Solving 4x4 on your own

I solved the 4x4 on my own back in the 80's. My first approach was to try layer by layer, but that gave me trouble. Then I realized myself that I could do centers and edges and solve it like a 3x3. That worked pretty well. Then I ran into a couple weird parity snags that I eventually figured out on my own.
 
Well, I paired up all the edges and was inserting the u layer centers. I was expecting to have to use the corner parity fixing algorithm that Chris gave me.
On the last layer I realised I just needed to cycle 3 edge pairs.

THEN IT WAS SOLVED!!!!!

I've never solved a 3x3 intuitively let alone 4x4 intuitively, actually, I did once experience an OLL and PLL skip.

Thanks alot everyone, and the next time I run into a corner permutation parity, I will remember what you told me Chris.
R2 Uw2 R2 Uw2 R2 Uw2.
 
Congrats! I have to say I have not yet solved a twisty puzzle on my own the first time. I'd like to. I'm trying to find a Rubik's UFO around here somewhere...
Anyway, I'm glad you did that and I think you will be too- once you start doing things like that you get a much better understanding of the cube and puzzles in general and all kinds of possibilities open up for you.
 
Can you scramble and resolve centers using an odd number of inner slice quarter turns? It's ok if this messes up some other things on the cube, as long as centers end up solved again.

Hint: It only takes 9 overall turns to do this

Chris

I tried my best. Couldn't come up with a way to do it in 9 quarter turns. Can you tell how.

However, I found that it is incredibly difficult to do this during the inspection time. After much practice I am down to maybe 25 seconds record for a single solve. However, there is a shorter way to do this than counting cycles of the 24 pieces, and more like counting the 12 cycles of a 3x3x3. I can explain this method if anyone is interested. Based on how fast the truly world class BLD people are at memorzing an entire 3x3x3, perhaps it is not so far fetched to count cycles using my shortcut pseudo-3x3x3 edge method to count cycles on a 4x4x4 and always avoid OLL parity.

Chris

Really interested in knowing how you find out by counting just 12 cycles instead of 24. Can you please explain.

cmhardw; said:
You have to consider what kinds of centers are cycled, meaning from which orbits. Remember that from a theory standpoint there are two kinds of centers on a supercube (and yes Maarten is right that any conclusion we're drawing here is really only valid on a supercube). When you turn an inner slice on a 4x4x4 cube you have done two 4-cycles of pieces, but you have to consider which center orbit those 4 cycles occur in.

Sorry for my limited understanding of what an orbit is. When you said there are 2 kinds of centers, did you mean

Kind1: Centers that lie on the middle slice
Kind2: Centers that are not of Kind1

or

Kind1: Centers that lie on the geometric diagonal of the cube's face
Kind2: Centers that are not of Kind1

or

Kind1: The center-most center
Kind2: Centers that are not of Kind1

And if you don't mind can you tell what an orbit actually refers to.

Thanks a ton.
 
I tried my best. Couldn't come up with a way to do it in 9 quarter turns. Can you tell how.

I don't think it can be done in 9 quarter turns. You can do this using 9 turns in face turn metric, but make sure that you use an odd number of turns in quarter turn metric.

Really interested in knowing how you find out by counting just 12 cycles instead of 24. Can you please explain.

What I do is sort of like memorizing a 3x3x3 cube blindfolded, so it will require counting only 12 "cycle" rather than the full 24. Basically you don't remember how many transpositions it takes to reduce the edges to the solved state, but rather to a state where edge groups are paired together, and oriented correctly in a 3x3x3 blindfolded orient first sense. For example:

Scramble your 4x4x4 cube with white on top and green on front with the following scramble:
Rw B2 D R2 B Uw2 B2 R B U' L' Uw2 Rw D' F2 D R2 Uw' B' Fw2 D' F2 Rw' R2 D2 F2 L' U2 L D B L2 R U R U L' B' F2 D2

If I was using my cycle counting method on this scramble I would do exactly the following.

1) Orient the cube using the cube rotation x, or turn the entire cube as if you were doing the turn R. This will place the paired yellow-red piece such that the yellow stickers are on the U layer (which is where yellow should go for me in order for an edge to be "oriented" in a blindfold solving sense).

