[Unofficial] 5x5 - 2:11.80 direct solving :D

[Dene]

Well, we don't directly solve the pieces do we? In 'direct-solving' you solve the edges to where they need to go. But we solve the edges wherever, to make it look like a 3x3, which we then solve..

This is a bad reason.
I'm going to use 4x4x4 as a more suitable example as it doesn't have fixed centers. If I solve the edges first, do they all have to be relative to each other for it to be "direct"?
When solving a 3x3x3, doing the OLL actually breaks up the rest of the puzzle, so the OLL is not a direct algorithm, by your standard.
In fact, no method could possibly be direct, as you are always going to have to move around pieces that are already in their correct positions to solve other ones.
In solving a puzzle, you are always making progress (if you are not, then you are not actually solving it, you are either getting further away from a solved state, or keeping it at the same distance from a solved state). Even if the piece is not in its correct position relative to another piece, you are still moving forwards.
Thus I conclude that any movement towards a solved state of the puzzle is direct movement.
At the very least, it is only the edges that are not being solved relative to the solved state of the cube, but they are still being directly paired up.

 

[TMOY]

At the moment, the top times are our proof. And since *all* the top big cubers use reduction of some sort, it 'proves' that reduction is the best method of the moment. I dont think *all* reduction users use reduction cos all people use reduction. Some people have tried other methods, but are still just faster with reduction.

Yes, it means reduction is the fastest method of the moment. It doesn't mean it's the fastest possible method.There's still a lot of work to be done on direct solving methods to make them go really fast (I discover new improvements on center everyday, for example.)
But of course if nobody ever does that kind of work because everybody feels compelled to use reduction it will never happen

 

[rachmaninovian]

bleah, where is per...well per was the fastest in big cubes in the good old days..

 

[mrCage]

Hi

Let's get 2 facts straight:

1 - reduction methods are most likely the most turn efficient methods
2 - reduction methods also require quite exceptional recognition skills

Without the skills from 2 solving with such method will be relatively slow.

With my cage method i am probably 20% higher on turns than the average direct solver (5x5x5 cube).

Direct solving is opposed to reduction solving.

With reduction solving pieces are not (nesessarily) moved to their final position during a stage. For ce3 the centers stage is direct solving, then the pairing up (tr)edges is indirect solving. The last stage(s) are again direct solving. So ce3 has 2 stages with direct solving and one reductionist step.

For my cage method all steps are direct solving.

For Rachmanovian's methods all steps are also direct solving.

For K4 almost all steps are direct solving. (there is a bit of edge pairing involved AFAICR ...)

I still don't believe we found the most effcient way of solving 5x5x5 (or larger). And i am not speaking og gods algorithm "solving"

Per

 

[AvGalen]

bleah, where is per...well per was the fastest in big cubes in the good old days..

Only unofficially where he claims to have gotten a 1:05 solve on 5x5x5. In real life he gets times likes this: http://www.worldcubeassociation.org/results/p.php?i=2004FRED02#555
Even when everyone else still sucked, Per was never among the top 3. The only time he actually won a big cube event was with sup 3 minute times this year.

I do realise he "might not practice" anymore, but claiming 1:05 solves (on this forum) or 1:41 solves (unofficial world record list) without any official solve or video to back up those claims is a sign of a vivid imagination at least. And don't think 5x5x5 is the exception. He has made similar claims about 4x4x4 with the same discrepancies officially

...There's still a lot of work to be done on direct solving methods to make them go really fast (I discover new improvements on center everyday, for example.)
But of course if nobody ever does that kind of work because everybody feels compelled to use reduction it will never happen

So, what work have you done so far, which improvements have you found? If you want help, just publish what you have found so far and get other people (like me) involved

 

[JLarsen]

Well, we don't directly solve the pieces do we? In 'direct-solving' you solve the edges to where they need to go. But we solve the edges wherever, to make it look like a 3x3, which we then solve..

This is a bad reason.
I'm going to use 4x4x4 as a more suitable example as it doesn't have fixed centers. If I solve the edges first, do they all have to be relative to each other for it to be "direct"?
When solving a 3x3x3, doing the OLL actually breaks up the rest of the puzzle, so the OLL is not a direct algorithm, by your standard.
In fact, no method could possibly be direct, as you are always going to have to move around pieces that are already in their correct positions to solve other ones.
In solving a puzzle, you are always making progress (if you are not, then you are not actually solving it, you are either getting further away from a solved state, or keeping it at the same distance from a solved state). Even if the piece is not in its correct position relative to another piece, you are still moving forwards.
Thus I conclude that any movement towards a solved state of the puzzle is direct movement.
At the very least, it is only the edges that are not being solved relative to the solved state of the cube, but they are still being directly paired up.

