The New Method / Substep / Concept Idea Thread

[BaMiao]

Is there a method of reduction for the 4x4 where you do edge pairing a little different and instead of solving it like a 3x3 you can solve it like a 2x2? I'd like to see something like this.

You can do this, but it's much harder. Here's a link to one strategy, from SuperAntonioVivaldi: http://www.youtube.com/watch?v=CGgwWiRWhto

 

[Renslay]

Is there a method of reduction for the 4x4 where you do edge pairing a little different and instead of solving it like a 3x3 you can solve it like a 2x2? I'd like to see something like this.

Of course there is.
But since you have to handle different type of pieces at the same time (matching centers+edges+corners to form the "corner of the 2x2x2"), it is complicated and possibly highly inefficient. But you can try to reduce it to a 2x2x2, it is fun.

 

[mark49152]

Very true. I was trying to think this through before burning too much energy looking up algorithms. Unless the cases are remarkably easy to identify, this idea may be DOA. Also, all cases would need to be bounded by an acceptably low number of algorithms, all of which would need to be of an acceptable length (maybe 7 moves at most?).

Yes it would have to be a small number of short, easily fingerable algs to be worthwhile. If the objective is to reduce the number of potential OLL cases/algs, it would defeat the purpose to have to learn lots of LS algs instead. Plus, there will be circumstances where the algs won't apply and you may need to deal with other OLL cases anyway.

For my sune/antisune idea, I was hoping that with more target states available, all that would be required would be correct choice of a simple insert such as RU'R', RU2R', sledgehammer etc., perhaps with simple semi-intuitive setup moves. It wasn't that simple, so I dropped it.

iMHO the best potential for an LS step is in solving hard F2L cases. If your last slot is going to take 11 moves and a rotation or two to solve, why not learn a few options and do something else useful with those 11 moves too, like solve LL edges? One example would be ZBLS when both pieces are stuck in their slot. The EJLS concept could be extended such that if you encounter a connected pair such as F2L cases 11/12 or 23/24 mid-solve, and solving it would be awkward perhaps because a rotation would be needed, you could quickly insert it as-is, enabling simpler lookahead, and come back and solve it via ZBLS as the last slot.

 

[AvGalen]

Is there a method of reduction for the 4x4 where you do edge pairing a little different and instead of solving it like a 3x3 you can solve it like a 2x2? I'd like to see something like this.

not a speedsolving method. And if you would make such a method you wouldn't do 12*2 edge pairing but you would do 8 * 3 edge pairing where every group of 3 edges would be in a 2x2x2 block with the 3 correct centers and the correct corner attached. NOBODY would like to see something like that

 

[cubacca1972]

Yes it would have to be a small number of short, easily fingerable algs to be worthwhile. If the objective is to reduce the number of potential OLL cases/algs, it would defeat the purpose to have to learn lots of LS algs instead. Plus, there will be circumstances where the algs won't apply and you may need to deal with other OLL cases anyway.

I just checked how many cases there would be for my idea, provided that the last CE pair was set up as in WV. 216. Sigh. The problem is that you would need to scan these cases, and be able to pick out the one C-E-C group which can be solved using one of the algorithms. The problem wouldn't be the total number of algorithms, but rather the amount of signal noise as your brain sifts through those 216 cases, looking for the one C-E-C group that can be solved by the relatively few (and hopefully short) algorithms. You wouldn't even be able to look at one specific potential C-E-C group, as it might only fuse together efficiently in one particular combination of orientations (easy to see, hard to describe).

For my sune/antisune idea, I was hoping that with more target states available, all that would be required would be correct choice of a simple insert such as RU'R', RU2R', sledgehammer etc., perhaps with simple semi-intuitive setup moves. It wasn't that simple, so I dropped it.

iMHO the best potential for an LS step is in solving hard F2L cases. If your last slot is going to take 11 moves and a rotation or two to solve, why not learn a few options and do something else useful with those 11 moves too, like solve LL edges? One example would be ZBLS when both pieces are stuck in their slot. The EJLS concept could be extended such that if you encounter a connected pair such as F2L cases 11/12 or 23/24 mid-solve, and solving it would be awkward perhaps because a rotation would be needed, you could quickly insert it as-is, enabling simpler lookahead, and come back and solve it via ZBLS as the last slot.

