An analysis of cubing methods and times...

[waffle=ijm]

\ No one wants to be told that the method they've invested so much energy and time could possibly be inferior. But thanks again for your support

As an early adopter of Roux, I grew up in the cubing community with this "inferiority complex" but it doesn't stop me because I don't care about someone's opinion unlike a lot of people here.

Like Kirjava said, if that's how you define better so be it. Because it's a rather subjective word. This isn't an argument of methods per se. Rather is an argument of the definition of better.

That being said, I motion to move this to off topic because it has nothing to do with methods rather just the definition of "better"

 

[PenguinsDontFly]

Insult. The last resort of the ignorant and the intolerant. I expected better from you.

Not an insult: he is letting you know that you are displaying yourself in a manner that other people might think you are that, and that you should be more aware of how you come across to people.

 

[Escher]

5 pages of non-arguments and pissed off people since I last posted and still no reply...

Speedsolving.com users are epic trollbait it seems.

 

[Artic]

thankfully, the undesirable posts were dealt with, and now we can get back on topic.

I've reconsidered some excellent points people have brought up, namely, my definition of the word "better". Perhaps that was too strong and subjective a word to use. Additionally, despite trying to remain as objective as possible by including only move count, rotation-less, and time into account, it's clear now that other "subjective" criteria must be taken into account when defining a global metric for "better". These include lookahead and move ergonomics to name a few.

Can we instead use the word "favorable", meaning that certain methods create conditions more favorable to producing faster times. Taken from the dictionary, "favorable" here means "encouraging, inclined to help or support". In this context, since CFOP is more heavily researched than other methods, would it be fair to say that CFOP is more favorable than ZZ, Petrus, or Roux?

 

[tseitsei]

In this context, since CFOP is more heavily researched than other methods, would it be fair to say that CFOP is more favorable than ZZ, Petrus, or Roux?

That's more like it :tu

I agree that it's "easier" to learn cfop and be fast at cfop since it is the most developed and explored method. Altough I still think that roux has equal (or even greater) potential to be fast. People just haven't developed/researched it that much yet...

 

[PenguinsDontFly]

End the thread here before someone disagrees. We seem to have found a solution that cfop is easier/favourable not better than roux zz because it is more popular and easier to get help with.

 

[Ranzha]

Insult. The last resort of the ignorant and the intolerant. I expected better from you.

Insult. The last resort of the ignorant and the intolerant. I didn't expect better from you.

 

[Artic]

So an often mentioned issue people have brought up in discussion over methods has been "research", and the abundance of research that has been put into CFOP. Therefore, I'm curious, what is the state of the art research being performed on the other methods?

1) In CFOP we have things like WV, SV, ZBLL etc that have added additional performance value over the standard CFOP. Using these new improvements or strategies, decreases in time have been achieved, and indeed, getting below an average of 10 seconds seems to require them.

2) In ZZ, the latest research that I'm aware of is the ZZ-Edges First method mentioned here: https://www.speedsolving.com/forum/showthread.php?50964-ZZ-Edges-First-discussion&p=1046130#post1046130 From the discussion, it seems that it might prove to be an improvement over the standard ZZ, and thus provide some speed benefit.

3) For Roux, I'm not aware of any research being currently performed for this method. I read the Roux discussion thread regularly, and I've also tried to investigate if anything new or worthwhile has been proposed, but there does seem to be anything. No real growth in the method has occurred, and the standard, F2B, CMLL, LSE is what people have continued to use.

Is there any new research being performed for Roux? Or has the research for Roux been maxed out?

 

[pewpewrawr]

roux is already perfect so we don't have to research anything

 

[GuRoux]

Is there any new research being performed for Roux? Or has the research for Roux been maxed out?

there's not any real good way to add huge "alg sets" apart from kcll, and affecting lse with different cmll. other than that is optimizing intuitive parts. of course getting better at blockbuilding first two blocks. 4a and 4b optimization. nmblocks and nmcmll. misorientated centers. UF and UB instead of UR and UL. 4b and 4c at the same time.

 

[DeeDubb]

3) For Roux, I'm not aware of any research being currently performed for this method. I read the Roux discussion thread regularly, and I've also tried to investigate if anything new or worthwhile has been proposed, but there does seem to be anything. No real growth in the method has occurred, and the standard, F2B, CMLL, LSE is what people have continued to use.

Is there any new research being performed for Roux? Or has the research for Roux been maxed out?

There's a lot of new stuff that can be worked on. Not many people are though. If Roux had a CFOP-size base, i think there definitely would be people focusing on some new things. Here's some examples where we can still grow.

For first two blocks:

So much of Roux block building is intuitive, so it's a little hard to expand block building research on a global level. What is happening now is a lot of collective sharing. GuRoux, Waffo, myself, and others are putting together walkthrough solve videos to help expand our own knowledge and others as well.

