Proposal: New megaminx scramble notation

[Carrot]

For people saying the ++, +-, -+, -- is too hard. it's 4 (+2=6) unique moves... 3x3x3 is 18. Pyraminx is 16. Skewb is 12.
Also, think of it as a coordinate system, first digit is the y axis, the second is the x axis, that should give you a direction vector (+- = (-1,+1) = Up - left = R++ D--; ++ = (+1,+1) = up - right = R++ D++). It just happens that directions are already handled like this, so north-east is always called north-east, not east-north. So +- could be treated like North-west, people with a megaminx in front of them should see how North-west would translate as R++ D--. (North=+, East=+)

Spoiler

Don't treat above as a tutorial or anything, it's just some messy explanation of a few ways to intuitively understand how the -+ system works. (it's recommended to think in order to interpret what I wrote, haha)

(note: The (R+ D-) ... etc notation would not really help me not getting lost, so I wouldn't see why change anything from what it is to that, but that's just my opinion)

I like how readable that notation stays when it's compact, and how the pairs always start with the same side.

Is this a proposal for the community, or for the WCA?

For official scrambles, we need to make sure the notation is easy to read (and remember) even for people who are not familiar with the puzzle. If we want to group moves, I'd prefer something like the following:

Community.

I'm pretty sure each line of a WCA scramble ends with U if the previous move was D++ and U' if the previous move was D--, so you dont even need the last U/U'/</>

I call hoax until documentation is provided.

 

[Stefan]

I call hoax until documentation is provided.

https://github.com/cubing/tnoodle/b...scrambles/src/puzzle/MegaminxPuzzle.java#L331
The value of variable dir is reused for the U-turn.

I don't remember why, though. It was probably in my proposal and maybe I just really preferred 2/5-turns over 1/5-turns.

 

[Lazy Einstein]

Just a quick weigh in here.
I find this notation very easy to follow compared to the normal notation. I don't lose my place as easy. I really like the idea of every other line being adjusted forward to prevent confusion.

I hope this get implemented because it fills a need in helping to correct commonly made mistakes in scrambling.

 

[Ranzha]

For official scrambles, we need to make sure the notation is easy to read (and remember) even for people who are not familiar with the puzzle. If we want to group moves, I'd prefer something like the following:

(R-- D++) (R++ D++) (R++ D++) (R++ D++) (R++ D--) U'
(R++ D--) (R-- D++) (R++ D--) (R++ D--) (R-- D++) U
(R++ D--) (R-- D--) (R++ D--) (R-- D++) (R-- D--) U
(R++ D--) (R-- D++) (R++ D++) (R-- D++) (R-- D++) U'
(R-- D++) (R-- D--) (R++ D--) (R-- D--) (R-- D++) U'
(R++ D++) (R++ D--) (R-- D--) (R++ D++) (R++ D--) U'
(R-- D++) (R++ D++) (R-- D++) (R-- D--) (R-- D--) U

What about Square-1?

 

[vcuber13]

What about Square-1?

What about it?

 

[Lucas Garron]

What about Square-1?

Good question. Two relevant facts for Square-1:

- There are different kinds of turns, and the (possible valid) amount of turning depends on the position.
- There is already a widely used notation to describe Square-1 algs that can be used for our method of scrambling.

I don't remember anyone ever saying we should change Square-1 notation, and I certainly haven't seen alternate proposals.

A closer example is clock, which we did change.

 

[Mikel]

I think a method to distinguish lines such as an offset or white/gray alternating lines would be great. I would recommend it for 6x6 and 7x7 along with megaminx.

 

[EMI]

Good question. Two relevant facts for Square-1:

- There are different kinds of turns, and the (possible valid) amount of turning depends on the position.
- There is already a widely used notation to describe Square-1 algs that can be used for our method of scrambling.

I don't remember anyone ever saying we should change Square-1 notation, and I certainly haven't seen alternate proposals.

A closer example is clock, which we did change.

Not to the good imho.
Square-1 is hard to scramble, but because of the puzzle, not the notation. The only way to make it easier would be to replace "/" with an "R2" and make (x,y) mean x *pieces* to the left, instead of x*30°. I'm not sure if that is easier though. We should probably just use Lego robots for it

 

[qq280833822]

Integrated in cstimer. See "megaminx" -> "Carrot"

 

[uberCuber]

make (x,y) mean x *pieces* to the left, instead of x*30°. I'm not sure if that is easier though.

