Article 4: Scrambling

JohannesLaire (2008-03-20 17:45:02 +0000)
[quote="cada":1kziu5vo]Also, it [Cube Explorer] is not available to run on my computer. I would like to see a platform independent scramble generator.[/quote:1kziu5vo] I agree, and I also think that it should be open source so it could be verified that it works correctly. Why is 2x2x2 still scrambled with 25 moves? [i:1kziu5vo]"All scramble programs kindly provided by Jaap Scherphuis and Syoji Takamatsu (Pyraminx)."[/i:1kziu5vo] This should be updated.
Ron (2008-03-21 07:32:44 +0000)
Clément is working on a platform independent scramble program. For now we stick to Cube Explorer.
VooX (2008-03-24 16:18:46 +0000)
Currently the best math states that all 3x3 cubes can be solved in a mimimum of 26 turns (half-turn metric). I strongly feel in order to ensure a random scramble, we must exceed this number when scrambling cubes for competition. I propose that all 3x3 scrambles be 30-35 moves (half-turn metric) to ensure that all cubes in competition have a random pattern with no significant advantages to any starting cross colour. We must exceed the minumum turns to solve a cube to ensure we have reached a random state, and we must add a few moves to account for the slightly non-random nature of any computer program (i.e. cube scrambler). I feel strongly about ensuring we have random solves, and I feel we should only need one piece of software to determine our scrambles. Having one software generate a pattern and another generate the scramble to get their is an unneccessarly complicated process. The standard scrambler used in the past with more moves to scramble should be sufficient until one single piece of software is written.
Lucas (2008-03-26 23:54:56 +0000)
I've noticed that I don't get the cube in very consistent orientations at the start of a solve. I can't request an orientation anymore (for scrambling, which presumably applied to the start of inspection), but I don't nearly get UF=WG all the time. I can remember a few UF=WO from Stanford. Especially for BLD, it would be nice either to know the orientation, or that the orientation is random. I suggest that the regulations require a random orientation of the puzzle to randomize the orientation of non-moving pieces (e.g. centers on odd cubes) by adding 0/1/2 rotations to the end of an alg, or maybe requiring the orientation to be manually "randomized" for each competitor. (Note that random orientations, notwithstanding the "competitors area," make it harder to glean the state of a scrambled cube consistently) Either that, or can we add to 4d) or A2d) "The cube may be given to the competitor in any arbitrary orientation." ? Or does the WCA think this is not necessary?
Ron (2008-03-27 07:01:49 +0000)
@Voox [quote:1jrg5qn5]I strongly feel in order to ensure a random scramble, we must exceed this number when scrambling cubes for competition.[/quote:1jrg5qn5] In the draft version we propose Cube Explorer for scrambling positions of Rubik's Cube. For most other puzzles we have no knowledge about the depth, so we also do not know how to exceed it. @Lucas [quote:1jrg5qn5]Especially for BLD, it would be nice either to know the orientation, or that the orientation is random.[/quote:1jrg5qn5] It is logistically not possible to manage that all orientations are the same. So you have to assume that the orientation is random. Do you really think that we should add that to the regulations? Or just leave it empty? Thanks, Ron
anders (2008-03-27 21:02:17 +0000)
[quote="Lucas":266gdxk3]I've noticed that I don't get the cube in very consistent orientations at the start of a solve. [/quote:266gdxk3] My opinion is that the cube orientation at the start of the solve should be random. Furthermore, I think that the cube orientation [i:266gdxk3]during scramble[/i:266gdxk3] also should be random, and not according to 4d). Earlier 4d) made some sense since all cubers, at least in the final round, must have the same scramble. But in the new regulations, 4e) makes it possible for competitors to have different scrambles in the final.
Gilles (2008-03-27 21:10:34 +0000)
Maybe we should add to 1h) competitors cannot be divided into groups in final rounds?
