[Help Thread] Roux Discussion and Help

[shadowslice e]

This has probably been proposed before because it is pretty easy to come up with, but I found this Roux variation and would like to hear your opinions on why the original Roux method is better (or not) than this:

1 first block
2 second block
3a\ orient edges using <M,U> ( = step 4a in original method)
3b\ place UL/UR edges in DF/DB position (or place UF/UB in this position), center colors of M slice doesn't matter. (=part of step 4b)
4\ COLL
5a\ place UL/UR (or UF/UB) in their home position (=part of step 4b)
5b\ finish M slice ( = step 4c in original method)

Why I think this variation of Roux can be good:

A) Instead of learning multiple algorithms to influence EO you can influence EO during completion of step 2 and avoid a 6 flip case easily.
B) During step 3 you don't permute or orient the top layer edges. This means you can recognize the COLL case during step 3. This should eliminate any pause before CxLL recognition.
C) Petrus, ZZ and this variation of Roux use the exact same algorithms.

The only downside I can think of is that COLL is by definition worse than CMLL, so this method has more moves than original Roux. The question basically boils down to: is the easy recognition and easy avoidance of 6 flip case worth the additional moves that COLL requires compared to CMLL?

This is very similar to Pinkie Pie (which used OLLCP rather than COLL).
One major issue with it is that is it much easier to transition from CMLL to LSE than it is from blocks to LSE. In addition, you also lose the ability to use many tricks for LSE which can reduce movecount by a lot.

CMLL is also much better than COLL in terms of movecount and ergonomics and frequently if you learn extra algorithms, they will frequently be the same as the "standard" but with a couple of moves made wide so it is not really so much learning an additional algorithm.

Lastly, this method will have more pauses when compared to regular Roux (and be less efficient) so will be slower overall. This can be from various things such as regrips between MU and RUrM or recognition etc.

Overall, I would say that regular Roux is better than this variant (though you may still be able to find some who think pinkie pie is better than Roux though these seem to be in the minority and shrinking).

 

[Fyzzna]

I reconstructed some of my solves today and was wondering if anyone would be nice enough to give me some tips.

1) 14.07
2) 14.03
3) 14.27
4) 13.59
5) 12.91

Note: This is not from a single ao5, as I only bothered with a reconstruction if the solve turned out to be decent. In the case of 'bad' solves (17+ seconds) I can usually spot what's going wrong while I'm solving.

 

[TDM]

I reconstructed some of my solves today and was wondering if anyone would be nice enough to give me some tips.

1) 14.07
2) 14.03
3) 14.27
4) 13.59
5) 12.91

Note: This is not from a single ao5, as I only bothered with a reconstruction if the solve turned out to be decent. In the case of 'bad' solves (17+ seconds) I can usually spot what's going wrong while I'm solving.

Spoiler: Solve 1

FB is fine. Note that you could also have done y' x' R2 U r B to solve the square: some people prefer wide moves to slices, so just giving this as an option. It also happens to lead to a shorter FB (finish with F' M' R' F), though that's just by luck in this case.


SB: There are two more efficient ways I can think of to solve the first square. The first one uses the fact that the green-white edge and green-white-orange corner are already paired together. Start with r' U' r2 to put the green-orange edge into DR. Note that doing r2 at the end not only preserves the pair, but also orients it correctly, so you can insert with U2 R U' R2.

The other way to solve the squares starts the same way as your solution does, with r U R to begin solving the DR edge. However here, instead of doing U' R U2 M2 to make the pair, you can do U M2 instead, and then finish the square with U' R.

Your solution to the pair is perfect.

CMLL: Try to avoid AUFing too much. Note that after just a U move, you should be able to see all the stickers you need to recognise the CMLL (they would be UBL, UFL, RUB, RUF). This would save you having to AUF twice.

LSE is perfect too. Nothing I'd do differently.

Spoiler: Solve 2

FB: Again, starting with R2 r' u might be good to avoid slice moves. But everything else is good.

SB: Generally, 2-gen is faster than non-2-gen. Although it is one move longer, the solution R U R U2 R U' R' U R U R' would let you get more TPS and have fewer regrips, so is a better solution.

