[Help Thread] 4x4 Discussion

[Stefan]

at the moment, I don't think there is a better solution than just solving the parities during the LL stage.

Yeah, maybe. Though if you use Petrus and thus recognize the flipped edges already after the 2x2x3 block, you can handle the "OLL parity" right there already. At that point you have more freedom, so a shorter alg might be possible. Someone asked this a while ago in the twistypuzzles forum and I showed an alg I had developed earlier: http://www.twistypuzzles.com/forum/viewtopic.php?f=8&t=9502

Also, the typical reduction method might not be the best way to go, the 4x4x4 offers so much opportunity for other ways. Some ideas I might've mentioned before (not sure) and that fit into this thread:

Holey method:
Step 1: Solve centers
Step 2: Solve as 3x3x3, solving as many edges as you can and all corners. With edges I do *not* mean edge pairs. Just single 4x4x4 edges. So you solve the 4x4x4 except it'll appear to have unsolved holes of 4x4x4 edges.
Step 3: Solve the remaining edges, with the r2 blindsolving method or 3-cycles or whatever. Excellent look-ahead here. One kind of parity might occur, namely swapping two edges. Or if you recognize it earlier and are left with a 4-cycle, then (rU2)4.5 might be usable.

uLL method:
Somewhat similiar to normal reduction, except your last layer is not U but u. So the normal edge pieces play the corners in that layer, and the normal center pieces play the edges in that layer. Parity is treated by a quarter u-turn.

One way to begin the uLL method (or others) could be to solve the complete bottom half of the 4x4x4 first. For example start with the bottom center, then solve all eight pairs of D-edges and d-centers, then solve all four pairs of D-corners and d-edges.

Or many other ways. I'm sure we'll eventually have computers and programs capable of aiding in the development of 4x4x4 solutions like we can already do now for the 3x3x3, and more. Then we can seriously study very different 4x4x4 methods easily, and I wouldn't be surprised if we find something better than today's normal reduction.

Btw, when is it okay to use the word "fix"? Is it only okay when you talk about parities?

Nothing really wrong with how you used it, so if people understand you, it's alright. It's just unusual to say that there instead of "solve", so I wasn't sure.

 

[Robert-Y]

Ah about your holey method: I was thinking about something like that before because I remember (I think it was) Milan Baticz who posted something about sub-40 (maybe it was sub-45) being easy with a solve the centres, then the corners, and finally the edges method.

 

[Stefan]

Yes, that was probably Milan, he told me about that as well. I have no idea how he's solving the centers+corners, though I think he solves them simultaneously (at least that's the impression I got).

 

[AvGalen]

Yes, that was probably Milan, he told me about that as well. I have no idea how he's solving the centers+corners, though I think he solves them simultaneously (at least that's the impression I got).

I have never seen Milan do that in competition. Milan is one of the better 4x4x4 solvers in the world, but I don't see him doing 45 seconds solves with another method

I love the Holey method, simply because of the name

 

[byu]

I really want to try Holey method, but I can't because of 2 reasons:

1) My 4x4x4 is broken

2) I can't find a good tutorial for r2. Erik's the only one I know of, is not uploaded.

 

[qqwref]

The Holey method seems cool. I'm not a r2 user so I prefer to just do normal comms. Second solve was 1:28.96; I think consistent sub-1:20 is definitely doable with this but I'm not so sure about a sub-1 average. To be faster than reduction you would basically have to be able to do the last 12ish edge comms faster than normal edge pairing (plus half a PLL parity on average).

uLL, I'm not so sure. The last step seems like it could be cool but I can't imagine the algorithms are very nice, and I also think solving the other 3 layers would be a real pain in the ass, especially if you do the bottom two layers first (since to place the top layer you'd have to basically put every piece in with commutators, as there is no freedom of movement).

 

[Mike Hughey]

I had to try the Holey method. My first five tries:

4:47.14, 3:20.05, 2:52.85, 3:41.78, 3:09.26

I was generally finished with all but the "holes" in less than 2 minutes (except for the first solve - it's really confusing at first doing OLL and PLL!). I think this clearly shows why I'm still so much slower than I should be at 4x4x4 BLD. Maybe I should just switch to this for my 4x4x4 speedsolve method - I might get a lot better at BLD edges that way!

 

[Stefan]

2) I can't find a good tutorial for r2. Erik's the only one I know of, is not uploaded.

The link from my M2/R2 page still works:
http://erikku.er.funpic.org/rubik/r2.html

But really, that's just one way to do it, you could come up with something better, too.

 

[Stefan]

To be faster than reduction you would basically have to be able to do the last 12ish edge comms faster than normal edge pairing (plus half a PLL parity on average).

Execution is probably slower, but recognition is probably faster because you just know where to put pieces and where to look. For r2 more than for 3-cycles. I'm not good at twisting r2 but on the 3x3 I can do M2 sighted without pause because of perfect look-ahead.