Ron (2008-03-27 21:20:10 +0000)
Hi Gilles, That would be a bad idea. If an event has one round with 50 competitors, you should/must use groups for practical reasons. Even for WC 2007 we used groups in the finals. Thanks, Ron
Gilles (2008-03-27 22:06:50 +0000)
Oh yes. WC '07. You're right.
anders (2008-03-27 22:10:54 +0000)
[quote="Gilles":jenm9i15]Maybe we should add to 1h) competitors cannot be divided into groups in final rounds?[/quote:jenm9i15] This is the case according to [i:jenm9i15]the current[/i:jenm9i15] regualtions: 4e) Competitors must solve the same scrambles per round. At the main judge's discretion, scrambling algorithms in preliminary rounds may be randomly chosen from a pool of scrambles, for example to prevent cheating in large competitions.
Ron (2008-03-27 22:25:28 +0000)
Hi Anders, Yes, I am aware of that. For the new regulations 4e) was changed. Thanks, Ron
VooX (2008-04-01 16:31:42 +0000)
Not to be on a soapbox but... I don't think anyone has given a good reason why we should use Cube Explorer which will require multiple pieces of software (without the cube solver program telling someone a scramble alg will be impossible without using like 50-80 moves which is a typical solve!!) and adds complexity to a system that can be kept simpler and much more efficient. Ron, you mentioned that you are not aware of the mathematical depth necessary to scramble cubes other than 3x3 randomly. The 3x3 value of 26 has been researched extensively by PhD's and others at M.I.T, I.B.M., and others who provided extensive analysis of the cube since the early 80's. The number of turns has been reduced to 26 over many years of refining research and computer analysis methods. By using a scramble program which strives to randomize turns as much as possible (we currently use an excellent algorithm generator with the official WCA scrambler written by Jaap) we keep the system random yet simple. [b:2llb7oiq]In order to make our scrambles truly random we should strive to use our current scrambler in a more mathematically random way by increasing the number of turns per scramble to around 35 turns.[/b:2llb7oiq] THE PROBLEM LIES NOT WITH OUR TOOLS (Jaap's scrambler) BUT WITH HOW [b:2llb7oiq]WE[/b:2llb7oiq] ARE USING THEM. As far as listed in the regs, the 3x3 would be the only type of cube to use Cube Explorer, so not knowing the depth of turns necessary for randomness on other cubes should not be an issue. And, please take this as constructive criticism, if Ron, Harris, Joel, or any other board members do not know the mathematics of cube randomness, I implore you all to do some research. Some of our greatest mathematical, computer engineering, and scientific minds in the past 25 years have taken an interest in studying the cube extensively. Contact M.I.T., or I.B.M., or other groups that have PUBLISHED RESEARCH PAPERS on the very topic for further information about random states in cubing and God's algorithms. Ignorance is not bliss, and should not be an excuse to complicate a system that is very effective and efficient using only Jaap's scrambler. It is in the best interest of the cubing community to keep things simple. Cube Explorer does not work on Mac computers and will exclude anyone who does not have a PC or Linux. How can this be a universal solution then? People at home would not be able to fairly replicate competition scrambles to practice with at home if they do not have the appropriate computer. Jaap's scrambler works in HTML which is a UNIVERSAL STANDARDIZED language understood by all platforms of computer. Please WCA board, refrain from using Cube Explorer as it add considerable complexity to a system that currently just needs a little tweak (35 moves!). Any response from the board or others I would love to hear.
Ron (2008-04-01 20:34:47 +0000)
Hi VooX, Sorry, I do not like the tone of your message. I am very much aware of research on this subject. And you obviously missed the latest news (25 moves for Rubik's Cube) and all the other discussions and investigations that we had on using scramble sequences or scrambled positions. I am not aware of good research on the depth of puzzles like 4x4 and 5x5. We chose to use random scrambled positions. At the moment only Cube Explorer has the required features. Hopefully we will soon have other options too. So that we can also run the program on Linux or Mac. For the time being it should not be a problem to have a Windows pc available at a competition. I know some people want to use other platforms. Clément Gallet is already working on solving this problem. Have you actually tried the new version of Cube Explorer, with the WCA scramble feature? Thanks, Ron
Pedro_S (2008-04-02 00:14:04 +0000)
I got a pyraminx yesterday, and I'm finding it quite cool I did solves using the official scrambler...and I was wondering... why some scrambles have just one vertex turn while other have 3 or 4? :? it would be fairer if they all had the same (2 or 3) or no?