For the pair, did you think the edge was flipped the other way? If so then your solution was fine, since you corrected your mistake. But if you hadn't seen the edge was there then you shouldn't have AUFed until you'd determined where the edge was, since that's just an unnecessary move.

CMLL: Fine.

LSE: Given your speed, it might be good to learn some of the easier EOLR cases - especially the arrow cases. Here, after M2 U, the EOLR case is one move: M'. Doing this saves four moves, which is almost 10% of your solve! Once you know CMLL, EOLR should definitely be the next alg set you learn.

Spoiler: Solve 3

FB: Once more your solution is good. But if the FB is ever this short, then this should be a chance to plan where your DR edge is going to be. Find where it is and, in inspection, follow where it'll move to after each move of your FB. That way you won't need to start SB with things like R' R2, because you'll know where it is already!

SB: Another solution to the square (after the R' R2) is r' U r U' R' U' R. The pair is good.

CMLL: Again, try to learn to recognise T (and U) cases from the side. The stickers are still all visible if you're looking at the U/R/F faces.

LSE: You were unlucky, there's not much more you could have done here, not even with EOLR.

Spoiler: Solve 4

FB: Square is good, but try to use E moves instead of things like u U'. Again, M' U r2 F' might be better than using M2 to avoid slices.

SB: Same comment as last solve for the square, because it's the same case Not much to be said about the pair, as usual.

CMLL: did you... AUF away from your normal recognition angle?

LSE: Good assuming you don't know EOLR. It's not a very good EOLR (only saves 1 move) but still, it's definitely better than not saving a move at all!

Spoiler: Solve 5

FB: Same comment about finishing with r2 F' as a possibility, but again, a good solution.

SB: After the R U, you should be able to notice the orange/blue pair, which looks much easier than the other one (the edge is oriented, which is never a bad thing). U' R' U2 R U r U R U R or r U' R' U R U R U2 R would be good continuations from here, but it's definitely an awful case for either pair.

For your pair, U2 R' U R U' R' U' R is faster. I know it's the same as the CFOP solution, but that's not a bad thing!

CMLL: Do you know r U' r2 D' r U' r' D r2 U r'? It flows a bit better imo.

LSE: Exactly how I would have done it.

 

[Fyzzna]

Wow, thank you for this extremely in depth analysis

Spoiler

FB is fine. Note that you could also have done y' x' R2 U r B to solve the square: some people prefer wide moves to slices, so just giving this as an option. It also happens to lead to a shorter FB (finish with F' M' R' F), though that's just by luck in this case.

Yeah I've been trying to use more r moves (feels much more ergonomic to me) but it's a habit that's been surprisingly difficult to break.

SB: There are two more efficient ways I can think of to solve the first square. The first one uses the fact that the green-white edge and green-white-orange corner are already paired together. Start with r' U' r2 to put the green-orange edge into DR. Note that doing r2 at the end not only preserves the pair, but also orients it correctly, so you can insert with U2 R U' R2.

That's a really neat way of solving the square, I'll definitely need to look into that more!

For the pair, did you think the edge was flipped the other way? If so then your solution was fine, since you corrected your mistake. But if you hadn't seen the edge was there then you shouldn't have AUFed until you'd determined where the edge was, since that's just an unnecessary move.

I did think that the edge was flipped. Still a very weird mistake as I definitely should have seen it when finishing the square.

LSE: Given your speed, it might be good to learn some of the easier EOLR cases - especially the arrow cases. Here, after M2 U, the EOLR case is one move: M'. Doing this saves four moves, which is almost 10% of your solve! Once you know CMLL, EOLR should definitely be the next alg set you learn.

Holy ****, a four move save is pretty insane. Though after watching the video it sounds like EOLR is an intuitive step rather than an alg set or am I missing something?

CMLL: did you... AUF away from your normal recognition angle?