There are other trade-offs, too. From some experiments a while back I'd say you usually get around nine edge pieces solved in the 3x3x3 step, not twelve. Because at the end you have to be lucky to have the right edges left and not already stuck in let's say the cross. On the other hand, when you solve a dedge, you have the choice between two so you can choose the easier of the two cases. For example if you build a white/red/blue F2L pair, you have the choice between two red/blue dedges.

So yeah, it has pros and cons and I don't know how good it is. Just mentioned it as an example for the many possibilities the 4x4 gives us.

Here's yet another I just remembered, btw:
1) Solve centers
2) Direct-solve as many edges as you can, treating the 4x4x4 as a 3x3x3. So similiar to Holey. Just don't solve corners, only edges.
3) Direct-solve the remaining edges in three-cycles. But not pure 3-cycles of edges. No, 3-cycles of edge-corner pairs. That's ok because the corners get solved later. And it allows wide turns rather than inner slice turns.
4) Solve corners

uLL, I'm not so sure. The last step seems like it could be cool but I can't imagine the algorithms are very nice, and I also think solving the other 3 layers would be a real pain in the ass, especially if you do the bottom two layers first (since to place the top layer you'd have to basically put every piece in with commutators, as there is no freedom of movement).

I must admit I didn't go much further than coming up with the idea. I just like the idea of u PLL. About the freedom of movement: You can also build the bottom half of the puzzle except one side, i.e. build a 2x3x4 block, then you have let's say <U,u,R> freedom. But again, mainly mentioned to say there's much more out there.

 

[Stefan]

Maybe I should just switch to this for my 4x4x4 speedsolve method - I might get a lot better at BLD edges that way!

Yeah, why not... Except for speedsolving you can improvise more. Especially you don't need a fixed buffer. When I do it and use r2 and have finished the cycle, I can just rotate the cube and continue from another buffer. Or switch to l2.

One improvement for Holey would be to solve the l-slice and r-slice edges as dedges earlier. So that you only end up with holes on L and R only, which are much nicer.

 

[Mike Hughey]

Maybe I should just switch to this for my 4x4x4 speedsolve method - I might get a lot better at BLD edges that way!

Yeah, why not... Except for speedsolving you can improvise more. Especially you don't need a fixed buffer. When I do it and use r2 and have finished the cycle, I can just rotate the cube and continue from another buffer. Or switch to l2.

Yeah, I thought of that, but I decided not to for my first five solves because I'm so used to looking for the pieces in my usual orientation. It would probably take me quite a few solves to get comfortable with rotating the cube - silly, but true. I did rotate once or twice on the last few pieces in those solves, though - by that point it's easy.

One improvement for Holey would be to solve the l-slice and r-slice edges as dedges earlier. So that you only end up with holes on L and R only, which are much nicer.

Yes, I can see where that might be a good idea, and it would really be pretty easy to do. Of course, that takes away from the benefit as BLD training, so maybe I wouldn't want to do that...

 

[CubeFreak122]

What I use for my 4x4 is the cage redux method, it's fast and I can get under 3 minutes with it. It has no parity and barely any algorithms and the long ones aren't as long as parity, go to youtube and search cage redux method and look for the one by jonowl. I'm going to make another tutorial for it but I can't until my computer stops overheating. The method is just like the 5x5 method

 

[rachmaninovian]

whats wrong with parity? it's just a 15 move long algorithm!
yes i HATE it when I get it, and depending on which cage system you use you may get parity.
cage redux? xD

 

[qqwref]

Maybe by "cage redux" he means pairing up all the edges, solving the cage like a 3x3, and then doing all the centers? Sorry, though, but whatever the method is "under 3 minutes" isn't too impressive on 4x4... I can do that with pretty much any speed method, even LBL or block building. It's a long way from sub-1 in any case

 

[rachmaninovian]

Maybe by "cage redux" he means pairing up all the edges, solving the cage like a 3x3, and then doing all the centers? Sorry, though, but whatever the method is "under 3 minutes" isn't too impressive on 4x4... I can do that with pretty much any speed method, even LBL or block building. It's a long way from sub-1 in any case

lol go see the vid and you would understand xD

 

[TsColin]

hey i got lucky a few times and solved it w/o parity

 

[Three Days Grace Fan]

4X4 parity help

For some reason, i am not able to get the parity algorithms right. i use the correct algorithm and i do the algorithm right, but all it does is scramble the cube. help?

 

[miniGOINGS]

are you doing slice moves instead of double layer turns?

 

[Kyle™]

which parity?
I made this a million years ago for a friend http://www.youtube.com/watch?v=C5QCmEG7ZSM

 

[AndyK]

I suggest checking out my algorithm list for the 4x4x4. It is nice because it contains all the parity algs that you need in a convenient printable format. It also has a complete notation guide and nice pictures of all possible parity configurations.

You can check it out here: http://kungfoomanchu.110mb.com/guides/andy-klise-4x4x4-guide.pdf
Or read about it more in this thread