Roux Method?

[waffle=ijm]

I use the Roux method to speedsolve after completely stopping trying to learn learning fridrich because the OLL algs annoy me so bad. I've got all the CLL algs down for the U corners, but i still can't get an easy sub-20 because I'm having trouble with the 1x2x3 blocks for both L and R sides.

can anyone help me?

 

[pjk]

What are you having trouble with exactly? Finding pieces? Optimizing moves?

What is typical breakdown?

 

[waffle=ijm]

usually tracking the pieces are easy, but the way i move them is too inefficient. i stick with M slices but i just need a more efficient way to do it

 

[DonQuixote]

With Roux it's all about practise, you won't stop finding more optimized ways to build the first two blocks even after a year. So just wait for you getting better thus being able to find better sequences.
What you could do, too: find the cases you can handle worst and exclusively spend time on them. After some months, do it again, your worst case scenarios will be different ones.

Kind Regards
Martin

 

[waffle=ijm]

Very true. I'm taking the block building slower than anything because the most efficient way for me to do it is to get a 1x2x2 in one sequence and attaching the pair to complete the 1x2x3. That last pair is really slowing me down. So I guess I'll start practising there

 

[fanwuq]

Does 2 1*2*3 blocks use significantly less moves than Fridrich F2L?
How many moves on average?

 

[rubiksfriend]

About 48 (or more) in speedcubing. I also (not that I'm an expert) recommend experimenting with other block methods (Heise, Petrus) just to find some new was to put pieces in general together. Then, go back to Roux and it will seem easier than building a deep block (1x2x3 vs. 2x2x2).

 

[pjk]

Waffle: Yeah, just practice, like Martin said. Work on minimizing movecount, minimizing delays, etc.

Fanwuq: Oh yes, by a long shot. I'd say 2 1x2x3 blocks is 15-20 moves on avg.

 

[fanwuq]

Wow! and that's during a speedsolve, not FMC?!?!!

 

[waffle=ijm]

Fanwuq: It takes me about 8-10 moves for easy cases to make the first block and "prepare" the second block which takes about 5-8 moves for easy cases. Other cases, I panic and quickly do an F2L alg to get a pair. Roux is suitable for fewest moves even though I'm not into that.

 

[pjk]

Wow! and that's during a speedsolve, not FMC?!?!!

It is actually around 18-24 moves avg.
Here is a typical breakdown of a solve for me:
First 2 Blocks: 23 Moves
CMLL/COLL: 10 Moves
Solve Remaining 6 edges/centers: 14 moves

Getting better cases happens quick often where you end up with a 40-43 move solution. Yesterday, while doing solves with scrambles off of JNetCube, I did a solve, which popped a couple pieces, I was able to put the pieces back in correctly and finish solving in 22 seconds, which is decent for me. I went back and did the solve again, and it ended up being 36 moves. So it was a 36 move speedsolve. At full speed, like Roux himself speed, I think it would be a 10.xx or faster solve.

 

[rubiksfriend]

Waterman averages fewer moves.

 

[brunson]

Waterman averages fewer moves.

Fewer, not less. You should consider changing your handle to watermanfanboi. ;-)

I use the Roux method to speedsolve after completely stopping trying to learn learning fridrich because the OLL algs annoy me so bad.

That's funny. I know all my OLL cases from Fridrich, but I've just started learning a simplified Roux and I find it really hard not to use my OLLs to solve some of the cases I come across.

 

[rubiksfriend]

I know, but there is a reason. Guess what speedcubing method I read/found out about first? A bit strange, you might say, but I've loved it ever since. Of course, Roux is pretty beastly. Pretty good move-count. Plus, not much to memorize. I hate memorizing, unless the moves contain only three types of turns (U,M,R).

 

[brunson]

I know, but there is a reason. Guess what speedcubing method I read/found out about first? A bit strange, you might say, but I've loved it ever since.

That's actually awesome. I'm a Fridrich user and I'm starting with some ZBF2L, but I've played with Petrus and a just now a little Roux. I know I've given you grief about Waterman in another thread, but it's from a scientific standpoint, I'm not actually all that interested in defending Fridrich. I love seeing people excel with methods other than Fridrich. Fridrich LL is kind of braindead and robotic, I think Roux is totally fun for the last six edges, I'm trying to do it completely intuitively and it's a great break from practicing lookahead in Fridrich F2L with my iPod playing a recording of a metronome in my ear.

What I think is interesting is the convergence of the advanced forms of all the methods. Lars Petrus himself uses a one look PLL in competition, I believe, and the more advanced LBL methods, like ZB, VH and MGLS are all about edge orientation before LL. I've only been playing with Roux for a little, but every now and then during edge permutation I'll hit a spot where I can pull a set up move + PLL to "cheat" and finish the solve.