Hahaha I'm not quite sure what happened there myself

SB: Generally, 2-gen is faster than non-2-gen. Although it is one move longer, the solution R U R U2 R U' R' U R U R' would let you get more TPS and have fewer regrips, so is a better solution.
[...]
SB: After the R U, you should be able to notice the orange/blue pair, which looks much easier than the other one (the edge is oriented, which is never a bad thing). U' R' U2 R U r U R U R or r U' R' U R U R U2 R would be good continuations from here, but it's definitely an awful case for either pair.

For your pair, U2 R' U R U' R' U' R is faster. I know it's the same as the CFOP solution, but that's not a bad thing!

I actually never learned proper F2L (switched straight to Roux after learning the beginner's method) and just tried to work out how to solve certain cases myself which is why I'm probably using way too many M moves in my second block. I'll definitely have to do some slow solves to try and find more efficient ways to for a lot of cases.

CMLL: Do you know r U' r2 D' r U' r' D r2 U r'? It flows a bit better imo.

I use r U' r2 D' r U r' D r2 U r' for the Pi Columns case (basically the same alg) so this one shouldn't be a problem to learn, it really does flow better.

EDIT: Put my answers in a spoiler so they don't take up the entire page.

 

[TDM]

Holy ****, a four move save is pretty insane. Though after watching the video it sounds like EOLR is an intuitive step rather than an alg set or am I missing something?

Well, it's a bit of both. Some people say it's intuitive, others say it's a mix of intuitive and algorithmic. For me, the "algorithms" definitely feel intuitive since I can see how the pieces are moving around, but some of them aren't obvious to me without having seen the case in the table/video beforehand. I don't really know if it counts as fully intuitive or not... but algorithms or not, it's good to know I still don't know around half of the cases but I'm gradually getting there. I wish I'd started sooner.

I actually never learned proper F2L (switched straight to Roux after learning the beginner's method) and just tried to work out how to solve certain cases myself which is why I'm probably using way too many M moves in my second block. I'll definitely have to do some slow solves to try and find more efficient ways to for a lot of cases.

Ahh, interesting! Kian did that too (switched straight to Roux from beginners). Untimed solves are definitely good ways to find new tricks, but also take a look at example solves and reconstructions if you want. I find untimed solves more fun since you're exploring for yourself, but there are always loads of things you'd never think about doing.

 

[applezfall]

what do you guys think of my 10.72 (5th best solve )
https://alg.cubing.net/?setup=_B-_F...r R_U-_L-_U_R-_U-_L_ U2_M-_U_M_U_M2_U-_M_U2_M

 

[Underwatercuber]

what do you guys think of my 10.72 (5th best solve )
https://alg.cubing.net/?setup=_B-_F2_R-_U2_B2_R-_F2_D2_R2_U2_L_R2_F-_D_U2_B_D_F-_L-_D2&alg=z2 D-_L-_D_U_R_U2_R-_U_F- U-_R-_U-_R-_U2_R-__U_r_m-_U2_r-_U-_r R_U-_L-_U_R-_U-_L_ U2_M-_U_M_U_M2_U-_M_U2_M

Looks pretty efficient, rip on tps tho

 

[applezfall]

Looks pretty efficient, rip on tps tho

lot of pauses during 2nd pair on sb and l10p and I turned pretty slow

 

[SpectralChimaera]

So I began learning the roux method, my average is about 2.5 minutes at the moment. Liter ally just finished learning the basic algorithms today and I'm still using notations to remind me which corners go where for the niklas.

My curiosity stems to how fast roux can be and how difficult it would be to transition to CFOP from roux I'd i ever felt like it.

 

[TDM]

So I began learning the roux method, my average is about 2.5 minutes at the moment. Liter ally just finished learning the basic algorithms today and I'm still using notations to remind me which corners go where for the niklas.

My curiosity stems to how fast roux can be and how difficult it would be to transition to CFOP from roux I'd i ever felt like it.

Many people would agree that Roux and CFOP are very close for two handed solving. Depending on your current CFOP times, the transition can take anywhere between a couple of weeks and a few months, if you practise effectively.

 

[SpectralChimaera]

Oh I don't know CFOP, I've Literally only done roux.