Ron (2008-04-02 04:18:13 +0000)
Hi Pedro, The scrambles for the vertices are also random. Each vertex has 3 options: clockwise, counter clockwise, none. Sometimes that means that no vertices are incorrect. Thanks, Ron
Lucas (2008-04-02 06:36:24 +0000)
[quote="VooX":sjlem1n6]Not to be on a soapbox but...[/quote:sjlem1n6] Ummm...? [quote="VooX":sjlem1n6] I don't think anyone has given a good reason why we should use Cube Explorer[/quote:sjlem1n6] I do, at least. That's someone. [quote="VooX":sjlem1n6] which will require multiple pieces of software[/quote:sjlem1n6] No. [quote="VooX":sjlem1n6] (without the cube solver program telling someone a scramble alg will be impossible without using like 50-80 moves which is a typical solve!!)[/quote:sjlem1n6] No? The program WILL give an alg. And 50-80 is not necessary, and not necessarily a typical solve (I average low 50's). [quote="VooX":sjlem1n6] and adds complexity to a system that can be kept simpler and much more efficient.[/quote:sjlem1n6] No. And it will be. [quote="VooX":sjlem1n6] Ron, you mentioned that you are not aware of the mathematical depth necessary to scramble cubes other than 3x3 randomly. The 3x3 value of 26 has been researched extensively by PhD's and others at M.I.T, I.B.M., and others who provided extensive analysis of the cube since the early 80's.[/quote:sjlem1n6] He is aware of it, he just does not know any conclusive evidence for a specific a value for God's number. First of all, what "the 3x3 value" of 26 (which, as noted by Ron, is 25 now)? PhD's do not research any more than American Medals of Congress do. MIT? IBM? If you want to put it that way... (Throw in Stanford for 25.) I think you're trying to say that there has been a lot of research to find a lower bound on the maximum random-state scramble length. This is somewhat irrelevant if empirically we always get around 20, and the theoretical maximum output from CE (2-phase) is still pretty low... Also, note that he is aware that 25 moves are indeed enough to scramble a cube randomly. We're concerned about the distribution, which appears to be biased in returning scrambles with (slightly) significantly less entropy than desirable. [quote="VooX":sjlem1n6] The number of turns has been reduced to 26 over many years of refining research and computer analysis methods.[/quote:sjlem1n6] Number of turns in what sense? You mean "[known bound for] the [maximum] number of turns [necessary to scramble a cube into any given state]." I has indeed required refinement and research, but so? [quote="VooX":sjlem1n6] By using a scramble program which strives to randomize turns as much as possible (we currently use an excellent algorithm generator with the official WCA scrambler written by Jaap) we keep the system random yet simple.[/quote:sjlem1n6] We are planning to make the system certifiable random yet simple. [quote="VooX":sjlem1n6] [b:sjlem1n6]In order to make our scrambles truly random we should strive to use our current scrambler in a more mathematically random way by increasing the number of turns per scramble to around 35 turns.[/b:sjlem1n6][/quote:sjlem1n6] 1) That will still not make them completely random. However, on the Yahoo group last summer I posted that I empirically found scrambles to be sufficiently random after 35-ish random moves. (I need to do that math over again now that I've taken a Statistics class and see what I get now.) 2) The scramblers will not be happy, nor the time schedulers and competitors. How about reducing the number of turns [i:sjlem1n6]and[/i:sjlem1n6] providing higher fidelity? [quote="VooX":sjlem1n6] THE PROBLEM LIES NOT WITH OUR TOOLS (Jaap's scrambler) BUT WITH HOW [b:sjlem1n6]WE[/b:sjlem1n6] ARE USING THEM.[/quote:sjlem1n6] 1) What is the problem? 2) If it's randomizing scrambled states for competitors completely, the problem does lie with what tools we are using. 3) We can use our tool, the computer, to use random scrambles. So, really, I agree with you here. [quote="VooX":sjlem1n6] As far as listed in the regs, the 3x3 would be the only type of cube to use Cube Explorer, so not knowing the depth of turns necessary for randomness on other cubes should not be an issue.[/quote:sjlem1n6] Shouldn't? Anyhow, other cubes will either get a random-state scrambler soon, or solving times for them depend less on the scramble (say, 4x4x4, Megaminx, or 5x5x5, for which we would certainly use random-state scramble if someone had a good program with short output). (Oh, and we do know it for quite a few puzzles, which should be getting corresponding scramblers soon) [quote="VooX":sjlem1n6] And, please take this as constructive criticism, if Ron, Harris, Joel, or any other board members do not know the mathematics of cube randomness, I implore you all to do some research.