[Help Thread] Roux Discussion and Help

[GuRoux]

I can't imagine recognition being that insanely hard... BD is hidden, but the other 5 are at least partially in view. If PLL can be recognized by looking at 2 sides, I can't imagine that some method for 1 looking LSE when 5 out of the 6 pieces are visible isn't doable. And 200 sounds like way too many to me. Permuting 6 pieces that are already oriented? 200? Really? Well, I don't know either. Can somebody calculate the number of algorithms it takes to permute L6E all at once?

Even if they're not used by anyone now, do the algorithms for one looking exist?

On a complete sidenote, can I just say that you have the most amazing username ever?

let me give a go at calculating this.

assuming one center orientation:
6!/2 = 360 with orientated edges
6!/2(2^5) = 11520 with any orientation edges

thanks, i'm surprised that you're actually the only person to have ever complemented it.

 

[GuRoux]

I'm averaging 13 seconds now, but I always have a long break before CMLL.
Is there any advice to avoid this break?
Do you fast guys recognize the CMLL case during SB?

I can't speak for the really fast guys but i when i practice CMLL i also time the recognition to get better at it.

 

[Bestsimple]

Do any of you have any advice for using less moves for LSE?

 

[GuRoux]

i always wondered how long LSE is suppose to be. anyone know?

 

[Bestsimple]

On the roux page it says it should be less than 15 moves.

 

[GuRoux]

On the roux page it says it should be less than 15 moves.

I think I am 18-19 in a speedsolve. It must have something to do with solving the orientation in a way to help the other steps.

 

[Bestsimple]

I have a lot of difficulty with 4c. In speedsolves i usually have a massive pause in the transition. Is there a method to predict 4c?

 

[Renslay]

I think I am 18-19 in a speedsolve. It must have something to do with solving the orientation in a way to help the other steps.

Mine is around 15 on average (usually between 12 and 18).

 

[already1329]

I have a lot of difficulty with 4c. In speedsolves i usually have a massive pause in the transition. Is there a method to predict 4c?

Predict the colour of the BU sticker.

 

[Kirjava]

Apologies in advance if this is a dumb question, or if it's already been asked and/or answered. Is there really no L6E 1 look permutation?

Of course there is, but if it was worth it it's likely that people would be doing it already.

You are another user blinded by the holy grailness of step concatenation enough to completely ignore the lack of advantage that comes with it.

It seems kind of ridiculous to have a step based on permutation of only the UL and UR pieces. I don't know how many cases 1 look L6E would be, but I can't imagine it'd be that incredibly many, would it? It's permutation of 6 pieces at once, and permutation of 8 pieces at once (PLL) is only 21 algs. I mean, people learn stuff like EG which is way more algorithm intensive than 1 look L6E permutation would ever be.

I've looked around a bunch and can't find any info on 1 looking it. Does this substep exist? If so, somebody please point me toward info about it, cause I sure as hell can't find any. If not... why not? I can't imagine recognition or execution being that tough.

I can't imagine recognition being that insanely hard... BD is hidden, but the other 5 are at least partially in view. If PLL can be recognized by looking at 2 sides, I can't imagine that some method for 1 looking LSE when 5 out of the 6 pieces are visible isn't doable. And 200 sounds like way too many to me. Permuting 6 pieces that are already oriented? 200? Really? Well, I don't know either. Can somebody calculate the number of algorithms it takes to permute L6E all at once?

Even if they're not used by anyone now, do the algorithms for one looking exist?

On a complete sidenote, can I just say that you have the most amazing username ever?

Algorithm quantity isn't the sole issue at hand here.

The current 'two-look' system we use for solving the permutation of edges gives solutions that are extremely close to optimal and essentially speed-optimal. It has a tiny recognition overhead allowing you to recognise the next step while completing the previous, leading to no recognition based pauses.

Recognition for this will be too hard, as a much much better alternative is available. You don't want to be taking a pause to look at 5 edges. Recognition for the current system we use requires you to look at two pieces, so it can be traced during the previous step. You'd need to trace 5 edges before you got to the step to be even close to performing like the "two-look" system we use. Then you have the issue of recalling and executing an algorithm from a large subset instead of following an intuitive pattern that gives basically the same solution.

I don't see the advantages in using a system like you propose. You have terrible algorithm quantity, terrible recognition and terrible recall for what are obstensibly exactly the same solutions.

 

[pijok]

Do any of you have any advice for using less moves for LSE?

Look at the Optimization part: http://grrroux.free.fr/method/Step_4.html
Learn at least the cases without *

i always wondered how long LSE is suppose to be. anyone know?

With some 4ab tricks sub 15STM avg isn't very hard

 

[5BLD]

Also predict 4a+4b in cmll

 

[Bestsimple]

So is that the BU sticker after you put the UR and UL edges in place?

 

[already1329]

So is that the BU sticker after you put the UR and UL edges in place?

Yes.

 

[IRNjuggle28]

Of course there is, but if it was worth it it's likely that people would be doing it already.

Well according to that logic, every new method being proposed would already be in practice if it was worth it. TCLL is doomed.

