[Help Thread] 4x4 Discussion

[xyzzy]

Did an ao12 on six different methods on a whim:

Redux - 1:02.20
Yau - 1:07.12
Yau variant - 55.20 (cf. example solve)
Hoya - 1:05.17
Cage - 1:35.94
Sandwich - 1:20.40

I don't practise with Hoya at all, and somehow I still manage to get (marginally) better times with it than Yau. Maybe more surprisingly, I'm significantly faster with redux. This might be related to how much of F2C I manage to plan out; with redux I have 3 choices for F2C, versus 2 for Hoya and only 1 for Yau.

1:20 for sandwich is somewhat faster than when I was using it as my main method (1:30-ish in early 2016), although I do know full CLL now along with way more centre commutators. (I'm still somewhat amazed that I had a classmate who could sub-minute with sandwich, and with much worse hardware than we have today.)

 

[Piotr Grochowski]

I use reduction, but reduction is actually a beginner's method. First I solve the centers, quite easy, and then pair edges, using this algorithm: l' U L' U' l (lowercase is slice). Also, I make sure last 3 edges are placed properly so that they all get paired with this alg.

Then I solve the remaining 3x3 using beginner's method.
1. Cross
2. I use beginner's methods to solve F3L, taking advantage of free slots.
3. First step of 2-look OLL (I may use other algs to directly solve OLL cases that I know)
4. Second step of 2-look OLL
5. First step of 2-look PLL (I may use other algs to directly solve PLL cases that I know)
6. Second step of 2-look PLL

OLL parity: Rr U2 x Rr U2 Rr U2 Rr' U2 Ll U2 Rr' U2 Rr U2 Rr' U2 Rr'
PLL parity: Uu2 Rr2 U2 r2 U2 Rr2 Uu2

 

[T1_M0]

I use reduction, but reduction is actually a beginner's method. First I solve the centers, quite easy, and then pair edges, using this algorithm: l' U L' U' l (lowercase is slice). Also, I make sure last 3 edges are placed properly so that they all get paired with this alg.

Then I solve the remaining 3x3 using beginner's method.
1. Cross
2. I use beginner's methods to solve F3L, taking advantage of free slots.
3. First step of 2-look OLL (I may use other algs to directly solve OLL cases that I know)
4. Second step of 2-look OLL
5. First step of 2-look PLL (I may use other algs to directly solve PLL cases that I know)
6. Second step of 2-look PLL

OLL parity: Rr U2 x Rr U2 Rr U2 Rr' U2 Ll U2 Rr' U2 Rr U2 Rr' U2 Rr'
PLL parity: Uu2 Rr2 U2 r2 U2 Rr2 Uu2

Hmmm I don't think it's necessary to do oll and pll after F3L

 

[Piotr Grochowski]

Hmmm I don't think it's necessary to do oll and pll after F3L

1. It's 2-look OLL and 2-look PLL, not OLL and PLL. These are very different things; the full OLL and PLL is for speedsolvers, and 2-look OLL and 2-look PLL are for intermediate solvers.
2. On 4x4, there are 4 layers. So solving F3L leaves one more layer to solve, and I use 2-look OLL, 2-look PLL and parity algorithms for that.

 

[greentgoatgal]

Hmmm I don't think it's necessary to do oll and pll after F3L

After the first 3 layers of the 4x4 yes it is.

 

[SavageCuber]

Hey everyone. My name is Saad. I'm into speedcubing since more than a year now and I average around 16 seconds on a 3x3. On a 4x4, my yauduction gets completed around 55 seconds but I average around 1:20. I'm practicing pretty hard and learning new things but can't get faster. Honestly speaking, I really like the 4x4. Any tips?

 

[xyzzy]

You average 16 seconds on 3×3×3 but spend 25 seconds on F2L+LL stage? I'd say the easiest, most immediate improvement you can get is to figure out why you're so much slower there, and fix whatever you're doing wrong. As you finish your last 2/3 edges, you should already be looking for your first F2L pair. (For reference, I also average around 16 seconds on 3×3×3, and I estimate that I spend around 20 seconds (including OLL/PLL parity) for F2L+LL.)

That said, there's also plenty of time to shave off in the rest of your solve. For the first two centres, always plan the first centre completely. (Even if you're fixed cross, you still have two colours to choose from for the first centre, and usually there's one choice that doesn't completely suck.)

Cross dedges are just a matter of not wasting moves (see J Perm's video). Don't do the "solve one edge wrongly" thing; it forces you to have to think about your colour scheme after you finish your centres, when you should be looking ahead for pieces.

I'm pretty bad at last-four-centres, so I can't give useful advice here, but it seems that doing half-centres is pretty popular: solve three bars on the F, D and B faces (in whatever order, but make sure the colours are correct relative to each other), then use Rw and U moves to finish off centres. Also, look ahead to the last cross edge as you finish, and prepare for a z' rotation.