[/quote:sjlem1n6] Harris is a board member? :P And what is "knowing" it? What will research help with fairer scrambles? [quote="VooX":sjlem1n6] Some of our greatest mathematical, computer engineering, and scientific minds in the past 25 years have taken an interest in studying the cube extensively.[/quote:sjlem1n6] They have, indeed. What's it with your love for MIT and IBM? MIT is as well qualified for cube research as any other good university, and cubing is not Deep Blue (well, kinda, but no-one at IBM engineers cube-solving computers). (If I'm not attending MIT or working for IBM, may I still study the cube?) May I note that Kociemba has actually made a practically usable optimal solver (you just need a good processor and lots o' RAM) included with CE? Who cares what we know about what algs it gives if we know that it has to be sub-26 (and now thus at worst on par with our current number of twists)? (in regard to scrambling.) [quote="VooX":sjlem1n6]Contact M.I.T., or I.B.M., or other groups that have PUBLISHED RESEARCH PAPERS on the very topic for further information about random states in cubing and God's algorithms.[/quote:sjlem1n6] Uh? [quote="VooX":sjlem1n6] Ignorance is not bliss, and should not be an excuse to complicate a system that is very effective and efficient using only Jaap's scrambler.[/quote:sjlem1n6] Ignorance? Ron has already demonstrated he is less ignorant of the current know bound than you. :wink: "excuse" - uh, reason? "complicate" - it's not more complicated "effective" - not for what we want. [quote="VooX":sjlem1n6] It is in the best interest of the cubing community to keep things simple. [/quote:sjlem1n6] We care about the competition scramblers and competitors here. And as previously stated, they [i:sjlem1n6]will[/i:sjlem1n6] be kept simple. [quote="VooX":sjlem1n6] Cube Explorer does not work on Mac computers and will exclude anyone who does not have a PC or Linux.[/quote:sjlem1n6] Not natively, yes. I've been told WINE can run it, so if you make me insist, I'll say you can indeed run it on a Mac if you install Linux, or whatever.. How can this be a universal solution then?[/quote] Cube Explorer is also temporarily the only desirably effective solution. The WCA is surely working on this. Could you contact your friends at MIT and IBM and ask them to make a robust scrambling/solving program for Macs? The WCA would surely love to approve it. I have also suggested officially allowing 50-move (negotiable, 35-40 is probably fine) scrambles to be used as an alternative (if no appropriate computer is available), but this would come into little use and is not really necessary. You can always print out scrambles in advance, too. [quote="VooX":sjlem1n6] People at home would not be able to fairly replicate competition scrambles to practice with at home if they do not have the appropriate computer.[/quote:sjlem1n6] Not true for most, as I've argued. You can install Linux on anything, anyhow (as long as your computer is appropriate enough for that). And I don't think you mean "replicate," but rather "simulate" or "generate official scrambles." Also, they can. At home, you can use any program to generate scrambles, and theoretically the result should be the same. Batch a state-generator with ACube if you want. Or pre-generate thousands of random scrambles for yourself, whatever. ("sq1optim.exe -r", anyone? :) ) Anyhow, most people will not notice a difference, and should not (and I've never heard a fast cuber who proudly stated that he used WCA-official scrambles for practice - though some may do so). However, we want competitions to be demonstrably fair, and it's easy to use give them (pseudo)random states to solve. It will be relatively impeccable system -a scheme for eliminating easy starts will be difficult to define, and still rather ineffective, so this is the best we can do. [quote="VooX":sjlem1n6] Jaap's scrambler works in HTML which is a UNIVERSAL STANDARDIZED language understood by all platforms of computer.[/quote:sjlem1n6] IT DOES NOT; IT USES JAVASCRIPT (which is indeed a UNIVERSAL STANDARDIZED language [i:sjlem1n6]understandable[/i:sjlem1n6] by all platforms of computer). [quote="VooX":sjlem1n6] Please WCA board, refrain from using Cube Explorer as it add considerable complexity to a system that currently just needs a little tweak (35 moves!).