I like the idea of a deliberate misplacement of a cubie or cubies to force certain paterns to happen. I used this in an Roux LSE idea, where you don't bother orienting edges, but rather solve RU and place the LU cubie flipped at LU. This only leaves 24 cases (12 with 3 M edges flipped, 12 with 1 M edge flipped), which then get solved along with LU.

Finding that holy grail of a new and sexy LS method for F2L solves seems to be elusive if not impossible.

My best idea was Petrus up to and including twisting the edges, then WV then PLL. Hasn't exactly taken the world by storm, but it still pleases me that WV gets mentioned from time to time.

 

[manyhobbyfreak]

New method for solving a 3x3?

I have been searching for different approaches to solve a 3x3 recently.

Trying out variations of CFOP, Petrus and Heise i came up with my own method.



Step 1. Solve a 2x2x2 block, (just blockbuilding/intuition)

Step 2. holding the block at DLB, solve "f2l-pairs" at FL and BR. (since nothing is solved in R L or U, feel free to rotate R L U for easy pair making)

Step 3. solve the front and right "cross-pieces". (be sure to use r/f moves for easy inserting, also you might want to try orienting some edges for the next step)

Step 4. solving the last 5 edges, (i tend to orient atleast a few edges during step 3 for easier solving/recognition)

Step 5. solving 2/3 out of the last 5 corners using commutators/algorithmes.

Step 6. Just a 3 cycle of corners or 2 flipped corners.



Has anyone thought about this way of solving before?
Do you have any tips for improving this method?

 

[Michael Womack]

Sound like this http://www.speedsolving.com/wiki/index.php/Hahn_Method or this http://www.speedsolving.com/wiki/index.php/Triangular_Francisco

 

[TheNextFeliks]

Sound like this http://www.speedsolving.com/wiki/index.php/Hahn_Method or this http://www.speedsolving.com/wiki/index.php/Triangular_Francisco

Not really. It's like noob heise.

 

[elrog]

If I recall correctly, it takes only 16 algorithms to orient the top 8 edges on a 4x4. The permutation of the edges, however, may take many algorithms because of the permutation possibly requiring and odd number of 2-cycles to solve. This bring me to the idea making the permutation of the edges even while orienting them. There are 2 possibilities between even and odd for both edge cycles making this require 4x16 or 64 algorithms. A large amount of these would be mirrors. 64 is definitely a learnable amount. After the permutation has been made even, the edges would have (4!/2)x(4!/2) or 144 permutations requiring 143 algorithms. This, I think is learnable, but I'm just questioning weather it is efficient to have to recall 1 out of 143 algorithms along side the fact that you cant master all 143. This is the only reasonable system for a 2 look last 8 edges though.

As a side note - I've been wandering why there hasn't been some version of Roux for the 4x4? I've been messing with it some and it seems pretty efficient.

 

[tseitsei]

If I recall correctly, it takes only 16 algorithms to orient the top 8 edges on a 4x4.

4x4 wing pieces don't have orientation. They only have permutation...

 

[Stefan]

4x4 wing pieces don't have orientation. They only have permutation...

Wrong. While you can't flip them in place, they certainly do have orientation. I can even tell you exactly where one is just by telling you which of the 24 possible ways it is oriented. And if for example your last layer is yellow and you want the yellow stickers of its eight wings to point up, then it's even *straight-forward* to think in terms of orientation, not permutation.

 

[tseitsei]

Wrong. While you can't flip them in place, they certainly do have orientation.

Well, IMO the fact that they can't be flipped in their correct places means that they don't really have an orientation. I mean you can't have a piece that is permuted correctly but not oriented correctly... That's why I think they have no orientation. I think it depends how you view the cube You CAN "orient" last 8 edges so that all U-face stickers are facing up, but to me that's not orienting but just permuting them differently...