Why does this strike me as obsessing over minor technicalities? Like really, is this anything to debate about? I don't see why people flip out over minor little titles to methods. It makes at least marginal sense, it's name isn't smash the cube on the ground, it's Direct Solving, and if the name seems THAT horribly flawed that you're going to get excited over it, then have a fun life on this forum, debating constantly.

 

[mrCage]

Hi Arnaud !!

No need to get personal. I have had similar discussions in the past. And it is not pleasant. There are many reasons i didn't really achieve good times in public. Firstly i never practiced in public. Secondly i always needed a long warmup to get really good times. Thirdly i didnt go to many events.

This was a discussion about methods, not about people. I was mentioned perhaps because i was one the first to "popularise" a non-reductionist method feasible for larger cubes.

Reductionist or not. There is always a stage where you can choose to get advanced or not. Most claim that pairing edges is not dependant on algs. It IS if you learn all the special cases for the last few edges. Similar for my cage method i do use some algs for the LL corners. I use that simple because i have known that for 20+ yrs. I wouldn't say i have gone tabular even i use a few known algs - perhaps not originally mine. My LL edges for cage are indeed "intuitive". Commutator 3-cycles by and large. My parity fix is very easy, but has side effect on centers. So i have "semi-switched" to a 12-move alg where the centers side effect stays on that layer. It can avoid some undesired side effects.

Actually what i have tried to claim is that cage is fun BECAUSE of all the thinking involved. Small tricks i can use to make the method faster.

Direct solving is in NO way the same as intuitive ,although most of my steps are intuitive. The reason i like cage - and will continue to do so - is because of the great lookahead. For every algorithm i use only a few pieces are affected. Sometimes i can see the next TWO (or sometimes THREE) algos that i am going to perform. I do not have to reinspect simply beause i know the pieces will not move "randomly" after an alg.

Every one should use the method that he finds best, most efficient. Or even most fun. Some people feel forced to use ce3 simply because that is what is common.

I claim that direct solving has the most undicovered potential. Reductionism doesn't have too many options for the stages. Yes i cannot prove that - just a gut feeling.

And finally if i picked up practising again, and getting used to v5 i would after a while catch up my old times, maybe even improve - but no way would i beat the current top guys. The current level is simply unbelievable!!

Do cubing for fun, not for the glory ;-) Actually speeeding is probably my least priority. Have fun!!

Per

 

[joey]

Why does this strike me as obsessing over minor technicalities? Like really, is this anything to debate about? I don't see why people flip out over minor little titles to methods. It makes at least marginal sense, it's name isn't smash the cube on the ground, it's Direct Solving, and if the name seems THAT horribly flawed that you're going to get excited over it, then have a fun life on this forum, debating constantly.

Yes it is, he was trying to understand why it is called direct solving and indirect/reduction solving.

 

[Dene]

Why does this strike me as obsessing over minor technicalities? Like really, is this anything to debate about? I don't see why people flip out over minor little titles to methods. It makes at least marginal sense, it's name isn't smash the cube on the ground, it's Direct Solving, and if the name seems THAT horribly flawed that you're going to get excited over it, then have a fun life on this forum, debating constantly.

You haven't been around for very long have you? Ah you young'ns give me a good laugh

I'm glad no one has been able to rebut against my argument though. I vote for a new name for a new name for these kinds of methods. (Something other than non-reduction methods, preferably).

EDIT: Oh and by the way, it may sound like I'm "flipping out", but it's best if you remember for all future reference: when you read a post by me, imagine that I have a big smily face on the whole time (even if I use a smilie that says otherwise). I'm not the type to "flip out" although I do have trouble trying to get across what I mean without it sounding arrogant/ranty.

 

[Kenneth]

I still don't get how reduction doesn't count as "direct"? How can you non-directly solve a puzzle?

My definition is direct if the pice is affected only once (this makes OLL/PLL a type of redux but not CLL/ELL)

My method is both, first I direct two centres, then I reduce by paring and placing two dedges, reduce again by makeing 4 "pairs" + place, direct CMLL, direct two centres, reduce a dedge + place, direct last centre + 2 edges (end F2L), direct ELL.