[/quote:sjlem1n6] It would not. And that's a 40% increase! Do you have any idea how inconvenient it will be for the scramblers (and organizers, who have timetables to keep, and scramblers to convince to hep them) to use longer scrambles that are still not random, instead of actual random [i:sjlem1n6]and[/i:sjlem1n6] shorter scrambles? [quote="VooX":sjlem1n6]Any response from the board or others I would love to hear.[/quote:sjlem1n6] You're welcome. I think your main point consisted of your misunderstanding of Cube Explorer's functioning (and some weird mention of how MIT and IBM relate to it). As far as I understand, we want: 1) To be able to claim that our sport is doing its best to high-integrity randomness and fairness in scrambles. 2) To give fair scrambles and thus ensure that we eliminate luck given to competitors through a faulty scrambling scheme (if you use 3OP for BLD, or pre-orient for speedsolving, having all edges oriented 3 times as likely as the ideally equally likely 12-edges-misoriented is not insignificant with the number of scrambles being used - and a no-EO BLD scramble has apparently already been given in competition). This way, we can endeavor to judge the skill of competitors as best as possible, by presenting them with scrambles not possibly more biased in their favor than we can prevent. Using CE (or other low-move-count alternatives as they become available) will reduce randomness to the pseudo-RNG employed, which I would consider safe enough (the seed is certainly greater than 10^19, I believe). If you bring in a quantum-random-number-thingy, we can make sure that competitors are given truly random scrambles, but I don't think we'll need to do that anytime soon. Simply, switching to random-state scrambling is a simple way to ensure greater randomness and fairness, and simultaneously speed up scrambling slightly. We cannot afford to stay below this standard indefinitely, and should move immediately, even if this results in a temporary period of time where the software is not ideally available and developed (for it is still practical and effective). -Lucas Garron
VooX (2008-04-02 07:05:31 +0000)
[quote="Ron":1xvw3eyx]Hi VooX, Sorry, I do not like the tone of your message. I am very much aware of research on this subject. And you obviously missed the latest news (25 moves for Rubik's Cube) and all the other discussions and investigations that we had on using scramble sequences or scrambled positions. I am not aware of good research on the depth of puzzles like 4x4 and 5x5. We chose to use random scrambled positions. At the moment only Cube Explorer has the required features. Hopefully we will soon have other options too. So that we can also run the program on Linux or Mac. For the time being it should not be a problem to have a Windows pc available at a competition. I know some people want to use other platforms. Clément Gallet is already working on solving this problem. Have you actually tried the new version of Cube Explorer, with the WCA scramble feature? Thanks, Ron[/quote:1xvw3eyx] Ron, Didn't intentionally try and irk you with my post, as I respect your contributions to the community I could have chosen my words with more care. On this, the official WCA site, there is no explanation behind the reason for changing systems. I cannot run the Cube Explorer program as I am on the Mac (non-Intel). I have however examined the programming code and mathematical explanations posted on the website. It is an excellent program, no doubt, but I feel adds too much complexity to the competition. In the explanations listed on this site, it sounds like a two step process, one to find a pattern, the other to find an optimal solve (i.e. scramble). A quote from the Cube Explorer web site: [quote:1xvw3eyx]Cube Explorer 4.20 needs 128 MB of RAM and Windows 98/ME/NT/2000/XP/VISTA to run. I also have a special version Cube Explorer 4.20s available, which uses more than 2 GB of RAM for the huge optimal solver tables. It is about 15 times faster than the standard optimal solver and optimally solves a random cube in less than two minutes on average on a 3 GHz machine. [/quote:1xvw3eyx] This is not an efficient answer is it? Only a high powered machine with a fast processor and a lot of RAM (>2GB) can solve ONE scrambled state in less than 2 minutes (lower powered machines would take 30 minutes by the author's calculations). Given the high number of scrambles required for the 3x3 cube with the regular event, blindfolded, one-handed, multi-blind, etc., we do not have a system that can QUICKLY generate the scrambles if we need to on-site (as we might if scrambles were somehow leaked and needed to be changed). In comparison, Jaap's program requires no memory, is completely platform compatible and quick to generate the scramble algorithm; at least while Cube Explorer still takes so long to calculate the scramble I still think CE should not be used. Again, sorry for coming off rude, I am really ill right now, and have less tact and more temper while ill than usual. VooX
