The New Method / Substep / Concept Idea Thread

[mDiPalma]

This applies to a lot of stuff.

JK guys, don't hunt me down.

Yeah lol. What if I posted in this thread:

1) solve the first layer but don't solve 2 adjacent corners and the edge in between.
2) solve the other layer across the cube in the same way. hold the cube so these two layers are on the Bottom Left and Bottom Right. I call this the "anti-equator step" because you are leaving a belt of pieces unsolved in the middle. But don't worry, the more pieces we leave unsolved in the middle of the cube, the better.
3) solve the corners of the cube with a 3-gen algorithm that will make any decent speedcuber throw up
4) solve the rest of the cube with intuition and the most awkward moveset I can think of [M,U]. This will take a lot of moves. This movecount will also have to be multiplied by 1.5 because I picked a horrible moveset


you'd think it was a silly approach.

that's because it is.

 

[Chenkar]

Yeah lol. What if I posted in this thread:

1) solve the first layer but don't solve 2 adjacent corners and the edge in between.
2) solve the other layer across the cube in the same way. hold the cube so these two layers are on the Bottom Left and Bottom Right. I call this the "anti-equator step" because you are leaving a belt of pieces unsolved in the middle. But don't worry, the more pieces we leave unsolved in the middle of the cube, the better.
3) solve the corners of the cube with a 3-gen algorithm that will make any decent speedcuber throw up
4) solve the rest of the cube with intuition and the most awkward moveset I can think of [M,U]. This will take a lot of moves. This movecount will also have to be multiplied by 1.5 because I picked a horrible moveset


you'd think it was a silly approach.

that's because it is.

Lol I had to reread it to understand where you were going. But it did make me consider going back to that method.

 

[mDiPalma]

holy crap! MrMan had a really good idea. I modified it slightly

1) EOLine,

2) insert 3x [E-slice edges with random D layer corners in random orientations] and 1x [F2L pair] (make sure the last edge+corner you insert is with the speed-heise algs here. The last insertion must be with these algs, but doesn't have to be the true F2L pair), note that the D layer corners must be in an exactly solvable state (in that their orientations and permutations are not affected by any other corners on the cube)

EDIT: you must also insert the cross edges, per standard ZZF2L procedure. I'm sorry; I thought that was obvious, my b

3) 2 distinct commutators

step 2 will take some getting used to, particularly in the last 3 F2L inserts

omg that's actually a fantastic method. props to MrMan.

i say we call it ZZ-Belt for the lols

 

[TDM]

holy crap! MrMan had a really good idea. I modified it slightly

This does sound like it could be good, but won't recognition be absolutely terrible for step 2?
E: actually, no it won't... you just need to orient the third "random" corner to make a possible case. So actually it wouldn't be that bad at all.
I've been more busy than I expected and haven't remembered to learn speed heise yet. Also, I can't commutator. Have you tried using this method? How fast are you with it?

 

[Chenkar]

holy crap! MrMan had a really good idea. I modified it slightly

1) EOLine,

2) insert 3x [E-slice edges with random D layer corners in random orientations] and 1x [F2L pair] (make sure the last edge+corner you insert is with the speed-heise algs here. The last insertion must be with these algs, but doesn't have to be the true F2L pair), note that the D layer corners must be in an exactly solvable state (in that their orientations and permutations are not affected by any other corners on the cube)

3) 2 distinct commutators

step 2 will take some getting used to, particularly in the last 3 F2L inserts

omg that's actually a fantastic method. props to MrMan.

i say we call it ZZ-Belt for the lols

What do the commutators solve? The other d edges and 7 corners? Sorry I'm not understanding it all

 

[deadmanlsh]

holy crap! MrMan had a really good idea. I modified it slightly

1) EOLine,

2) insert 3x [E-slice edges with random D layer corners in random orientations] and 1x [F2L pair] (make sure the last edge+corner you insert is with the speed-heise algs here. The last insertion must be with these algs, but doesn't have to be the true F2L pair), note that the D layer corners must be in an exactly solvable state (in that their orientations and permutations are not affected by any other corners on the cube)

3) 2 distinct commutators

step 2 will take some getting used to, particularly in the last 3 F2L inserts

omg that's actually a fantastic method. props to MrMan.

i say we call it ZZ-Belt for the lols

But it isn't COLL you use for the last layer (wait, it's actually not the last layer, it's the top layer), it's L4C, which has more than 80 algorithms and really sucks in both efficiency and speed.

