Online since 2002. Over 3300 puzzles, 2600 worldwide members, and 270,000 messages.

TwistyPuzzles.com Forum

It is currently Wed Jul 23, 2014 1:09 am

All times are UTC - 5 hours



Post new topic Reply to topic  [ 30 posts ] 
Author Message
 Post subject: New 3x3 Speedsolving Method
PostPosted: Thu Mar 15, 2007 5:21 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Hi all,

I've been thinking about this for a while, so I've decided to share my idea with you and see if you have any suggestions! This method is similar to Ryan Heise's Human Thistlethwaite Method up until a certain point...

1) Orient all edges (this step is easy, and I can explain my near-optimal method if you'd like)
2) Put E edges in E (layer between U and D) without affecting their orientation (i.e. using U D R L F2 B2 moves)
3) Orient all corners in 1-2 algs (you could do a 3/5 corner orientation approach (similar to Guimond, but the algs wouldn't affect edge orientation), a 4/4 corner orientation approach (almost like doing 2 OLLs with some setup moves), or something else)
4) Separate corners/edges into their appropriate layers (pretend you are solving 2 opposite centers on the 5x5)
5) There are many ways to finish off the cube:
A) Solve E, fix parity, do 2 PLLs
B) Solve E, solve U/D corners in 1 alg, solve U/D edges in 1 alg
C) Solve E, solve U/D edges in 1 alg, solve U/D corners in 1 alg

If anyone has comments, questions, or suggestions, please feel free to post them!!! If people would like algs for each step, I'd be happy to find those.

Thanks!

I would also like to add that there are already pure intuitive methods, and part intuitive part algorithm methods, but there are no (almost) pure algorithm methods. This method would fit into the third category unless you would like to do the first few steps without algs.


Top
 Profile  
 
 Post subject: new method
PostPosted: Thu Mar 15, 2007 7:52 pm 
Offline

Joined: Tue Mar 04, 2003 11:17 am
Forget it.


Last edited by David J on Sun Mar 18, 2007 3:33 pm, edited 1 time in total.

Top
 Profile  
 
 Post subject:
PostPosted: Thu Mar 15, 2007 8:08 pm 
Offline
User avatar

Joined: Sat Aug 12, 2006 6:40 pm
Location: California
Have you tried the Human Thistlthwaite?
http://www.ryanheise.com/cube/human_thistlethwaite_algorithm.html

_________________
www.garron.us
Nothing takes time from expanding your knowledge like doing your homework and applying to college...


Top
 Profile  
 
 Post subject:
PostPosted: Thu Mar 15, 2007 8:18 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Yes, it is similar to my method. However, his method eventually gets the cube into a state where only double turns are needed to solve it, which seems rather slow to me... That's why I stop when the cube gets into a domino type state (although the final algs would not necessarily keep it in this state), which is easier to work with than only double turns in my opinion. A domino state also makes cases very easy to recognize!


Top
 Profile  
 
 Post subject:
PostPosted: Thu Mar 15, 2007 10:40 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
It would be great if you could explain your near-optimal method for stage 1. I'm interested.

Have you considered using a strategy like Pochmann's for the Domino stage?

Personally, I'd try something like this (after you orient all edges, put E edges in E, and orient all corners):
- Solve all first-layer edges.
- Solve all first-layer corners.
- Fix parity and solve E layer.
- PLL.
It's not fully algorithmic, but it would be interesting to compare it to your method.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Fri Mar 16, 2007 6:38 am 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Here's my near-optimal method for step 1:
I will consider edges to be oriented if:

1) If they are U/D edges and are on U/D, then the U/D part of the sticker should be on U/D

1) If they are U/D edges and are on F/B, then the U/D part of the sticker should be on R/L

1) If they are F/B edges and are on U/D, then the F/B part of the sticker should NOT be on U/D

1) If they are F/B edges and are on F/B, then the F/B part of the sticker should be on F/B

Using this logic, any single R/L turn would flip the 4 edges on that face.

First, get 4 flipped edges (using U D F B R2 L2) onto either R or L, then simply do R/L.

If there are 4 more (or more) flipped edges, do the same thing.

If there are 2 edges remaining, do the following:
Get one of them onto R/L, and one of them NOT on R/L.
Do R/L, then "replace" the newly correct edge with the other flipped edge.
Then, do R/L again.
Example: R F R would flip the DR and UF edges

After all edges are oriented in this manner, then do y/y' (cube rotation in direction of U/U'), and continue solving.

If anyone needs more details or has questions, I'd be happy to answer them!


