods.
8) you can tell people you use ZZ on a megaminx and see their expressions ;-)

Cons:

Spoiler

1) Building something other than an F2L may take some getting used to although for an experience solver there should be no problems.
2) without a fixed colour scheme it could be difficult to work out edge orientation for step 3

Evaluation:
[spoiler
The method has some potential but needs to be tested (like all methods) to determine the overall usefulness as well as ironing out any flaws in addition to finding new tips that could help.

How the method was created/ reasoning

Spoiler

Basically, I was trying to manipulate the rules for creating and maintaining oriented edges (mainly that an edge cannot be turned by 3 adjacent faces - like L,U &F on a 3x3x3 or else it will lose orientation- extending to mega we get the rules 1. Each face can only have 2 adjacent mobile faces which are not adjacent to each other & 2. The faces cannot be positioned in a closed ring with an odd number of faces) and finding that the maximum number of oriented faces it is possible to have on a megaminx is 5 in a Horseshoe shape or 6 in a ring. However, I found that it is difficult to get all these edges oriented in an efficient way and basically impossible to keep all the faces in view ( it is possible to keep 4 moving faces oriented and in view as well although it basically removes the last layer always being on top which can be disorientating for people who are not used to it (I'll post a description on how to do this as well at some point as I have created the beginnings of a method but hat needs even more work for improvement than the method I proposed above)) and so concluded that the best way to do this was to use the L3F as presented above using multiple VHLS inserts

Sorry if this is a bit confusing. I don't dealt know how to describe the shapes of blocks on a Mega. If this is the case, then the basic idea is this: solve until there are 5 faces left. Build a block like Yu Da Hyun while orienting the edges of the last 3 faces using something like VHLS or ZBLS. ZZ until last layer. Solve last layer however you would like.

 

[shadowslice e]

: it is possible to keep 4 moving faces oriented and in view as well although it basically removes the last layer always being on top which can be disorientating for people who are not used to it (I'll post a description on how to do this as well at some point as I have created the beginnings of a method but hat needs even more work for improvement than the method I proposed above))

Ok, here is that method ;-)

There are two options: LS orientation and a more ZZ style Eo.

LS option

Spoiler

1) Choose 4 centres and solve the corners and edges between them (this is like a 2x2x3 on a 3x3x3) as well as the opposite edge on the centres that only have 2 adjacent centres in the block rather than 3 (like the F and B faces in Petrus)
2) close one of the two centres with the extra edge and create a star with it (basically, solve the two remaining edges adjacent to the centre) in preparation for F2L.
3) Pair the edges for the F2L and move to the face above/adjacent to their slot
4) VHLS/ EOLS/ZBLS etc. You should now have all edges oriented (apart from two cases- one with 2 mosoriented edges (fix with a sexy) and one with 4 (two sexys)
5) you should now have 5 faces which are unsolved (4 in a horseshoe and one above all of them. You should now be have a 4-gen solve remaining (do not turn the U-face) you could do this in quite a few ways- building to an LL or you could do a 2 face final phase (Technique I'm currently working on although tbh I am only doing so because I like symetric patterns and my OCD would kill me to do this step non-symetrically.)

ZZ Eo option

Spoiler

1) Same as above
2) Just solve the F2L straight away without worrying about LS or Eo Etc
3) Using U moves, orient the edges in the same way as ZZ but remember to preserve the edge on the bottom.
4,5) Same as above

Pros:
1) 4-gen is very ergonomic and can be placed on the table for easy turning
2) For LL, case reduction
3) Less building for S2L/second stage (I'm not really sure what to compare it to or call it...)
4) ZZ on mega!!!!!

Cons:
1) OH GOD NO THE Eo DETECTION WHEN TRYING TO BE CN (for the ZZ variant)

Could be a fun method in my opinion and I think it has some potential to be fast although it would be intersting to see which Eo method is the best (out of the 3 presented in the 2 posts)

 

[2180161]

I looked at HTA, and this is similar but not quite the same

EO

place, in any order, 3 e-slice edges in e-slice

orient 3 d-layer corners

insert final e-slice edge with a corner correctly oriented

OCLL

F2l Using [R2,L2,F2, B2, U, D, M, U] but preserving EO

PLL

Like I said very similar to HTA, but with less algs

 

[shadowslice e]

I looked at HTA, and this is similar but not quite the same

EO

place, in any order, 3 e-slice edges in e-slice

orient 3 d-layer corners

insert final e-slice edge with a corner correctly oriented

OCLL

F2l Using [R2,L2,F2, B2, U, D, M, U] but preserving EO

PLL

Like I said very similar to HTA, but with less algs

More similar to SSC... https://www.speedsolving.com/forum/showthread.php?54056-SSC-(Shadowslice-Snow-Columns)-3x3x3-Method

The only real difference is that I combine a couple of steps (Eo+EoEdge, orient 3 corners+place 3rd eslice piece & insert the final e-slice piece+ orient all corners) and do LEE->LSE rather than LEE->EPLL as the last step (LFE- Last Four Edges) is faster than EPLL.

So not a new method really, but I would like some help to continue to improve SSC (both SSC-M and SSC-O).