Edge pairing is all about fluidity and not pausing, even when you get "weird" cases. 3-2-3 is the most common "style", as it greatly reduces the likelihood of getting weird cases compared to 6-2, while being only slightly less efficient with normal cases.

 

[SavageCuber]

You average 16 seconds on 3×3×3 but spend 25 seconds on F2L+LL stage? I'd say the easiest, most immediate improvement you can get is to figure out why you're so much slower there, and fix whatever you're doing wrong. As you finish your last 2/3 edges, you should already be looking for your first F2L pair. (For reference, I also average around 16 seconds on 3×3×3, and I estimate that I spend around 20 seconds (including OLL/PLL parity) for F2L+LL.)

That said, there's also plenty of time to shave off in the rest of your solve. For the first two centres, always plan the first centre completely. (Even if you're fixed cross, you still have two colours to choose from for the first centre, and usually there's one choice that doesn't completely suck.)

Cross dedges are just a matter of not wasting moves (see J Perm's video). Don't do the "solve one edge wrongly" thing; it forces you to have to think about your colour scheme after you finish your centres, when you should be looking ahead for pieces.

I'm pretty bad at last-four-centres, so I can't give useful advice here, but it seems that doing half-centres is pretty popular: solve three bars on the F, D and B faces (in whatever order, but make sure the colours are correct relative to each other), then use Rw and U moves to finish off centres. Also, look ahead to the last cross edge as you finish, and prepare for a z' rotation.

Edge pairing is all about fluidity and not pausing, even when you get "weird" cases. 3-2-3 is the most common "style", as it greatly reduces the likelihood of getting weird cases compared to 6-2, while being only slightly less efficient with normal cases.

Thanks. This was pretty useful. I'm pretty good at last 4 centres. The only place I'm bad at is when I get oll or pll parity. That's where I get bad timings

 

[Piotr Grochowski]

I solve centers completely different than everyone. Note that the last four centers are sideways. Lowercase means slice or wide turn, it doesn't matter.

Hope the 3-face, 4-face and 5-face views were all clear.

 

[Oliver1010]

If there is 1 edge oriented on the last layer, is there an algorithm to orient the 3 remaining edges?

 

[xyzzy]

If there is 1 edge oriented on the last layer, is there an algorithm to orient the 3 remaining edges?

Don't bump ancient threads, and also B' R' singleparity R B. (3-flip OLL parity isn't worth it unless you know COLL, imo, and even then it's not that useful.)

 

[Piotr Grochowski]

A shorter PLL alg than r2 R2 U2 r2 R2 u2 r2 R2 u2 U2 (at least in OBTM/WTM): u2 r2 U2 r2 R2 U2 r2 u2 (not bumping)

 

[cuber314159]

how many parity OLL cases are there?

 

[Lid]

how many parity OLL cases are there?

54

 

[Piotr Grochowski]

Don't bump ancient threads, and also B' R' singleparity R B. (3-flip OLL parity isn't worth it unless you know COLL, imo, and even then it's not that useful.)

Ancient?

 

[Oliver1010]

Don't bump ancient threads, and also B' R' singleparity R B. (3-flip OLL parity isn't worth it unless you know COLL, imo, and even then it's not that useful.)

I'm relatively new, so I haven't seen the ancient threads
I know COLL
And is there an alg, even if it isn't worth it?

 

[Piotr Grochowski]

I'm relatively new, so I haven't seen the ancient threads
I know COLL
And is there an alg, even if it isn't worth it?

Didn't expect you to know Corners Of Last Layer!

If the correct edge is on the right side, the alg is B' R' (OLL parity) R B. Of course, if you only know 2-look OLL, like me, you won't get the "not worth it" stuff.

Also, if you're curious, adjacent edge swap (front and right) is R2 D' x (PLL parity) F R2 x', though it's useless if you use edges first last layer.

 

[greentgoatgal]

Ancient?

Oliver1010 originally posted on an ancient thread. But then the posts got moved to a more appropriate location.

 

[Oliver1010]

Didn't expect you to know Corners Of Last Layer!

If the correct edge is on the right side, the alg is B' R' (OLL parity) R B. Of course, if you only know 2-look OLL, like me, you won't get the "not worth it" stuff.

Also, if you're curious, adjacent edge swap (front and right) is R2 D' x (PLL parity) F R2 x', though it's useless if you use edges first last layer.

Thanks

 

[xyzzy]

And is there an alg, even if it isn't worth it?

People like Feliks and Mats probably know full COLL and I don't think I've ever seen them use a 3-flip alg before, though. (I do use it, but only to increase the chances of getting ZBLL.)