2) I will look at the piece in UBl and see which piece it is. It is the orange yellow wing edge. I now find the other yellow orange wing, which is at RBd. The yellow is on the B face, which is where yellow should be in my 3x3x3 edge blindfold orientation scheme to be oriented "correctly". So I will imagine that I am sending the piece at UBl to RBu, which will pair it up with the other yellow orange and make an oriented 3x3x3 edge pair. Now I don't memorize this, nor do I count how many cycles there are. All I do is think that I just did one cycle, so I toggle the parity of the permutation I am analyzing. When beginning I have imagined that I have done no cycles at all, and I now imagine that I will swap my UBl edge with the one at RBu, which is one transposition. So now the parity is odd, and I verbally say "odd"

3) The piece that is in RBu is the yellow green edge, and it's pair mate is at DFr with the yellow on the D layer. This is already oriented, so I send RBu next to it to DFl, one transposition which toggles parity. I now say "even" meaning the permutation I am creating has even parity.

4) The piece that was at DFl is green red, and it's mate is at DBl. This piece is "flipped" in a blindfolded sense for 3x3x3 edge orientation because the green is on the B face. So now I will do two transpositions. I imagine that I am sending the piece at DFl to DBl, and the piece at DBl goes to DBr. Since I have done two swaps in a row, I have not toggled parity. So I don't say anything, and the permutation I am creating is still "even" as it was at the end of the last step. Now, when actually doing this, if edges are flipped I simply don't do anything and just move on, but the reason is because I am actually doing two transpositions.

5) The piece at DBr goes to UBl to pair with it's mate (which is where I started). This piece was already sent to UBl when I did the swap earlier. So this tells me that I completed a partial cycle and I must start at a new piece somewhere else. The parity of the permutation is still "even"

6) I now need to start from a new location, because I completed a cycle. I will now start at URb. This piece is white orange, and pairs with the other one at DRb. The DRb piece is flipped in a 3x3x3 sense, so I will do two transpositions, or I will not toggle parity. Therefore I simply don't say or do anything and just skip along to the next piece. My permutation parity is still "even" as it was at the end of the last step.

7) DRf pairs with UFr, which is flipped. Again do nothing and skip to the next piece.

8) UFl goes to LBd. I say "odd"

9) LBd goes to LFu. I say "even"

10) LFu goes to LRu goes to LRd. I say nothing and move on.

11) LRd goes to DLb goes to DLf. I say nothing and move on.

12) DLf is the piece that pairs with URf (the place adjacent to where I started). So I know the parity of the wing edge permutation is the last word I said, so the parity of the wing permutation is even. This means that the wing edges on the cube are in an even permutation, and I must solve the centers using an even number of quarter turns to restore the parity of the wing edges to even. This will effectively skip the OLL parity case, and I will know that it is impossible to have the OLL parity during the solve, so I save time.

13-optional) You may wonder how I would know that I am done at step 12. Back when I was trying to do this seriously I used a base 5 counting system with my feet to count how many edge pairs I had built during my analysis. The reason I used base 5 was because it allows you to count to 24 on two feet, and this way I could also use it when counting the full 24 cycles (which I did for a while before coming up with the idea I just described).

To verify that the wing edges actually are in an even permutation go ahead and just count the cycles in the usual sense. After the initial cube rotation, here is the wing permutation using my blindfolded lettering scheme.

(QFJBIG)(NPHOUXATCLV)(RK)(DSW) and two wing edges are solved. If you analyze the actual wing edge permutation you will find that it is indeed in an even permutation.

Sorry for my limited understanding of what an orbit is. When you said there are 2 kinds of centers, did you mean

I meant this option that you wrote:

Kind1: Centers that lie on the geometric diagonal of the cube's face
Kind2: Centers that are not of Kind1

Basically every inner slice contains 8 x-centers from the same orbit, and all the rest of the centers are in groups of 4 oblique centers from the same orbits.

And if you don't mind can you tell what an orbit actually refers to.

Thanks a ton.

I wonder if we are using this in the correct sense, as I have heard some people say that cubers misuse this term in a pure mathematical sense. By orbit, I mean all possible locations on the cube that a certain piece may occupy. So for example the corner orbit consists of 8 locations that the corners may occupy. The wing edge orbit on a 4x4x4 consists of the 24 locations that the wing edges may occupy.

Chris
 
Hey,
I have a 4x4x4 cube and haven't been spending much time on it. I haven't bothered to learn a method and I play around with it every now and then. So far I can intuitively solve the first 3 layers, all corners, but havn't quite figured out how to orient and/or permute individual last layer edges.
The closest I've got to it being solved was all but 2 pieces solved, they were oriented correctly but needed to switch places.