Top
 Profile  
 
 Post subject:
PostPosted: Fri Mar 16, 2007 9:20 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Hey everyone,

I just thought of something that might make this method a lot easier in the beginning (and easier to recognize/solve).

For the new step 1, get all E edges onto E, and make sure they're oriented.
For the new step 2, you could either use a variation of Gilles Roux's edge orientation method (M' U M'), and repeat that until all the edges are oriented, or you could easily generate MU algs or regular algs.
Then you could finish the solve as described before.

Any thoughts?


Top
 Profile  
 
 Post subject:
PostPosted: Fri Mar 16, 2007 11:15 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
I calculated that one of the best averages with this method would be around 17 seconds. Needless to say that this is not what I would have hoped for. To make this method faster, some steps have to be changed/combined/switched. Any ideas?

I think the concept of this method could be very fast but it definitely needs some improvements right now.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Mar 17, 2007 3:22 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
Where do you get 17 seconds from?


For your new Step 2, there are 20 cases for orientation, most of which will be quick, so you can probably do it in one algorithm which will probably take a serious speedcuber under 2 seconds each. Remember, though, you'd want them to be UMD-group, since you are going to orient all U/D edges in one step.

In total, the new step 1 and step 2 should probably take about 4 seconds.

But then we have to orient all of the corners, which may take some time... I'm not sure how fast this can be done, but I think I remember seeing somewhere that you can use one of 35 algs to orient all top-layer corners, do an x2, and then use one of 7 algs to orient all top-layer corners again. I think that this will probably end up taking about 3 seconds.

After that it's Domino + E-layer. It would probably be better to do E-layer last, since that can be done very quickly even if everything else is solved (~1 second). So the question is now: what is the fastest way to solve the oriented U and D layers? It should have the potential to be sub-8.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Mar 17, 2007 3:28 pm 
Offline
User avatar

Joined: Tue May 30, 2006 10:53 pm
Location: FL
An intermediate method is 15-20 seconds with this.That is like saying Fridrich should be sub 8 because you don't factor in recognition and difficult cases.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Mar 17, 2007 6:05 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
No, I mean that the domino stage should probably average under 8 seconds. I guess that the other steps take 8 seconds total, so if you want the solve to be under 16 seconds (say) then that particular stage should have sub-8 potential.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Sat Mar 17, 2007 11:48 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
OK, I've been thinking and here's my idea for an optimization of the method:

1) Get all E-layer edges oriented and placed in E layer.
2) Orient all corners.
3) Orient all edges in U and D layer (1 alg).
4) Solve first layer Domino-style.
5) Solve E layer and fix parity on U layer at the same time (1 alg).
6) PLL (1 alg).

Notes:
I think that you could combine steps 1 and 2 somehow, but it would take some practice and I'm not sure quite how to do it. I'm doing step 2 before step 3 because using a UMD-group algorithm (for orienting edges) does not affect corner orientation and it's easier to orient corners when you don't care about the edges. It may even be easier to orient corners before getting E-layer edges oriented. For step 4, I'd solve edges one at a time and then corners one at a time, but there's probably a faster way. I think that step 5 has as few as 6 nonsolved cases, which can be solved by (M2 E M2 u), (R2 U2 R2 U2 R2), (R2 u2 R2 u2), (R2 u R2 u'), (R2 u' R2 u), (R2 u U R2 u2 R2 u). Step 6 is standard PLL.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 2:44 am 
Offline
User avatar

Joined: Tue Jan 10, 2006 11:15 am
Location: Helsinki, Finland
watermelon wrote:
For the new step 1, get all E edges onto E, and make sure they're oriented.
For the new step 2, you could either use a variation of Gilles Roux's edge orientation method (M' U M'), and repeat that until all the edges are oriented, or you could easily generate MU algs or regular algs.

This can't be a good idea... Orienting just 2 edges takes way too many moves using UDM. And orienting all edges first in fact makes getting all E edges onto E easier because you don't need to care about their orientation. I can't see the point in changing the order.

Orienting corners is not a nice step, but using tons of algs it maybe could be done fast...

I have never tried speedsolving domino, but Petrus seems to work very well. You'll just skip some steps.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 11:40 am 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
qqwref wrote:
1) Get all E-layer edges oriented and placed in E layer.
2) Orient all corners.
3) Orient all edges in U and D layer (1 alg).
4) Solve first layer Domino-style.
5) Solve E layer and fix parity on U layer at the same time (1 alg).
6) PLL (1 alg).


That seems like it could be very good, and I also like Johannes's suggestion of Petrus.

If you solved F2L Petrus-style, you could correct the parity and combine steps 4 and 5 all at once!

