RoFL method

[scottishcuber]

Yeah. I guess with EG each set contains the same LL orientations which means recognition and familiarity aren't affected, but with this there are far more sets with unique orientations.

And I think I was mixing up recall with recognition. I recall algs in inspection if they are recently learnt and the solve will be just as fast because there is no recog. time.

 

[KongShou]

this is a cool method, but too many alg

too many!

 

[Stefan]

To clarify again maybe: The idea was *not* to learn all ~700 algs of all 18 sets. The idea was to learn the 43 algs for "the one best" set of the 18 sets. Of course you *can* combine several, like EG does, but you don't have to, and the number of sets can be low.

Seems like recog would be a pain for this method.

Don't know how much recog of the first layer is affected, but if you meant the second layer, you can just pick a RoFL case where the second layer isn't "misoriented", i.e., where the cases are the exact same as in normal CLL, so recog would be the exact same (have a look at its HTM and QTM pages). As mentioned on the bottom of my page, my case FDL_FRD is the best of the 18 (saving on average about 0.86 moves during CLL).

43 * (18+3)

= 903 - 1 = 902, whoops.

That doesn't make sense. First of all, 43 * (18+3) = 903 - 1 is false.
Also, EG-0 and EG-1 are already included, so you only have to add EG-2.
Also, only-DFR-misoriented and only-DLF-misoriented are equivalent.

43 * (18+1-2) - 1 = 730

this is a cool method, but too many alg

too many!

It's only one more than CLL, which I believe many people use.

 

[KongShou]

i see, but surely we have decided that EG1 and 2 are the best sets?

also what is the point of naming all 18 cases and finding algs for them then?

 

[Stefan]

i see, but surely we have decided that EG1 and 2 are the best sets?

Define "best". Movecount-wise, EG-0 (CLL) is the worst among the 18 in my list, EG-1 is the second-worst, and I believe EG-2 is about as bad as EG-0 (from vague memories of my full RoFL study).

also what is the point of naming all 18 cases and finding algs for them then?

Well, in order to possibly find the best, of course I have to analyse all of them.

If you meant to ask why I *showed* all of them:

• To present all results of the analysis, so we can see which cases are how good (this also allows to judge combinations of sets).
• Pure optimal HTM/QTM counts don't necessarily reflect how well they work in practice (after optimizing algs for finger-friendliness), so the best sets judged that way might not actually be the best in practice.
• There isn't "the one best" set, because:
  • The top two (RDF_FDL and FRD_LFD) have the lowest movecount, but leave the second layer misoriented, which people might not like.
  • Those two cases also lose against DLF_RDF and LFD_DFR, which have better best cases (allowing 4-moves last layers, might be optimal for people interested in non-lucky single records).
  • In the full study (up to all four pieces rotten), there's no case that wins both HTM and QTM.

Edit: Maybe I shouldn't have called this thread "RoFL method" but rather "RoFL 2x2x2" like the webpage, as it's not really a fixed method (at least I haven't picked one) but more of an analysis which can lead to different actual methods. And if eventually the one and only actual "RoFL method" does get picked, I should probably rename the thread to "ROFLMAO" (ROtten First Layer Method And Others (or I have to get Tyson or Toby to use RoFL (yes, I know, I'm trying too hard now))).

 

[KongShou]

Define "best". Movecount-wise, EG-0 is the worst among the 18 in my list, EG-1 is the second-worst, and I believe EG-2 is really bad, too (from vague memories of my full RoFL study).



Well, in order to possibly find the best, of course I have to analyse all of them.

If you meant to ask why I *showed* all of them:

• To present all results of the analysis, so we can see which cases are how good (this also allows to judge combinations of sets).
• Pure optimal HTM/QTM counts don't necessarily reflect how well they work in practice (after optimizing algs for finger-friendliness), so the best sets judged that way might not actually be the best in practice.
• There isn't "the one best" set, because:
  • The top two (RDF_FDL and FRD_LFD) have the lowest movecount, but leave the second layer misoriented, which people might not like.
  • Those two cases also lose against DLF_RDF and LFD_DFR, which have better best cases (allowing 4-moves last layers, might be optimal for people interested in non-lucky single records).
  • In the full study (up to all four pieces rotten), there's no case that wins both HTM and QTM.

really interesting, this might be better than EG

 

[elrog]

I once had a similar idea, but never thought to only use a few of the sets and just choose the best ones. To find the "best" set though, you would not only have to get good algorithms for all of the cases, but see which ones are easier to form the first layer. There may be some sets that cover bad cases in another sets first layer making them work well together despite the actual algs.

 

[Stefan]

Something that just occurred to me: If intentionally "missolving" the first layer more and more allows better and better average move counts to solve the rest in one step, what if I go to the extreme of "missolving": not solving anything at all. Just leave the cube scrambled. The average move count needed to solve a random cube is about 8.756 (using Jaap's table). That's not much worse than the 8.463 after fully solving one layer, which I think didn't count AUFs. So solving one layer on average only gets you slightly closer to a solved cube, or maybe (if you do count AUFs) actually further away from a solved cube. Weird.

 

[scottishcuber]

Weird indeed.

I guess the premise is that partially solving the cube (either a first face, layer or RoFL) greatly reduces the amount possible positions of the remaining corners.

There are 3,674,160 possible positions on a 2x2x2 cube and 730 cases for RoFL.

How many moves on average does it take to build any type of rotten FL?
In a previous post on this thread, you quoted the max number of moves required to make the rotten FL at 2. Let's go with that.

Basically, we reduce the cube to a RoFL case with 2 moves. In doing so we are using a method that lets us solve the cube after the rotten FL in an average of 7.769 moves QTM (I averaged the QTM column on the table on your website).
What if we were able to reduce the cube in a different way such that we were even closer to reaching 8.756? What would that method be? More radical than RoFL for sure, but what is it?

So instead of reducing the cube to 730 cases, this number would be different (certainly bigger)...the number of moves needed to reduce the cube would be less (less than 2 anyway), and on top of that ,each case after reduction will take even fewer moves to solve on average.

p.s. I'm not necessarily asking you all these questions Stefan, it's just what I'm thinking. Thats why it's a little disjointed as well lol.

 

[Stefan]

[post=911869]Stats for three rotten corners[/post] added in Chris Olson's TCLL thread.

 

[BlazingDragon]

Ok i really dont get this, like whats the difference between RDF and FRD? arent they the same thing? they are both right down front corners right? I mean switching the order of the right or down doesnt affect the orientation right? or have i been missing things the entire time?

Edit: also when you say DBL and DBR are solved do you mean just oriented or completely solved?

Edit2: Oh! for RDF do you mean the D holds the R color, the F holds the D color and the R holds the F color?

 

[Stefan]

I mean switching the order of the right or down doesnt affect the orientation right?

Um, that's precisely the one thing it does.

Edit: also when you say DBL and DBR are solved do you mean just oriented or completely solved?

When I say solved, I mean solved.

 

[Stefan]

Edit2: Oh! for RDF do you mean the D holds the R color, the F holds the D color and the R holds the F color?

Yes. Because I don't just say "RDF", I say "DFR holds RDF" (the first being a position, the second being a piece).

It's also well-known in blindcubing, maybe you've seen people say stuff like "shoot to FLU" or "shoot to FLU" and actually meaning the F sticker position there (the "LU" following it makes clear which of the eight possible F sticker positions is meant). "Shoot to LUF" thus means a different orientation.