I was wondering who here managed to solve the 4x4 all by themselves?
If so, how long did it take you?
And how fast can you solve it now? (perhaps even with your own method)

Took me one week to figure out the parities that was in the 80's, after I thought about it -I had some knowledge about permutation- the solution was just one turn away

I solve the 4x4 now in about 1:18
 
I tried my best. Couldn't come up with a way to do it in 9 quarter turns. Can you tell how.

I don't think it can be done in 9 quarter turns. You can do this using 9 turns in face turn metric, but make sure that you use an odd number of turns in quarter turn metric.

Chris

u R2 u R2 u R2 u R2 u ?
I completely randomly got that. :eek:
Its actually nice, only 3 swaps of edges induced. Or to put it in other words one 3 cycle commutator away from parity case.

And nice way of finding the odd/even parity of dedges. I didn't really get how you planned to use toes to count base 5 numbers. Is it something like left toes for 5's place and right for 1s place? That sounds as a good idea actually. I will give it a try.

I feel like sharing my way of memoing. Tell me if I should change it.
Your (QFJBIG)(NPHOUXATCLV)(RK)(DSW)
would be FJBI GNPH OUXA TCLV NRKR DSWD for me with Q as my buffer.

1) My memo seq doesn't have a letter for buffer.
2) Once the first cycles close, the rest of the cycles will be starting and ending with the same letter.
3) I memo them in groups of four. (I find it better than 2 as I map these four letter words to some word with some vowels added. If the last group is a 1 or a 3 long one, then it is odd parity)
4) I find this easy because all I need to do is pick 2 letters, and cycle them with buffer. I don't have to think of things like when I need to break into a new cycle.

Shanmukh
 
Zane, if you really understand what you have done so far you should realise that your second problem is almost like the first. A few setup moves is all you need to transform between them. Me like others would like to express my doubt as to how you get get so far and not to be able overcome the final small hurdle ??

Per
 
I tried my best. Couldn't come up with a way to do it in 9 quarter turns. Can you tell how.

I don't think it can be done in 9 quarter turns. You can do this using 9 turns in face turn metric, but make sure that you use an odd number of turns in quarter turn metric.

Chris

u R2 u R2 u R2 u R2 u ?
I completely randomly got that. :eek:
Its actually nice, only 3 swaps of edges induced. Or to put it in other words one 3 cycle commutator away from parity case.

Shanmukh

Following is an observation when I did r U2 r U2 r U2 r U2 r
All this while I was thinking that the centers were intact and somehow I was not comfortable with the idea. On careful observation I found this:

Considering the four centers in each face as 2 pairs. An offset of one quarter turn is induced in the r slice pairs and There is an odd permutation in the pairs. The pairs on the U face end up swapped at the end of the algo.

So I don't know if we can say that it actually re-solves the centers. If you are okay with the odd permutation in centers then thats the best I can come up with. Let me know if there is some other algo is in your mind.

Shanmukh
 
My mistake for having put it that way. I was just considering using this for BLD odd dedge parity, which is when it bothered me because I usually solve dedges before centers.

In the current context, it answers the question. Thanks and apologies.
 
I was just considering using this for BLD odd dedge parity, which is when it bothered me because I usually solve dedges before centers.

As long as you only use this on a 4x4x4 you could use an alg like (R' U' R U')*5 to rotate the center back to it's original state. On a 5x5x5 you will end up also swapping some t-centers (and on a supercube you would have rotated the centralmost center 180 ;) ). For 4x4x4 though those swapped rows are equivalent to rotating the center 180 degrees. This way you can still solve dedges before centers.

Chris
 
As long as you only use this on a 4x4x4 you could use an alg like (R' U' R U')*5 to rotate the center back to it's original state.

Thanks for this algo. I always use the commutator [RLdw2R'L' U] and a U2

On a 5x5x5 you will end up also swapping some t-centers (and on a supercube you would have rotated the centralmost center 180 ;) ). For 4x4x4 though those swapped rows are equivalent to rotating the center 180 degrees. This way you can still solve dedges before centers.