 

[Berkmann18]

I came up with a method similar to OLS except it solves either the EPLL or the CPLL while inserting the R U R' or the R U' R' pair.
This method is more like a supset (not subset) or should I say a group constituted by EPLS (Edge Permutation and Last Slot) and CPLS (Corner ...).
There's a total of 12 cases for EPLS and 12 cases for CPLS (both pairs including) so 24 algs in total that leave a KhLL (Keyhole Last Layer aka pure COLL) which contains 80 algs.
So it's quite similar to KhLS (Keyhole Last Slot) except that it requires EO to be solved for EPLS (therefore using a TSEOLL method or simply ZZ would be a must) and CO to for the CPLS (as the recognition system is really tricky when the corners are not oriented the same way).
The aim of both is to have only one type of cube to solve during the LL

EPLS
Recognition order: RF-LU-RU
-O: opposite edge to RF
-R: adjacent edge to RF on the right side (when looking on the RF's original face, e.g: RF is green-yellow, the Radj would be orange-yellow)
-L: same as above but for the left side

Joint pair (the pair is in UL-ULF)
OL: R U' R'
RL: U R U2 R'
LO: U R U' R' U R U2 R' or R U R' U' R U' R'
RO: L' U2 R U R' U2 L
LR: L' U2 R U' R' U2 L U R U' R'
OR: U' R U' L R' U' R U L' R'

Split pair
OL: R L U' R' U L' U2 R U' R' U'
RL: R' D' L F2 L' D R2 U R' U or L U L' U2 R U' L U2 L' U R' U
LO: D2 L2 D F2 D' R L2 U L U L' U' R' D2
RO: R L' U R' U2 L U' L' U2 L
LR: L' D2 L U L' D2 L U' R U R' or R F U2 F D F' U2 F D' F2 U R'
OR: F2 U' L2 D' L2 U L D L' F2 U' or U' L2 U R U2 L2 U' L2 U' R' U' L2

CPLS
Recognition order: RFD-FUL-RUB
O: Oppposite sticker to RFD (e.g: RFD is blue then it would be a green sticker)
L: sticker of the left adjacent face to the original RFD face (e.g: RFD is blue then it would be an orange)
R: same as above but for the right
S: same color than the RFD's sticker

Joint pair
LS: R' U D R2 D' R' D R2 U' R D'
OO: U' L' U2 R U' R' U2 L or y' U r D r' U' r D' r'
LR: U2 R2 F' D2 F R' U2 R F' D2 F R2 or L2 U L' U2 R U R' U2 L' U' L2 U' L2
OS: U' F' L2 U' L' U2 L F U' F' L' U' L' F or R2 U2 R U2 R' U2 R U' F2 U' R2 U2 R2 U F2
RR: U L' U2 R U R' U2 L or d2 F' U L U L' U' F
RO: l' U2 R U2 r' U2 L

Split pair
Recognition order: FDR-FUL-RUB
SR: R L U' R' U L' U2 R U' R' U'
OO: L D' L2 U2 L U L D F2 L' U'
LR: U L' U2 R U' L U2 L' U R' U2 L
SO: R2 U' R' L' U2 L U L' U R L U2 R2
LL: R F D' F' U2 F D F' U' R'
OL: U R U R F U R' U' R F' R U2 R' U2 R


Pros of EPLS:
-You will end up with 0-4 corners to solve so it's useful for FMC
-Easy recognition
-No need to worry about LL corners

Cons of EPLS:
-You need to pre-orient the edges before using one of the cases

Pros of CPLS:
-no need to have EO done (so it reduce the LL to ELL so 29 cases)
-if you use the algs I listed and have EO than you would only get an EPLL or a skip)

Cons of CPLS:
-CO needs to be done before

Pros of PLS:
-you will end with either a KhLL case or an ELL case

Cons of PLS:
-for each sets the Orientation stage of the aimed pieces needs to be done before using PLS
-the F2L pair needs to be a 3-mover

If you have any suggestion about algs or anything that I could do to improve that method aside from getting a list of all possible PLS that includes AUF (so about 168-ish cases if I calculated it well)

 

[joshsailscga]

Sorry if I misunderstood, but isn't that something like 1000+ algs?

I'd guess even more than that, compare 3x3 PLL (21) to 3x3 ZBLL (492ish), and then remember that mega has somewhere close to 200 algs for full PLL...that doesn't even consider the VHLS.

 

[shadowslice e]

I came up with a method similar to OLS except it solves either the EPLL or the CPLL while inserting the R U R' or the R U' R' pair.
~snip~

https://www.speedsolving.com/forum/showthread.php?54615-Full-List-of-LPELL-Algorithms I believe that this is what you are doing when you say EPLS and the CPLS sounds vaguely like an LS extention of CFCE so it could be useful there.

The only other question I have is: how do you decide which case to use if you are using PLS? Surely you would have to look at almost all the pieces to recognise which to use (and so the recognition would be almost as bad, if not worse, than ZBLL)

 

[Berkmann18]

https://www.speedsolving.com/forum/showthread.php?54615-Full-List-of-LPELL-Algorithms I believe that this is what you are doing when you say EPLS and the CPLS sounds vaguely like an LS extention of CFCE so it could be useful there.

The only other question I have is: how do you decide which case to use if you are using PLS? Surely you would have to look at almost all the pieces to recognise which to use (and so the recognition would be almost as bad, if not worse, than ZBLL)

Kind of... And it's also useful for Keyhole (more importantly EPLS than CPLS).

As for choosing which set I'm gonna use, depends on how I want my LL to be and as some algs are both in the CPLS and EPLS I could predict the PLL and starting making cancellation straight-away.

And also you don't have to learn both sets, if you know KhLL than EPLS would be more suitable but if you know CFCE or just ELL than CPLS would be more suitable.


Sent from my iPod touch using Tapatalk