 

[sub20cuber]

How to do a CFOP and Roux hybrid method

this is a new method I thought of whilst playing about with my cube. I am a cfop solver and was trying to get better with roux (4 fun) and came up with this hybrid of roux f2l and edge orientation with cfop pll with semi automatic coll.

STEP 1
firstly solve the two cross edges on the right and left(aka the DL and DR layer)and then build up 2 1x2x3 blocks on the left and right using either roux or cfop style depending on your current method.

STEP 2
Line up the M slice centers and do either corner orientation etc. if your bottom colour is white the make all the corners on the U layer have yellow facing up using either cmll ithink or 2 look using 2look oll cases to orientate the corners and then the y perm or the tperm to permute the corners.

STEP 3
orientate the rest of the edges so etc. if the top and bottom colour of your cube is white and yellow the edges have to have their red or yellow side facing down for the D layer edges or up for the U layer edges. to do this watch a roux tutorial as explain would take 2 long.

STEP 4
put the remaining D layer edges in the correct spots using M' U2 M to insert an edge in the front slot with the edge above where it needs to go or M U2 M' to insert an edge in the back slot with the edge above its correct slot

STEP 5
now you will have a pll edge cycle case hence the semi automatic coll I mentioned earlier. so the pll you will now be either the u perm the other u perm the h perm the z perm or a pll skip and a solved cube. all that is now left is auf.

thank you for reading this. this took a long time to type down so I hope you could at least try this out for yourself. I know this is not the fastest way and definitely not the best method but I have gotten sub 35 avgs with this method which is faster than my normal roux over 1 minute times.

please tell me what you think of this method and a new better name for it.
sorry if you thought of this method before me

 

[scottishcuber]

I don't know I looked it up and didn't find any other mentions of it

I think almost every single CFOP/Roux cuber at some point tried mixing them both. They don't mix well. You've taken away the difficulty and efficiency of block-building, as well as the intuitiveness of L6E with this hybrid. Ben no happy.

This isn't original at all.

also your Youtube link doesn't work.

 

[sub20cuber]

I made this for experimental purposes only I was only sharing this for fun I would of thought ithers would of done this before but they hadn't posted it also I changed my youtube name and link and stuff

 

[TDM]

STEP 1
firstly solve the two cross edges on the right and left(aka the DL and DR layer)and then build up 2 1x2x3 blocks on the left and right

Or you could just use Roux for this step.
And for 4 and 5, you've just done 4b as two different pieces. 4c is the same thing but, again, slightly different pieces. And longer algs.
Also you don't permute edges with an A perm.
You're only getting faster times because this is closer to your main method than Roux.

 

[sub20cuber]

I agree I am faster on this than roux because its nearer to cfop and thanks I meant a u perm not a a perm

 

[Noahaha]

Making a CFOP-Roux hybrid is like mixing milk and orange juice.

 

[elrog]

That would be very gross...

The way I view methods is by their last layer system with the best way to build the F2L for that system.

So CFOP is doing OLL/PLL with the most efficient way to build F2L up to that.

Roux is doing CLL/ELL, but builds the F2L with a missing M layer to shorten the CLL (and as a bonus it doesn't affect how you solve the M layer) and because it is a good lead into doing ELL. It also opens up the possibility of doing orientation for the 6 edges followed by permutation, though I personally prefer a more ELL approach. Not that this has anything to do with how good a system is, but it is possible to do L6E intuitively.

 

[Noahaha]

Roux is doing CLL/ELL, but builds the F2L with a missing M layer to shorten the CLL (and as a bonus it doesn't affect how you solve the M layer) and because it is a good lead into doing ELL. It also opens up the possibility of doing orientation for the 6 edges followed by permutation, though I personally prefer a more ELL approach. Not that this has anything to do with how good a system is, but it is possible to do L6E intuitively.

This is not a good description of Roux. You should expand your definition so that not every method has to have an LL system, especially since a lot of methods don't.

Roux is a method trying to get to LSE as quickly as possible.