Here's the up to date method:
1) Orient all edges and get all E-layer edges into E
2) Orient all corners using 1-2 algs
3) Solve F2L Petrus-style
4) PLL

Fortunately, although step 3 is completely intuitive, recognition will be very easy. Step 1 is also intuitive, but it can easily be planned during pre-inspection.

Is there anything else that could be improved with this method?


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 12:26 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Johannes Laire wrote:
Orienting corners is not a nice step, but using tons of algs it maybe could be done fast...

Well... here's what I'm thinking. Get 3 (or 4) corners oriented and on D. A simple sune will work for this if this step is not already done.

Then, use a quick RU alg to orient the rest. There wouldn't be too many algs and this step would be quite fast.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 2:13 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
For orienting corners, here's an idea:
- 1 alg to get at least 3 corners oriented on D (<10 algs)
- 1 alg to orient last D corner and orient all U corners (~35 algs).

OK, but here's my question: how do we orient all edges and get all E-layer edges onto E quickly?


Here's an idea for orienting corners and edges:
1) Orient all corners in some way. For example, do the first two steps of Guimond on the corners. This is a very fast step.
2) Hold the correctly oriented corners so that the color you've oriented them towards (say white/yellow) is on R/L. Solve the R/L centers, and then put the R/L edges into the R/L layers with the right orientation. Finally, orient the middle layer edges.
3) Solve F2L Petrus-style.
4) PLL.

The first two steps are like corners-first solving, but much more color-neutral.

Here's an algorithm for orienting middle-layer edges: U M' U M' U2 M U M U orients the edges on DB and UB (and moves things around a little).

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 2:31 pm 
Offline
User avatar

Joined: Tue Jan 10, 2006 11:15 am
Location: Helsinki, Finland
qqwref wrote:
OK, but here's my question: how do we orient all edges and get all E-layer edges onto E quickly?

That's easy. You just need to know that moving R or L 90 degrees flips those 4 edges. ~5 moves on average is reasonable, it takes some practice of course but not much. Getting E-edges onto E is very trivial, it takes just a few moves.

qqwref wrote:
2) Hold the correctly oriented corners so that the color you've oriented them towards (say white/yellow) is on R/L. Solve the R/L centers, and then put the R/L edges into the R/L layers with the right orientation. Finally, orient the middle layer edges.

I'd like to get some more details about this step, I find it very hard to do effectively.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 2:37 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
@qqwref

I believe we had the same idea for orienting corners :).

However, the method you just suggested would use a lot of slice moves, wouldn't it? I agree with Johannes that the 2nd step seems very difficult...

To answer your question:

Orient all edges using the method I described in an earlier post, then simply use a few UDRL moves to get the E edges onto E. Easy to recognize, and very few moves.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 3:05 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Just had an idea to slightly combine steps 3 and 4 of this method:

1) Orient all edges and get all E-layer edges into E
2) Orient all corners using 1-2 algs
3) Solve F2L Petrus-style
4) PLL

Maybe you could solve LL corners while solving F2L? Does anyone know how to do this?

Then you would be left with an easy EP for LL.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 4:15 pm 
Offline

Joined: Thu Nov 23, 2006 6:02 pm
Have you actually used the method?

_________________
3x3: PB 9.64
http://www.xanga.com/j_ey


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 4:23 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Well... I have done some slow solves but no timed solves yet. I will probably start speedsolving with this method once it is finalized/optimized...


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 9:18 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
Johannes Laire wrote:
That's easy. You just need to know that moving R or L 90 degrees flips those 4 edges. ~5 moves on average is reasonable, it takes some practice of course but not much. Getting E-edges onto E is very trivial, it takes just a few moves.


True. It would take some practice to be able to orient the edges quickly (especially if there are an odd number), but in the end it shouldn't be any more difficult than a cross or a 2x2x2. About the E edges, I see now that it's trivial after you've oriented everything. Good point.

Johannes Laire wrote:
I'd like to get some more details about this step, I find it very hard to do effectively.


I think you're right... it was an idea I had, but it doesn't seem to be fast, since there's simply too much to do.

watermelon wrote:
Maybe you could solve LL corners while solving F2L? Does anyone know how to do this?

Then you would be left with an easy EP for LL.


I don't think it will be at all easy to do that. Since you're doing F2L mostly intuitively, there's no uniform last step, and so solving the corners while finishing the F2L would be cumbersome. I'd just go with the PLL, which is only slightly slower than an edge-permutation anyway.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Mar 18, 2007 10:33 pm 
Offline
User avatar

Joined: Sun Dec 31, 2006 4:42 pm
Hmm... I just realized that unless you have at least 3 corners oriented every time after step 1, the move count for this method would be way too high. Does anyone know of a way to orient a few corners while doing step 1? Maybe you could do a quick R U R' or R U' R' before you insert the last E edge into E?


