Developing a better LL system

[Speedmaster]

Could anybody explain me this system, as easy as possible? All I know is, that there are lots of cases and you look at one and use two fast algs to solve one case and you don't have to learn the algs, but the combinations of them and the recognition. But there are so many cases, how is it possible to recognise all of them fast??!

 

[Zarxrax]

Just memorizing cases is much easier than memorizing a new alg for every case. 1000+ seems like a ridiculous number, but when every alg is easy and already familiar, (I'm assuming, I didn't look at every alg in the Misc section of the site), I don't see learning this as is as inherently difficult at all. Why? If I'm already really familiar with all algs involved, learning these cases is something I could easily do without a cube, which would tremendously speed up the process. If I wasn't busy memorizing parityOLL and stuff for 4x4, I'd start learning some of this myself right now.

Really? I've personally always found it to be MUCH more difficult to memorize the cases than it is to remember algorithms. But everyone is different I guess.

 

[bobthegiraffemonkey]

I'd like to eventually see this sorted by CLL. I had an idea before of learning this for bad COLL cases (I had ideas of EO hacks for big cubes, which kinda failed when I tried them, but I've not completely given up on it yet), but you could do this with bad CLL cases instead. Good CLL: CLL->ELL, bad CLL: this. I'm busy with other stuff so I don't see myself putting in any hard work on this unfortunately, if I had more free time I would seriously consider it.

 

[jeff081692]

Could anybody explain me this system, as easy as possible? All I know is, that there are lots of cases and you look at one and use two fast algs to solve one case and you don't have to learn the algs, but the combinations of them and the recognition. But there are so many cases, how is it possible to recognise all of them fast??!

Say for instance you decided to learn 1 look last layer. You would need to be able to recognize 1000+ cases and each case has a unique algorithm to learn so that is 1000+ algorithms. What this method does is makes it so you don't have to learn so many algorithms, instead all you would need to know to 1 look a case is to know which 2 algorithms to apply out of the small pool of fast algorithms. It is easier to see a case and recognize it as the combination of two known cases than to recall one alg out of 1000+.
But even this could be a bit extreme starting out. So to use this as a two look system all you would need to know when you arrive at a case is which is the first alg. Once you do the first alg, the cube will be in a familiar state (assuming you know the other algs in the pool) and you can finish the cube with the second alg.
That's the best explanation of the system I can come up with. Recognition is the only thing that needs to be learnt. It is definitely possible to recognize the difference between cases. The only way to get faster at that is practice. As people learn this system they will be able to give back information to the community that makes recognition easier.

 

[mDiPalma]

If all edges are oriented beforehand, exactly how many of these cases would then be required? 1/4 of the total?

 

[vd]

If all edges are oriented beforehand, exactly how many of these cases would then be required? 1/4 of the total?

1/8, if I am right?

 

[Speedmaster]

Say for instance you decided to learn 1 look last layer. You would need to be able to recognize 1000+ cases and each case has a unique algorithm to learn so that is 1000+ algorithms. What this method does is makes it so you don't have to learn so many algorithms, instead all you would need to know to 1 look a case is to know which 2 algorithms to apply out of the small pool of fast algorithms. It is easier to see a case and recognize it as the combination of two known cases than to recall one alg out of 1000+.
But even this could be a bit extreme starting out. So to use this as a two look system all you would need to know when you arrive at a case is which is the first alg. Once you do the first alg, the cube will be in a familiar state (assuming you know the other algs in the pool) and you can finish the cube with the second alg.
That's the best explanation of the system I can come up with. Recognition is the only thing that needs to be learnt. It is definitely possible to recognize the difference between cases. The only way to get faster at that is practice. As people learn this system they will be able to give back information to the community that makes recognition easier.

Thank you
Now I understand it ;D

 

[AHornbaker]

I've been following this thread for a while and have looked at the tables, but I'm not sure I understand it. What are the algs underneath each graphic in the full table/which algs do you choose from for the pool of algs?

 

[JasonK]

I've been following this thread for a while and have looked at the tables, but I'm not sure I understand it. What are the algs underneath each graphic in the full table/which algs do you choose from for the pool of algs?

Each of the "algs" underneath the case is a different way of solving that case. Everything between two > symbols is one "alg".

eg. The first case (H-Perm) can be solved with DoubleSune U' DoubleSune, or with TripleSune F'RD2R'FU2F'RD2R'FU2, or any of the other solutions in the list.

 

[AHornbaker]

So I'm assuming Kirjava is working on making a small pool of algs that can be used for every case? Without that the site is just a list of algs for every permutation of the last layer.

 

[Kirjava]

So I'm assuming Kirjava is working on making a small pool of algs that can be used for every case? Without that the site is just a list of algs for every permutation of the last layer.

I've already done this. There's an organised version of the system other than the full table.

Next step is to seperate into learnable subsets. At some point...

 

[elrog]

I know that you already wanted everyone to stop assuming you were not working on this, but people can't help asking because it is such a great idea. If used with a method that preorients edges such as ZZ or partial edge control used with each slot resulting in edges oriented by the time you reach the last slot, then you could use partial corner control with the last slot and only deal with cases with edges oriented and atleast one corner oriented correctly. This would further reduce cases and alg numbers. What I am wanting to say though, is that you could make cases like these that would be subsets of the whole thing first so that people can use these while you are generating the rest.

 

[redbeat0222]

Pretty gord. I really need the better Last Layer.

 

[Kirjava]

What I am wanting to say though, is that you could make cases like these that would be subsets of the whole thing first so that people can use these while you are generating the rest.

Yeah, I was planning on doing the "A" CLL cases subset first as a trial and seeing how it worked out after people had learned it. Then I could get the motivation to move onto other stuff.

It's just work that'll I'll do at some point.

 

[AHornbaker]

Here's a thought:
For recognition of all LL cases, you could use a modified 2 side PLL recognition. Not sure how well you would be able to recognize a PLL case (with any of 57 given orientations), but if someone went through all the cases and figured out how doable it is, it could be a workable solution. Someone with programming skills could map out all the LL cases, analyze each from the four 2-side recognition cases for each (showing U, F, and R faces), and see how many distinguishable LL cases can be recognized this way.

 

[elrog]

The permutation of the top layer is very easy to recognize so long as corners are oriented. Without corners being oriented, I think you'd need to see alteast 3 sides (not counting the top of course). Mapping them out would also be an enormous task. You do realise that you'd have to check every possible case with every possible AUF? (number of cases times 4)

 

[rj]

Once it becomes popular, then I will learn it--if it is less algorithms than fridrich.

 

[Speedmaster]

But that will never happen, just look at the number of algs now..

 

[aznanimedude]

Once it becomes popular, then I will learn it--if it is less algorithms than fridrich.

well you're in luck
it IS indeed less algorithms than friedrich, assuming you mean OLL+PLL.

and why not just be that person that makes it popular?

don't confuse # algorithms with # cases, #cases is mildly large, but the number of algorithms necessary to know is not

 

[mark49152]

The permutation of the top layer is very easy to recognize so long as corners are oriented. Without corners being oriented, I think you'd need to see alteast 3 sides (not counting the top of course). Mapping them out would also be an enormous task. You do realise that you'd have to check every possible case with every possible AUF? (number of cases times 4)

Two sides and top should be enough, because given all the information visible on those for each case, there's only one possible state for the information not visible.