You are another user blinded by the holy grailness of step concatenation enough to completely ignore the lack of advantage that comes with it.

I've been dissed by Kirjava. Somehow I feel like I'm truly part of the cubing community now.

Algorithm quantity isn't the sole issue at hand here.

The current 'two-look' system we use for solving the permutation of edges gives solutions that are extremely close to optimal and essentially speed-optimal. It has a tiny recognition overhead allowing you to recognise the next step while completing the previous, leading to no recognition based pauses.

Recognition for this will be too hard, as a much much better alternative is available. You don't want to be taking a pause to look at 5 edges. Recognition for the current system we use requires you to look at two pieces, so it can be traced during the previous step. You'd need to trace 5 edges before you got to the step to be even close to performing like the "two-look" system we use. Then you have the issue of recalling and executing an algorithm from a large subset instead of following an intuitive pattern that gives basically the same solution.

Well, are L6EO and L4EP really that easy to recognize either? It seems like you'd have the exact same issues with recognition of those that you would with one look: BD being hidden and UB and DF being partially hidden.

I don't believe there's nothing you can add onto CLL or L6EO other than knowing how to force the occasional skip. There has to be something more structured that can be used. There's tons of add ons and subsets of CFOP LL, (winter/summer variation, ZBLL, OLLCP, phasing) and I bet something like that could be created for Roux. And would have already been, if Roux was as widely used and as widely studied as CFOP. I have no clue what it'll be, but I can't believe there's no way that Roux LL could get more efficient without sacrificing recognition. I am aware that for the purposes of this discussion, that's pretty irrelevant, though. You have good points.

I don't see the advantages in using a system like you propose. You have terrible algorithm quantity, terrible recognition and terrible recall for what are obstensibly exactly the same solutions.

Fair enough. I was a bit put off by the 4 look LL/ML of Roux, but if the last two steps don't require a pause if you know what you're doing, I guess it's not as big a deal. The struggle over whether or not to switch to Roux continues... Thanks for the help.

 

[GuRoux]

Well according to that logic, every new method being proposed would already be in practice if it was worth it. TCLL is doomed.



I've been dissed by Kirjava. Somehow I feel like I'm truly part of the cubing community now.



Well, are L6EO and L4EP really that easy to recognize either? It seems like you'd have the exact same issues with recognition of those that you would with one look: BD being hidden and UB and DF being partially hidden.

I don't believe there's nothing you can add onto CLL or L6EO other than knowing how to force the occasional skip. There has to be something more structured that can be used. There's tons of add ons and subsets of CFOP LL, (winter/summer variation, ZBLL, OLLCP, phasing) and I bet something like that could be created for Roux. And would have already been, if Roux was as widely used and as widely studied as CFOP. I have no clue what it'll be, but I can't believe there's no way that Roux LL could get more efficient without sacrificing recognition. I am aware that for the purposes of this discussion, that's pretty irrelevant, though. You have good points.



Fair enough. I was a bit put off by the 4 look LL/ML of Roux, but if the last two steps don't require a pause if you know what you're doing, I guess it's not as big a deal. The struggle over whether or not to switch to Roux continues... Thanks for the help.

I don't know that there is much else than tricks in the LSE, knowing different CMLL algs and what they do and blockbuilding skills. Either way, I think roux is the best.

 

[IRNjuggle28]

I don't know that there is much else than tricks in the LSE, knowing different CMLL algs and what they do and blockbuilding skills.

Not much yet. There's potential for much more than there is. ZZ and CFOP have had way more efforts to engineer add ons and shortcuts than Roux. But for now, agreed.

Either way, I think roux is the best.

I'd like to agree with you, even though the WCA rankings don't. Roux seems like a lot of fun.

 

[GuRoux]

Not much yet. There's potential for much more than there is. ZZ and CFOP have had way more efforts to engineer add ons and shortcuts than Roux. But for now, agreed.

I'm hopeful, but not so sure. I've tried to think up some possible places for such things but never really came up with anything.

 

[uberCuber]

Well, are L6EO and L4EP really that easy to recognize either? It seems like you'd have the exact same issues with recognition of those that you would with one look: BD being hidden and UB and DF being partially hidden.

1. You can fully recognize L6EO without seeing the BD piece, the BU sticker, or the DF sticker.
2. L4EP recognition does suffer from the hidden sticker problem, but the fact that there are only a few possible cases makes it rather easy to lookahead and predict the case in advance so you don't have to deal with on-the-spot recognition (which would require either rotating the cube around a bit to look at hidden stickers or taking a guess and doing a move that may or may not be correct)

 

[GuRoux]

1. You can fully recognize L6EO without seeing the BD piece, the BU sticker, or the DF sticker.
2. L4EP recognition does suffer from the hidden sticker problem, but the fact that there are only a few possible cases makes it rather easy to lookahead and predict the case in advance so you don't have to deal with on-the-spot recognition (which would require either rotating the cube around a bit to look at hidden stickers or taking a guess and doing a move that may or may not be correct)

I think for the L4EP, you're suppose to recognize the case by looking ahead in the UL/UR permutation step leaving no pause for recognition. For me, I do an automatic M' after UL/UR permutation which gives me recognition.