CN discussion is also a big element of Roux because the benefit of x/y2 or y/x2 CN vs. a fixed block makes quite a big difference. My average first block went from 14 moves or so with fixed blocks to 8-10 with y/x2 CN.

For CMLL:

Knowing multiple CMLL algs for cases is something that very few really do, but it's something that I believe could have a big benefit. Knowing one alg that preserves EO and another alg that disturbs it would help you predict and control the first part of L6E. There's something there.

There's also NonMatching CMLL, where you have two different x rotations of your left and right blocks. This could help you reduce your second block move count drastically, but making recognition on CMLL very difficult. I don't think many people have fully invested in this yet.

For L6E:

It may be possible to learn algorithms to control and even place UR/UL edges while doing EO. I know Alex knows many of these intuitively, but I'm not sure if anything has been compiled yet.

Misoriented centers is another thing that some people are playing with, where you build around the wrong centers to save moves and fix it at the end.



Another thing I would like to see if people investing in block building methods for 4x4 and up, such as K4 and Meyer.

So I do believe Roux has a TON of growing to do. It's not perfect, and it isn't even close to being fully pursued.

 

[waffle=ijm]

There's plenty of potential steps in Roux that can be combined and therefore improved. Over the years plenty of ideas popped up, most are dismissed because they are sub-par to a "standard" Roux solve.

I remember at one point there was a discussion about doing WV to orient the corners. Then using PLLs permute corners and UL/UR. But it was dismissed because all we were doing was applying what we knew from CFOP onto Roux. It had nothing to do with Roux. In order for the method to advance we needed something that was intrinsic to Roux, which is why researching potentially new ways to improve Roux is hard. Application is easy, new concepts are not.

Take KCLL for example. It is CMLL while doing EO or changing it to something easier. No other method uses CMLL or LSE. This is intrinsic to Roux, this alg set is advances the method in terms of potentially new ways of improvement. It is not sub-par to standard and therefore not easily dismissed. "How come no one is using it or published algs?" one might ask. It's really not hard to derive these types of algs if you're clever enough. So instead of straight up just telling people, we'd rather encourage people to figure it out on their own. Documentation is nice, but really dumbs people down. For the clever ones, KCLL is assimilated into regular solves so you don't even notice it. No one talks because there's no need.

I'd rather be asked about concepts rather than just saying the alg for this case is x.

So to answer the question has the research for Roux been maxed out? The answer is no, because it's stupid to yell out ideas that doesn't make sense. Revelations that are intrinsic to Roux aren't as common and will take more users and even more time.

 

[PenguinsDontFly]

There's a lot of new stuff that can be worked on. Not many people are though. If Roux had a CFOP-size base, i think there definitely would be people focusing on some new things. Here's some examples where we can still grow.

For first two blocks:

So much of Roux block building is intuitive, so it's a little hard to expand block building research on a global level. What is happening now is a lot of collective sharing. GuRoux, Waffo, myself, and others are putting together walkthrough solve videos to help expand our own knowledge and others as well.

CN discussion is also a big element of Roux because the benefit of x/y2 or y/x2 CN vs. a fixed block makes quite a big difference. My average first block went from 14 moves or so with fixed blocks to 8-10 with y/x2 CN.

For CMLL:

Knowing multiple CMLL algs for cases is something that very few really do, but it's something that I believe could have a big benefit. Knowing one alg that preserves EO and another alg that disturbs it would help you predict and control the first part of L6E. There's something there.

There's also NonMatching CMLL, where you have two different x rotations of your left and right blocks. This could help you reduce your second block move count drastically, but making recognition on CMLL very difficult. I don't think many people have fully invested in this yet.

For L6E:

It may be possible to learn algorithms to control and even place UR/UL edges while doing EO. I know Alex knows many of these intuitively, but I'm not sure if anything has been compiled yet.

Misoriented centers is another thing that some people are playing with, where you build around the wrong centers to save moves and fix it at the end.



Another thing I would like to see if people investing in block building methods for 4x4 and up, such as K4 and Meyer.

So I do believe Roux has a TON of growing to do. It's not perfect, and it isn't even close to being fully pursued.

Nice post. I use meyer y neutral but I suck lol

 

[mark49152]

...it was dismissed because all we were doing was applying what we knew from CFOP onto Roux. It had nothing to do with Roux. In order for the method to advance we needed something that was intrinsic to Roux

Why is it a problem to take things from CFOP and apply them to Roux, if they are beneficial? That's purist to the point of being self-defeating.

"How come no one is using it or published algs?" one might ask. It's really not hard to derive these types of algs if you're clever enough. So instead of straight up just telling people, we'd rather encourage people to figure it out on their own. Documentation is nice, but really dumbs people down.