Many people would agree that Roux and CFOP are very close for two handed solving. Depending on your current CFOP times, the transition can take anywhere between a couple of weeks and a few months, if you practise effectively.

 

[TDM]

Oh I don't know CFOP, I've Literally only done roux.

oh whoops, I read that the wrong way round! But the same thing's true.

 

[SpectralChimaera]

oh whoops, I read that the wrong way round! But the same thing's true.

So basically roux and CFOP can be as fast as each other and it comes down to preference?

 

[SpectralChimaera]

Question for the group. I'm currently doing niklas into sune wine to lse etc. Is there a faster or more efficient way of permuting and flipping the corners then finishing?

 

[TDM]

So basically roux and CFOP can be as fast as each other and it comes down to preference?

Near enough. Some people will tell you CFOP is faster, other people will tell you Roux is faster. There isn't any way to measure it, so everything's based off theory/guesses.

Question for the group. I'm currently doing niklas into sune wine to lse etc. Is there a faster or more efficient way of permuting and flipping the corners then finishing?

It's better to orient the corners, and then permute them afterwards. This has better recognition, and also makes it easier to learn more advanced methods.

Orientation alg: Sune (same as you're doing now, only you do it first.)
Permutation alg: Hold the headlights (solved corners) on L, and do: R U R' F' R U R' U' R' F R2 U' R'.

After that, you can learn 2-look CMLL, which uses one (of 7) algs to solve orientation, and one (of 2) algs to solve permutation, i.e. 9 algs total.

 

[SpectralChimaera]

I wont directly quote since its a pain on phone.

So once you've done both blocks, you do sune to get headlights (is that blinkers on top, the sides or double headlights) the do that new permutation to finish? Or does that lead to LSE?

Although hang on that doesn't take into account corners not being where they're meant to be after blocks .

Unless I missed something, which is likely.

 

[TDM]

I wont directly quote since its a pain on phone.

So once you've done both blocks, you do sune to get headlights (is that blinkers on top, the sides or double headlights) the do that new permutation to finish? Or does that lead to LSE?

Although hang on that doesn't take into account corners not being where they're meant to be after blocks .

Unless I missed something, which is likely.

(there's an edit button if you want to add more information to a comment)

Sorry, that's my mistake, the word "headlights" has multiple meanings in cubing. Let me rephrase.

Step 1: Orient corners with sune. They won't be in the right places.
Step 2: That new algorithm I gave you swaps two adjacent corners. Use this to solve the corners correctly. (If you have a diagonal swap of corners, you will need to apply the algorithm twice).

 

[SpectralChimaera]

(there's an edit button if you want to add more information to a comment)

Sorry, that's my mistake, the word "headlights" has multiple meanings in cubing. Let me rephrase.

Step 1: Orient corners with sune. They won't be in the right places.
Step 2: That new algorithm I gave you swaps two adjacent corners. Use this to solve the corners correctly. (If you have a diagonal swap of corners, you will need to apply the algorithm twice).

So is this way faster or slower than niklas into sune?

 

[SpectralChimaera]

Near enough. Some people will tell you CFOP is faster, other people will tell you Roux is faster. There isn't any way to measure it, so everything's based off theory/guesses.


It's better to orient the corners, and then permute them afterwards. This has better recognition, and also makes it easier to learn more advanced methods.

Orientation alg: Sune (same as you're doing now, only you do it first.)
Permutation alg: Hold the headlights (solved corners) on L, and do: R U R' F' R U R' U' R' F R2 U' R'.

After that, you can learn 2-look CMLL, which uses one (of 7) algs to solve orientation, and one (of 2) algs to solve permutation, i.e. 9 algs total.

I think I need a diagram or video because as soon as I've done the left and right block I do sune but everything's out of whack then the next one does nothing D:

 

[SpectralChimaera]

Ok so after a bit of research I found lubixcubes website which denotes all the 7 Alg for orienting the yellows up then permitting the corners. So I'm now in the process of learning all the sexy moves and j/y perm. Next question though, is that the height of roux? Or are there algs which orient and perm at the same time?