Imagine getting a pure-flip Sune on the bottom. *shivers*

 

[TDM]

What do the commutators solve? The other d edges and 7 corners? Sorry I'm not understanding it all

They solve corner 3-cycles on the D face and on the U face. You could also get 2/3 corner twists.

But it isn't COLL you use for the last layer (wait, it's actually not the last layer, it's the top layer), it's L4C, which has more than 80 algorithms and really sucks in both efficiency and speed.

Imagine getting a pure-flip Sune on the bottom. *shivers*

Not COLL, commutators. Also it's three corners, not four, which means a single commutator is usually fine, but you can sometimes get pure twists cases... they aren't too bad though. They can all be done 2-gen, and there are often even faster algs e.g. left sune + right sune.

 

[mDiPalma]

This does sound like it could be good, but won't recognition be absolutely terrible for step 2?
E: actually, no it won't... you just need to orient the third "random" corner to make a possible case. So actually it wouldn't be that bad at all.
I've been more busy than I expected and haven't remembered to learn speed heise yet. Also, I can't commutator. Have you tried using this method? How fast are you with it?

during step 2, you also have to make sure the isolated D-layer CP is feasible. so you have to pay attention to which corners you're placing after the first one.

so basically, ignoring completely the true F2L pair (which can be successfully inserted at any moment during the F2L), after you insert the first random corner:

1a) if the first F2L corner is in the correct location, but misoriented, you must also insert the next 2 random F2L corners into the correct spots.

1b) if the first F2L corner is not in its correct location, make sure you don't insert the opposite corner-edge "pair" together into the same slot. (ie don't have a 2-swap of D-layer corners). for example, if u insert the red-green corner with the red-blue edge, dont insert the red-blue corner with the red-green edge

2) the last random F2L corner has to be inserted to make the isolated D-layer CO possible.


this method seems very fast. i havent tried any solves yet, but I will later. i know all the speed-heise algs for DFR oriented down.

you don't even need to use "commutators". There are only 27 possible cases. 18 are 3-cycles, but most of those are just the standard OCLL alg. 8 are pure CO (sune pairs to solve these cases). and 1 is a skip. Honestly, if you know CFOP, you should have no problem with commutators with corners that are isolated to a single layer.

 

[mDiPalma]

What do the commutators solve? The other d edges and 7 corners? Sorry I'm not understanding it all

Yeah, after step 2, you should only have 6 corners left unsolved on the cube (max), which can be exactly solved with 2 unrelated commutators/conjugates.

But it isn't COLL you use for the last layer (wait, it's actually not the last layer, it's the top layer), it's L4C, which has more than 80 algorithms and really sucks in both efficiency and speed. Imagine getting a pure-flip Sune on the bottom. *shivers*

COLL typically affects the permutation of the edges. In this method, the second step gaurantees that the edges are already correctly solved, as well as 1 corner in each the U and D layers. We won't use 2x L4C cases, but instead we will use 2x L3C cases.

 

[Chenkar]

They solve corner 3-cycles on the D face and on the U face. You could also get 2/3 corner twists.

Ok... so when do you solve the RD, LD and U edges? I'd try it out now, but I don't have a cube to experiment with atm.