Perhaps if you share your innovations it would promote interest in the method as well as encouraging other innovations. That clever guy could be working on something new, or just practising and getting faster, instead of reinventing the wheel by rediscovering your secret algs.

 

[Ranzha]

Perhaps if you share your innovations it would promote interest in the method as well as encouraging other innovations. That clever guy could be working on something new, or just practising and getting faster, instead of reinventing the wheel by rediscovering your secret algs.

People like being spoonfed.
People don't always make proper associations when they are spoonfed.
People don't ask why enough.

I hate it.

 

[mark49152]

People like being spoonfed.
People don't always make proper associations when they are spoonfed.
People don't ask why enough.

That's their problem, and of course not everyone is like that. I can't see how withholding a useful alg set is the best option for the community or the development of the method.

 

[waffle=ijm]

Why is it a problem to take things from CFOP and apply them to Roux, if they are beneficial? That's purist to the point of being self-defeating.

Maybe intrinsic came off too strongly (apologizes if it did). Let me rephrase it, if I may.

I don't have a problem if they are beneficial. It's just CFOP algs are not optimized for Roux and to me will never be. Simply put, CFOP has pieces to preserve and Roux does not. It is for the same reason Roux users use CMLL rather than COLL and Build blocks rather than F2L our F2B. We can easily do it. Like I said Application is easy. It's not purist rather a pursuit of optimization, both in speed and in move count. Non optimized concepts and algs hardly seem on par with what is already "standard."

If one were to optimize CFOP algs to Roux, then it would no long be usable on CFOP and rather intrinsic to Roux. Concepts are another thing but share the same sentiment as algs.

I'll talk about the WV to PLL, I mentioned. In short, if you go the new substeps/method thread, it's like weeding out what's good and what's bad ideas before they have time to grow. It goes back to the pursuit of optimization for a specific method.

WV alg preserve LL edge orientation. Why would I need that in roux? It's more moves than needed
Then I would have to EO.
Then I would have to use PLL if I didn't get CLL skip. Again more moves.

It's takes like at most 20 moves to complete "last slot" in second block and CMLL where as WV to PLL is probably more. But that's move count, what about speed. 2-gen CFOP stuff is fast, right?

Sure they are. But you have to account for setups in this case. for the WV part, we have to make a pair, place that LL, all while making sure 1x2x2 is also oriented properly (F2L style). Then you'd have to do the alg. Then you'd have to do EO, then You'd still have to position UL and UR pieces onto LL to use the PLL. It's fast to execute, but not fast when it comes to setup.

To me, WV to PLL in Roux is bad. Even worse than 2-Look CMLL. From my experience (and I hope I have some as I've been using the method for a really long time), any concept that deals with Last Slot, ends poorly if trying to applying to Roux.

Perhaps if you share your innovations it would promote interest in the method as well as encouraging other innovations. That clever guy could be working on something new, or just practising and getting faster, instead of reinventing the wheel by rediscovering your secret algs.

The thing is, it's not Secret. KCLL is documented (with some algs at that) and shared with everyone, conveniently on this link and if you haven't found it you're not trying hard enough. If you're not trying hard enough you shouldn't even consider learning KCLL. I'm not trying to conceal algs from people when I say clever people can figure it out. Quote from KCLL "it can be approached by learning 300+ algorithms, or next to none." In other words, if you're clever enough, you land on the second category.

Maybe you misunderstand when I say no one talks. It's because it's been out for 4 years. Some people have completely assimilated the concept of KCLL in their solves so it looks normal. It's like how CFOPers use ZBLLs in their solves (keeping in mind that full algorithmic KCLL is around 370 something algs). The concept is there, the algs are there. When I say no one talks its to avoid times like when a nub gets a PLL skip and goes "ZBLL ZBLL ZBLL" It cheapens the method, don't you think?

EDIT - When a new discovery is made, we post, we discuss, we deem if it's feasible. But right now the hard part is discovery. Not many people try to go that route. Making something out of nothing is hard.

 

[Ross The Boss]

A lot of the time no new CMLL algs need be learned to influence EO (and by that i mean looking up different algs online). You can experiment on certain cases by doing something like r instead of R, or counter slicing into your usual algorithm (e.g. R'+M' done in one fluid movement). Sure, there are times when it would be nice to know an extra alg or two for EO, but a lot of them can be written into muscle memory through repeated intuitive modifications to your normal alg. that -along with many other things- is what im trying to do right now.

 

[Kirjava]

I feel like I'm pretty full disclosure, you have no argument with "secret algs".

To accentuate waffle's suggestion, I've long felt embarrassed about putting my name towards that system because of how trivial it is to derive. I like to call it CLLEO now.