Lucas (2008-04-02 18:45:05 +0000)
[quote="VooX":wtf3w7hj] Didn't intentionally try and irk you with my post, as I respect your contributions to the community I could have chosen my words with more care.[/quote:wtf3w7hj] Then why didn't you? [quote="VooX":wtf3w7hj]On this, the official WCA site, there is no explanation behind the reason for changing systems.[/quote:wtf3w7hj] Where do you expect it to be? On the homepage? Explicitly inside the regulations, adding more to our lengthy document in order to justify a clear procedure? How about just inside this forum and thread? Also see my first post [url=http://worldcubeassociation.org/forum/viewtopic.php?t=373:wtf3w7hj]here[/url:wtf3w7hj]. [quote="VooX":wtf3w7hj] I cannot run the Cube Explorer program as I am on the Mac (non-Intel).[/quote:wtf3w7hj] Too bad for you (right now), but not competitions. If they have an issue with this, it will be resolved... [quote="VooX":wtf3w7hj] I have however examined the programming code and mathematical explanations posted on the website. It is an excellent program, no doubt, but I feel adds too much complexity to the competition.[/quote:wtf3w7hj] Feel? The coplexity is in its operation, not how one uses it. Would you like me to say that using a computer for Jaap's scrambler adds complexity because we are relying on circuits and processors and operating systems that the WCA has not investigated itself? [quote="VooX":wtf3w7hj] In the explanations listed on this site, it sounds like a two step process, one to find a pattern, the other to find an optimal solve (i.e. scramble).[/quote:wtf3w7hj] Get better hearing. It's not. Both are actually multi-step processes, involving bring a computer, booting it, starting a program, generating scrambles, printing, etc. And in case you noticed, OPTIMAL is [u:wtf3w7hj][size=200:wtf3w7hj]NOWHERE[/size:wtf3w7hj][/u:wtf3w7hj] required. It's not necessary, becaues the states are random. Optimal would only sve us a few (2-3) moves, anyhow. [quote="VooX":wtf3w7hj] This is not an efficient answer is it? Only a high powered machine with a fast processor and a lot of RAM (>2GB) can solve ONE scrambled state in less than 2 minutes (lower powered machines would take 30 minutes by the author's calculations). [/quote:wtf3w7hj] I still don't think you understand that this is not by any means necessary. There's no reason to go otimal unless we can get optimal solutions nearly as fast as sub-optimal, short, practical ones. [quote="VooX":wtf3w7hj] Given the high number of scrambles required for the 3x3 cube with the regular event, blindfolded, one-handed, multi-blind, etc., we do not have a system that can QUICKLY generate the scrambles if we need to on-site (as we might if scrambles were somehow leaked and needed to be changed). [/quote:wtf3w7hj] No. 1) The number of scrambels needed is roughly know, and it's easy to prepare by printing way too many scrambles. 2) We [u:wtf3w7hj][i:wtf3w7hj][b:wtf3w7hj]DO[/b:wtf3w7hj][/i:wtf3w7hj][/u:wtf3w7hj] have a system that can generate scrambles on site. Seriously, have you ever used CE or thoroughly informed yourself about its function? (Or do you need PhD's at MIT and IBm to tell you how it functions?) You implore us to do our research, but you clearly haven't done yours, being uninformed about the current bound and the functioning of CE. Leaked? Can't that happen at the competition, too? [quote="VooX":wtf3w7hj] In comparison, Jaap's program requires no memory, is completely platform compatible and quick to generate the scramble algorithm; at least while Cube Explorer still takes so long to calculate the scramble I still think CE should not be used.[/quote:wtf3w7hj] 1) It uses memory. (Little, but it does.) 2) CE DOES NOT TAKE LONG! [quote="VooX":wtf3w7hj] Again, sorry for coming off rude, I am really ill right now, and have less tact and more temper while ill than usual. VooX[/quote:wtf3w7hj] Again, sorry for coming off rude, I am really sane right now, and have less tact and more temper while people post stupid sutff.