 

[mDiPalma]

Ok... so when do you solve the RD, LD and U edges? I'd try it out now, but I don't have a cube to experiment with atm.

ya i'm sorry. that's totally my B. I forgot to write that in the original method description (i thought it was obvious). I updated the original description

The LD and RD edges are solved during the F2L. so you apply the standard ZZF2L blockbuilding strategy to insert these pairs. (or you can just do EOLine, followed by inserting the cross-edges. whatever floats ur boat)

The U layer edges are solved with the speed-heise algs that I linked to in the original method description. So during the last insert, you apply one of those 24 algs. and once you're done, you will magically have everything solved on the entire cube, except for a commutator on the U layer, and a commutator on the D layer.

 

[Chenkar]

ya i'm sorry. that's totally my B. I forgot to write that in the original method description (i thought it was obvious). I updated the original description

The LD and RD edges are solved during the F2L. so you apply the standard ZZF2L blockbuilding strategy to insert these pairs. (or you can just do EOLine, followed by inserting the cross-edges. whatever floats ur boat)

The U layer edges are solved with the speed-heise algs that I linked to in the original method description. So during the last insert, you apply one of those 24 algs. and once you're done, you will magically have everything solved on the entire cube, except for a commutator on the U layer, and a commutator on the D layer.

Okay makes more sense now. Thanks

 

[TDM]

during step 2, you also have to make sure the isolated D-layer CP is feasible. so you have to pay attention to which corners you're placing after the first one.
so basically, ignoring completely the true F2L pair (which can be successfully inserted at any moment during the F2L), after you insert the first random corner:

1a) if the first F2L corner is in the correct location, but misoriented, you must also insert the next 2 random F2L corners into the correct spots.
1b) if the first F2L corner is not in its correct location, make sure you don't insert the opposite corner-edge "pair" together into the same slot. (ie don't have a 2-swap of D-layer corners). for example, if u insert the red-green corner with the red-blue edge, dont insert the red-blue corner with the red-green edge
2) the last random F2L corner has to be inserted to make the isolated D-layer CO possible.

this method seems very fast. i havent tried any solves yet, but I will later. i know all the speed-heise algs for DFR oriented down.

So, here's how I would do it then (all of these corner-edge pairs exclude FR, obviously):
1. insert edge+random corner
2. insert edge+specific corner with random orientation
3. insert edge+only remaining corner with specific orientation
So you only need to really focus on CP for 2, and CO for 3. That isn't too bad... for step 2, there's only two D layer corners left that aren't DFR, so that's not hard. Step 3 is the "hard" bit, making sure it's oriented correctly, but with a bit of practise that shouldn't be very hard at all.

you don't even need to use "commutators". There are only 27 possible cases. 18 are 3-cycles, but most of those are just the standard OCLL alg. 8 are pure CO (sune pairs to solve these cases). and 1 is a skip. Honestly, if you know CFOP, you should have no problem with commutators with corners that are isolated to a single layer.

Yeah, thinking about it now, I think there's only one case I can't 1-look (the T case [x R2: [R U2 R', D2]] and its mirror), but that isn't a very hard one to learn.

Ok... so when do you solve the RD, LD and U edges? I'd try it out now, but I don't have a cube to experiment with atm.

Step 2 is like ZZ F2L, so when mDiPalma was talking about inserting 'pairs', he meant inserting the pairs and DL/DR at the same time as the pairs. The speed heise bit solves U layer edges.

 

[mDiPalma]

So you only need to really focus on CP for 2, and CO for 3. That isn't too bad... for step 2, there's only two D layer corners left that aren't DFR, so that's not hard. Step 3 is the "hard" bit, making sure it's oriented correctly, but with a bit of practise that shouldn't be very hard at all.

yeah this is actually a really good ZZ variant. think of all the moves you'll save during F2L. The only expense is a z2/x2 rotation followed by a quick 8 move "alg" at the end of your solve. i actually think this method is better than standard ZZ, which says a lot.

it might be an even better idea, statistically, to assure that the corners you insert during F2L are not permuted correctly. that would ease the recognition a bit.

another great advantage is that you basically only get 3 move inserts, up until the last "pair". And you get 4 chances for an easy ACTUAL F2L pair.