Chris

So to use the same method in 5x5 super bld:
Goal: swap UBl <-> UFl and swap USl <-> USr (I meant the tedges)

1) edge + t-center parity fix:
r U2 r U2 r U2 r U2 r

2) cyle (DBr DFr UFl) depending on the situation

3) correct the other centers:
[rld2r'l' U] U2

I don't have 5x5 supercube and currently my windows is not working so can't check it on gabbasoft.
Correct me if I am wrong.

anyway this method would fail for 6x6 or higher due to the introduction of oblique centers(or whatever they are called)
I can't think of a good alg for the case of cubes higher than 5x5
 
Last edited:
This thread inspired me a while ago to get a 4x4x4, and to do it a good number of times on my own until I got to a decent enough speed (maybe somewhere around 2-8 minutes?). Just got my 4x4x4 in the mail the other day, and scrambled it up (by hand, I still don't really know the notation), and got it in 6 hours. It's very strange though, I've had it take me 16 hours to do it and one solve in about 9-10 minutes, maybe I'll start getting a little more precision in a few weeks.

I'm still having trouble with a few key things, mainly figuring out whether or not I need to fix parity or not (I'm a little lost reading through the heavily technical portions of this thread that seem to explain what I'm missing). Oh well, I'll probably figure it out sooner or later, and that may even be some of the fun (?).

Zane, did you end up getting better with your own intuitive solves or did you start to collect some methods and algs, or fully pick up a method or what? I'm curious!

Oh, and if anyone cares, I've been doing my best to follow the loose steps of firstly fixing centers, then pairing edges, then making some blocks and getting a F2L minus a slot style thing, trying my damndest to detect if I need to fix parity, trying to fix parity, and then checking if I actually needed to fix parity, at which point I just sort of push through to the end. Sloppy, right? :)
 
This thread inspired me a while ago to get a 4x4x4, and to do it a good number of times on my own until I got to a decent enough speed (maybe somewhere around 2-8 minutes?). Just got my 4x4x4 in the mail the other day, and scrambled it up (by hand, I still don't really know the notation), and got it in 6 hours. It's very strange though, I've had it take me 16 hours to do it and one solve in about 9-10 minutes, maybe I'll start getting a little more precision in a few weeks.

I'm still having trouble with a few key things, mainly figuring out whether or not I need to fix parity or not (I'm a little lost reading through the heavily technical portions of this thread that seem to explain what I'm missing). Oh well, I'll probably figure it out sooner or later, and that may even be some of the fun (?).

Zane, did you end up getting better with your own intuitive solves or did you start to collect some methods and algs, or fully pick up a method or what? I'm curious!

Oh, and if anyone cares, I've been doing my best to follow the loose steps of firstly fixing centers, then pairing edges, then making some blocks and getting a F2L minus a slot style thing, trying my damndest to detect if I need to fix parity, trying to fix parity, and then checking if I actually needed to fix parity, at which point I just sort of push through to the end. Sloppy, right? :)

I hope I don't dissapoint you, but I havn't really touched it since I've solved it. Some moves I used to pair the last layer edges was just a simple r U r' U r U2 r'.
I solved the effected centers without disturbing the edges or corners at all using intuitive moves such as, b l' b' U' b U l U' b'.
I don't literally perform it as (b l' b' U' b U l U' b'), because it is intuitive I use cube rotations to change the 'b' into say, an 'r' or an 'l'

I also experimented with some Fridrich PLL algorithms, by applying some inner slice set-up moves and a PLL alg then undoing the set-up moves, I was able to minipulate individual edge peices, but that didn't work out to well.

I'm glad this thread inspired you :D

Next time I order some more puzzles I might get a better 4x4x4 cube and learn reduction or something
 
Last edited:
I had to skim your post, like most of the posts in the last few pages of this thread, just for the sake of avoiding influence (stealing algorithms or something without knowing it), hope that doesn't make me a bad person...

And if I knew more algs, I'd probably use them, I only really know niklas and sune, and their inverses. No biggie that you haven't touched it since you solved it though (I've done the same with sq1). I've been trying a sort of 2 gen type thing (not sure if it counts as a 2 gen, but using only u/U r/R, as in leaving a 2x2x4 block), that might keep me interested if I do end up figuring it out.
 
i found it out on my own the first time ^_^
then i had parity every other time. . .grr. . .
 
What you have is a parity. And there's nothing you can do without an algorithm (unless, of course, you could fix it intuitively, which is close to impossible to do). If you wanna solve it on your own, I would rescramble it and hope for the best that you don't get a parity case.
its not impossible to fix it intuitively all you need is a brain.
 
Back
Top