Ron (2008-04-02 20:06:44 +0000)
Hi Voox, Herbert Kociemba added a special WCA scramble feature to Cube Explorer. If you click on 'WCA Scrambles' in the menu, then 5 random positions are generated. The program also almost immediately shows solutions for these positions. Most of the time around 20 moves. You click on 'Print Scrambles' in the menu to get a nice print. That's it. Check out this thread for a discussion on the scrambling. http://games.groups.yahoo.com/group/spe ... sage/34928 Thanks, Ron
Max (2008-04-02 20:43:15 +0000)
The scramble pages misses a doctype. It should be added in order to verify standards-compliance. Also, all puzzles should be scrambled to a random state (that is, where orientations and permutations of all pieces i randomized). I think the best would be if someone writes a program that generates a random puzzle then computes a scramble that puts the the puzzle into that state. This program could be written in a well-known language (C or C++, perhaps?), be distributed in source and binary form and use a CLI to ensure biggest possible platform independence. If needed, a separate GUI could be written for Windows. I would do this myself, but i am quite a crappy programmer. ;)
Ron (2008-04-02 21:25:40 +0000)
Hi Max, OK, I updated the scramble pages for the DOCTYPE. We are using random positions now for: - Rubik's Cube - Rubik's Clock If we would have suitable programs for other puzzles, then we should use them. Requirements: - fast generation of random positions and the required move sequences to get there - not too lengthy move sequences (<=70 moves) and not significantly longer than current sequences - image of each generated random positions, with print option - preferrably available on all platforms - preferrably one program that handles all puzzles Thanks, Ron
StefanPochmann (2008-04-03 08:44:16 +0000)
[quote="Ron":2y58ws84]- not too lengthy move sequences (<=70 moves) and not significantly longer than current sequences[/quote:2y58ws84] I'd like to mention that the old megaminx scrambles had fewer moves than the new ones, but were significantly harder to execute and took a lot more time. Number of turns doesn't tell the whole story, ease and time of execution should be paramount. I do realize for many puzzles the number of turns is a good indicator for this, but not for all.
Clement Gallet (2008-04-04 11:32:26 +0000)
Hi Uniform scramblers that follow Ron's requirements would be not hard to write for Pyraminx, 2x2 and Square-1. For Megaminx, we would prefer a fast scramble to perform over an uniform scramble, except if someone manages to get very short uniform scrambles. So the only puzzles that cause problems are the 4x4 and 5x5. For the 4x4, sub-optimal solvers have already been created that can produces scrambles in less than 60 moves in average. Clément
Stefan Łapicki (2008-04-06 18:48:22 +0000)
[quote="Ron":1swesfu8] We are using random positions now for: - Rubik's Clock[/quote:1swesfu8] But it isn't written in regulations.
Ron (2008-04-06 20:29:51 +0000)
@Stefan Lapicki [quote:2ghdu8ww]But it isn't written in regulations[/quote:2ghdu8ww] Yes, you are right. I corrected this in version 4f, April 6, 2008. Thanks, Ron
Pedro_S (2008-04-14 00:17:27 +0000)
I was doing some square-1 solves, using the new scrambler (which I really like, since it's so much easier to read with the ( ) around the numbers) and I was doing it like (a,b), turn, (c,d), turn, and so on, starting always with the top and bottom layers, never with the right but I got this scramble: (0,3) (0,3) (-3,5) (2,1) (4,0) (4,3) (0,4) (0,1) (6,0) (0,4) (6,0) (5,2) (0,4) (6,5) (4,0) (6,4) (0,2) and if you do it like I was doing (and actually did for all other scrambles), it won't work. You can't the (-3,5) part. If you start with the right layer, like / (0,3) / (0,3)... it works this is really strange. Is it standardized like I was doing, starting always with top/bottom?
Ron (2008-04-14 20:03:47 +0000)
Dear Pedro, Yes, that is strange. I probably caused this when I changed the format of the scrambles. No idea where though. I asked Jaap to look into this. Thanks, Ron
Ron (2008-04-21 05:55:19 +0000)
@Pedro Jaap found the bug. Tonight I will upload the new version, so please try it again on Tuesday. Thanks for your feedback. @Stefan and Clément This weekend we used the new Megaminx scrambler for German Open 2008. It was a joy to use it, so much easier and faster. The scrambled positions look equally hard. That WR holder Erik set worse times than normal, was caused by something other than the scrambled positions. ;-) Thanks, Ron