not to mention the endless possibilities for FMC.

this deserves its own thread. i'll write one when I get off work

inb4 sub8 avg5 and WR single by Porkynator.

so do we call it ZZ-Belt, ZZ-Edges First, ZZ-F, ZZ-rulez, ZZ-Z or what??? NEED IDEAS PLS

 

[TDM]

Spoiler: Long post

yeah this is actually a really good ZZ variant. think of all the moves you'll save during F2L. The only expense is a z2/x2 rotation followed by a quick 8 move "alg" at the end of your solve. i actually think this method is better than standard ZZ, which says a lot.

it might be an even better idea, statistically, to assure that the corners you insert during F2L are not permuted correctly. that would ease the recognition a bit.

another great advantage is that you basically only get 3 move inserts, up until the last "pair". And you get 4 chances for an easy ACTUAL F2L pair.

not to mention the endless possibilities for FMC.

this deserves its own thread. i'll write one when I get off work

inb4 sub8 avg5 and WR single by Porkynator.

so do we call it ZZ-Belt, ZZ-Edges First, ZZ-F, ZZ-rulez, ZZ-Z or what??? NEED IDEAS PLS

Do you have any idea roughly how many moves this method would take on average?

I would call it ZZ-edges first, or more likely I would call it ZZ-EF.

 

[TheOneOnTheLeft]

The only expense is a z2/x2 rotation followed by a quick 8 move "alg" at the end of your solve.

I don't see why one couldn't predict the D case based on watching the three unsolved corners as you place them, or that in combination with looking at the case while executing the speed-heise insertion, and/or while executing the U-layer comm. If it is feasible to recognise the D case at one of these points, then the x2/z2 rotation isn't needed and we could just look for nice algs that directly solve L3C on D. Most likely some U/D mirrors of L3C on U will be nice for this anyway.

 

[mDiPalma]

Do you have any idea roughly how many moves this method would take on average?

because you can do the legit F2L pair whenever you want, the movecount would be consistently pretty low.

EOLine 7
square 4
pair 3
legit F2L square/pair 5
set up last pair 5
speed heise+aufs 10
commutators 17

looks like around 50 htm for a speedsolve. ill do a closer analysis later

I don't see why one couldn't predict the D case based on watching the three unsolved corners as you place them, or that in combination with looking at the case while executing the speed-heise insertion, and/or while executing the U-layer comm. If it is feasible to recognise the D case at one of these points, then the x2/z2 rotation isn't needed and we could just look for nice algs that directly solve L3C on D. Most likely some U/D mirrors of L3C on U will be nice for this anyway.

true. but frankly i don't think it's worth it to learn 20+ new algs from a new orientation. I think it's good enough to just pre-recognize the L3C case on D, and just hop right into it with the right grips once you finish the L3C on U.

 

[TDM]

because you can do the legit F2L pair whenever you want

I don't understand why you can do this; if you do F2L pair + U layer stuff and then do other pairs, surely that would affect the U layer? Unless you can do speedheise for all four slots? Or were you thinking of putting in the actual pair anywhere, and doing speedheise on any random pair in FR?

 

[MrMan]

I think we should create the thread for it and add the info in the first post while talking about it, it is not really clear now.
And for me ZZ-EH is nice !

 

[mDiPalma]

I don't understand why you can do this; if you do F2L pair + U layer stuff and then do other pairs, surely that would affect the U layer? Unless you can do speedheise for all four slots? Or were you thinking of putting in the actual pair anywhere, and doing speedheise on any random pair in FR?

you dont have to do speed-heise on the legit pair. as long as you do speed-heise on the last slot (be it in BL or FL or anywhere, with a random F2L corner in a random orientation that is paired with the correct F2L edge), this solution approach will still work.

the only requirement of speed-heise is that it be performed on the last slot (it does not matter where that slot is, nor the relevance of the F2L corner involved.

does that make sense?