Nice you are using my algs rachmaninovian. I'm currently working on a wiki page for all my ELL's, with images =)

Here

 

[JLarsen]

Why does this strike me as obsessing over minor technicalities? Like really, is this anything to debate about? I don't see why people flip out over minor little titles to methods. It makes at least marginal sense, it's name isn't smash the cube on the ground, it's Direct Solving, and if the name seems THAT horribly flawed that you're going to get excited over it, then have a fun life on this forum, debating constantly.

You haven't been around for very long have you? Ah you young'ns give me a good laugh

I'm glad no one has been able to rebut against my argument though. I vote for a new name for a new name for these kinds of methods. (Something other than non-reduction methods, preferably).

EDIT: Oh and by the way, it may sound like I'm "flipping out", but it's best if you remember for all future reference: when you read a post by me, imagine that I have a big smily face on the whole time (even if I use a smilie that says otherwise). I'm not the type to "flip out" although I do have trouble trying to get across what I mean without it sounding arrogant/ranty.

Ah, you know I have issues protraying my actual emotions in text. I can understand that.

 

[qqwref]

Looks like people are being confused by the difference between direct solving and reduction. Here's the way I understand it:
- When you direct solve a piece, you put it where it is going to be when you're solved, and none of the algorithms you do afterwards affect where it is (beyond, say, an AUF away). Intriguingly, these steps are often intuitive, but they can also be extremely algorithm-heavy when there is less freedom to move around without affecting what you've already solved! The cross and F2L of Fridrich, and the blocks and CMLL of Roux are direct solve steps.
- When you solve a piece by reduction, you don't actually put it in a solved position, but rather you do some moves that put it in a position which makes the puzzle easier to solve (sort of reducing the complexity). There are really two main types of reduction steps: those that put the pieces into a more convenient or simpler position (such as Petrus's bad edges step, or OLL), and those that pair up pieces so they can be solved in groups (the edge step of reduction).

I wish it were that simple all the time though. It turns out that some steps can be partially direct solving and partially reduction: CLS and ZBF2L, for instance, actually solve some pieces while orienting others. Trying to pair up a couple of 4x4 edges while you're finishing the last few centers would also be a good example of this.

As for methods, no method can be purely a reduction method (since you do have to actually solve the pieces sometime!), but we tend to think of methods as involving reduction when a large part of the pieces are oriented or paired before they actually get solved. The best examples of this are the reduction methods for bigcubes and the computer-solving 'phase' methods (such as Thistlethwaite) which don't solve the puzzle all at once, but in which each step brings the puzzle into a position that can be solved with a more restrictive set of moves.

 

[TMOY]

So, what work have you done so far, which improvements have you found? If you want help, just publish what you have found so far and get other people (like me) involved

I'm currently in the process of writing down my method. But it takes time because I am lazy

Looks like people are being confused by the difference between direct solving and reduction.

My own cage method (which is quite different from Per's) uses some bits of reduction (I'm pairing some edges before putting them in place, and for the last 4 centers, I first put everything on the right pair of opposite faces before solving those pairs). I'm still putting it in the category of direct solving methods because it's closer to pure solving methods than to reduction.

 

[Stefan]

I'm glad no one has been able to rebut against my argument though.

Can't speak for others, but I myself simply just didn't care enough to spend the time on you/this.

 

[joshuachan1995]

rachmaninovian, u should transfer your tutorial on direct solving from sgcuber t o here... i'm sure you will get more attention

 

[Cerberus]

I think it's a really nice solve with very fast movement, I would be interested in a turns/second statistic and at all how many turns are needed compared to reduction.

By the way, reduction is direct solving of centers and then indirect for edges.
But still there (at least for AvG I would say) it's not that indirect because you put the tredge piece where it needs to be at the edge and ending with restored centers or would it be direct if you do centers - 3x3 part - edges? (3x3 part with middle edges + corners and edges could be done with extended AvG algorithm where you undo the changes on the 3x3 part...)

 

[rachmaninovian]

this solve is ~240 moves at most. 240/131.80 =1.82tps if i achieve 2 tps it wld be great...haha

 

[mrCage]

rachmaninovian, u should transfer your tutorial on direct solving from sgcuber t o here... i'm sure you will get more attention

Hi

Is there a direct link to a forum post (or similar??)

Per

 

[Kenneth]

Is there a direct link to a forum post (or similar??)

If you have not found it yourself already? then look in the guides sub forum.

 

[joshuachan1995]

Is there a direct link to a forum post (or similar??)

If you have not found it yourself already? then look in the guides sub forum.

It is not available on this forum.....I will just post the link here


http://sgcubers.com/forum/index.php?topic=582.0