Anyway, I just don't think there's enough objective data to make a solid claim about Waterman being faster or slower than any other method. When I ask you to back things up it's because I want to see hard numbers to either confirm or deny the claim. I'd love to see some of the Fridrich champs dethroned by a different method, even if it's one of them converting, themselves.

But until it happens it's all speculation.

Of course, Roux is pretty beastly. Pretty good move-count. Plus, not much to memorize. I hate memorizing, unless the moves contain only three types of turns (U,M,R).

I started based on this posting: http://www.speedsolving.com/forum/showpost.php?p=62266&postcount=8 I'd been looking for a low algorithm method to use to get other people into the sport so I started playing with this "simplified Roux" and it's been a blast.

I've tried a couple of true Roux solves, but man... Petrus or Roux, it doesn't matter, I frickin' SUCK at block building. I end up building my blocks with a partial cross and then F2L corners. Maybe when PJK gets back to town we'll have a meet and he can give me some tips.

 

[rubiksfriend]

It's OK. When I first learned about Roux (summer, 2 yrs. ago?) I was horrible. I couldn't build a block to save my life; however, you can get better by just playing around. Look at the same position from different angles, move the centers around, take apart a finished 1x2x3 or 1x2x2 to see which positions are easy. Take a basic position and change the corner or edge orientation, change the position of the pieces, etc. Also, I recommend trying to be color neutral (with lim. of course, i.e. always use white/yellow blocks as sides) and recognize easy patterns. I've been using Petrus the past few days and it helped me to set up edges while pairing up corner better. As for Waterman... CFOP got 5-9 seconds taken off just through algorithmic optimization. I'm not saying that could happen for Waterman (since his is already pretty optimized ) but popularity can increase times. CLL takes what, 2-3 seconds? Solve redges with simple sequences, the recognition for solving one redge and orienting midges is really easy, only <UMR> turns for half the method, it can be good (I think). I also have an idea on face building but more on that later. Hint: it involves Roux.

 

[waffle=ijm]

That's funny. I know all my OLL cases from Fridrich, but I've just started learning a simplified Roux and I find it really hard not to use my OLLs to solve some of the cases I come across.

I actually just memorized all the CLL algs. 40 CLL algs (24 if looking at Roux solution) isn't as bad as 57 OLL agls

 

[brunson]

It's OK. When I first learned about Roux (summer, 2 yrs. ago?) I was horrible. I couldn't build a block to save my life; however, you can get better by just playing around. Look at the same position from different angles, move the centers around, take apart a finished 1x2x3 or 1x2x2 to see which positions are easy. Take a basic position and change the corner or edge orientation, change the position of the pieces, etc.

Those are some really great tips. I'm definitely going to try taking apart some blocks, that's how I got a feel for my intuitive Fridrich F2L, slowly putting a pair in place, then backing them out, then putting them back, etc, etc. Of course, like everything, it all boils down to practice. It'll come with time, I'm sure.

Also, I recommend trying to be color neutral (with lim. of course, i.e. always use white/yellow blocks as sides) and recognize easy patterns. I've been using Petrus the past few days and it helped me to set up edges while pairing up corner better.

By your post count I can see you missed several of "The Great Color Neutrality Debates" of 2008. I'm color neutral. I'd love to see someone color neutral Fridrich get in the top 10 in 3x3, because until someone does there are going to be a bunch of holdouts that don't believe it's a benefit. And unfortunately all of them are hella faster than me.

As for Waterman... CFOP got 5-9 seconds taken off just through algorithmic optimization. I'm not saying that could happen for Waterman (since his is already pretty optimized ) but popularity can increase times. CLL takes what, 2-3 seconds? Solve redges with simple sequences, the recognition for solving one redge and orienting midges is really easy, only <UMR> turns for half the method, it can be good (I think). I also have an idea on face building but more on that later. Hint: it involves Roux.

I'd love to hear more.

Thanks for the awesome advice.

 

[brunson]

That's funny. I know all my OLL cases from Fridrich, but I've just started learning a simplified Roux and I find it really hard not to use my OLLs to solve some of the cases I come across.

I actually just memorized all the CLL algs. 40 CLL algs (24 if looking at Roux solution) isn't as bad as 57 OLL agls

Yeah, I don't count them that way. 57 cases, but reflections bring them down to something like 35. How may CLLs are there in Roux if you don't count symmetries?

Edit: And how many of them are valid COLLs?

 

[waffle=ijm]

http://grrroux.free.fr/method/Step_3.html here roux says that there are 48 without symmetries. But i use the 40 CLL algs from when you do ELL and CLL as the last layer solution from Guus Razoux-Schultz.