Top
 Profile  
 
 Post subject:
PostPosted: Mon Mar 19, 2007 3:43 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
Well, you could use a working-corner approach (before inserting the last E edge, solve 3 corners with RUR' type moves). Other than that, though, you're just going to have to use an algorithm.

Do you guys know any short algorithms to twist a single corner on D that will keep all edges oriented and do any twists on the top? I thought F'RUR'U'F would work, but that flipped two edges.

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Mon Mar 19, 2007 6:42 pm 
Offline

Joined: Sat Feb 17, 2007 4:19 pm
Location: USA
qqwref wrote:
Well, you could use a working-corner approach (before inserting the last E edge, solve 3 corners with RUR' type moves). Other than that, though, you're just going to have to use an algorithm.

Do you guys know any short algorithms to twist a single corner on D that will keep all edges oriented and do any twists on the top? I thought F'RUR'U'F would work, but that flipped two edges.


This entire discussion is mostly way over my head, but what about Sune? Does that preserve this definition of edge orientation?

_________________
2x2 single 4.89, avg 15.05
3x3 single 17.39 avg 26.66
4x4 single 1:32, avg 1:56
5x5 single 3:19, avg 3:50
6x6 12:00, 7x7 47:51, 8x8 34:10, 9x9 43:28
megaminx 4:57
To do: get faster.


Top
 Profile  
 
 Post subject:
PostPosted: Mon Mar 19, 2007 6:49 pm 
Offline
User avatar

Joined: Sun Feb 26, 2006 12:40 pm
Location: Marske-By-The-Sea, UK
The Sune done as ( R U R' U R U2 R') rotates the UFR UBR UBL corners once clockwise, but does not affect edge orientation.
It also cycles UF-UB-UL, while still leaving the corners in the same permutation and then rotates U twice,
(that should be right. (This can also be explained better.))

_________________
List of Speedcubing methods
Speedcubing tutorial

@.=split(//,"J huhesartc kPaeenrro,lt");do{print$.[$_];$_=($_+3)%25;}while($_!=0);


Top
 Profile  
 
 Post subject:
PostPosted: Mon Mar 19, 2007 6:56 pm 
Offline

Joined: Sat Feb 17, 2007 4:19 pm
Location: USA
Yes, you'd need to do it on F and R, or use a y/y' setup move. But then you'd cycle three edges on F, preserving orientation, twist 2 non-relevant U corners, and twist one relevant D corner.

_________________
2x2 single 4.89, avg 15.05
3x3 single 17.39 avg 26.66
4x4 single 1:32, avg 1:56
5x5 single 3:19, avg 3:50
6x6 12:00, 7x7 47:51, 8x8 34:10, 9x9 43:28
megaminx 4:57
To do: get faster.


Top
 Profile  
 
 Post subject:
PostPosted: Tue Mar 20, 2007 9:50 am 
Offline
User avatar

Joined: Tue Jan 10, 2006 11:15 am
Location: Helsinki, Finland
qqwref wrote:
Do you guys know any short algorithms to twist a single corner on D that will keep all edges oriented and do any twists on the top?

I think I learned this from Ryan's Human Thistlethwaite Algorithm: R' D L2 D' R.

Very clever indeed.


Top
 Profile  
 
 Post subject:
PostPosted: Tue Mar 20, 2007 6:46 pm 
Offline

Joined: Sat Jan 22, 2005 12:12 pm
Location: NY, USA
Clever! I like it!

Another way you could do the same kind of thing is L' U R2 U' L. There's also its mirror, L U' R2 U L'. So a corners strategy could be:
1) Solve at least 3 corners on D
2) Solve at least 3 corners on U
3) Fix last two corners
Each step is 1 algorithm. Step 3 has a 1/3 chance to be skipped, so it's not lucky if you don't have to do it.


EDIT:
Here are some more options:
L (R U R' U R U2 R' U2) L' [on D layer, only effect is twist of DRB corner]
R U' L U R' U' L'

_________________
My official times
My youtube
Puzzle Solving Service! - a puzzle that has never been scrambled and solved has been wasted.


Top
 Profile  
 
 Post subject:
PostPosted: Tue Mar 20, 2007 11:04 pm 
Offline
User avatar

Joined: Tue Mar 13, 2007 11:24 pm
Location: MN
its all to confusing... i think i'll just learn fridrich before anything else

_________________
Image


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 30 posts ] 

All times are UTC - 5 hours


Who is online

Users browsing this forum: No registered users and 4 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  

Forum powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group