I would argue that that is harder/slower: with the current notation, I can follow a square-1 scramble quickly and easily without ever looking down at the puzzle. When I see a 4, I have a sense of how "far" that is and don't have to watch myself turning the puzzle. But if (x,y) referred to the number of pieces, then how "far" a 4 is depends on the state of the puzzle, requiring me to look down at the puzzle a lot, making me go slower and also potentially lose my place in the scramble more easily.

 

[qqwref]

Yeah, the number of pieces is harder to be fast with, even though it might seem easier for new people. Back when I did Square-1 sim the hardest part for me was dealing with the U/D turns by piece instead of by angle.

I do like this Megaminx notation. Although I wish we had a scrambling setup that could reach most of the puzzle states.

 

[cubernya]

Yeah, the number of pieces is harder to be fast with, even though it might seem easier for new people. Back when I did Square-1 sim the hardest part for me was dealing with the U/D turns by piece instead of by angle.

I do like this Megaminx notation. Although I wish we had a scrambling setup that could reach most of the puzzle states.

I know it can't reach all, but what percentage can it reach?

 

[Lucas Garron]

I know it can't reach all, but what percentage can it reach?

70 independent choices = 2^70 states
The number of states is about 2^226.

So, pretty much 0%.

Note that this information doesn't tell you anything about whether our scramble program is good or bad. If those 2^70 are indistinguishable from random (to a human), we're totally fine.
(I've been meaning to run calculations to see if certain simple measurements of randomness are satisfied, e.g. amount of separation of adjacent pieces.)

 

[qqwref]

Lucas is right. Even with 77 choices (choosing the U or y direction as well) the number of positions is negligible compared to the number of positions of the puzzle.

 

[Renslay]

70 independent choices = 2^70 states
The number of states is about 2^226.

Note that this information doesn't tell you anything about whether our scramble program is good or bad. If those 2^70 are indistinguishable from random (to a human), we're totally fine.

Then why is everybody so hyped about random state scrambles instead of fix 25 move scrambles in 3x3?
(Especially considering the huge argument in this thread: http://www.speedsolving.com/forum/showthread.php?48443-Best-way-to-randomize-a-cube )

 

[Lucas Garron]

Then why is everybody so hyped about random state scrambles instead of fix 25 move scrambles in 3x3?
(Especially considering the huge argument in this thread: http://www.speedsolving.com/forum/showthread.php?48443-Best-way-to-randomize-a-cube )

There is no contradiction. In both cases, the goal is to produce a state that is no easier/harder for a human to solve than a random state.

For 3x3x3 it's really easy to generate a short scramble that is indistinguishable from a (uniformly) random state... namely, because it *is* a random state. (By easy, I mean your smartphone can literally do it in the blink of an eye.)
If you're using a computer/smartphone, there is no good reason not to use random-state scrambles for 3x3x3 – especially since quite a few timers are available for each common platform that will do this for you by default.

25-move scrambles are bad for several reasons. One that is easy to grasp: EO is about twice as likely to be solved (compared to a random state) – this already makes it a bad choice for BLD practice.

For Megaminx, generating a short alg to produce a random state is far from practical (and it would take too long to apply), so *there* the notion of indistinguishability actually becomes relevant.

 

[Carrot]

https://github.com/cubing/tnoodle/b...scrambles/src/puzzle/MegaminxPuzzle.java#L331
The value of variable dir is reused for the U-turn.

I don't remember why, though. It was probably in my proposal and maybe I just really preferred 2/5-turns over 1/5-turns.

*facepalm* why? Why? That's a terrible reasoning, but okay. I was wrong about that, but no documentation was provided why it was reasonable to take that decision.

Integrated in cstimer. See "megaminx" -> "Carrot"

Sweet!

 

[Stefan]

*facepalm* why? Why? That's a terrible reasoning, but okay. I was wrong about that, but no documentation was provided why it was reasonable to take that decision.

I don't see you complaining about all the other 2/5-turns, so why about these?

For 2/5-turns in general, here's a bit (starting at "Now an improvement..."):
https://www.worldcubeassociation.org/forum/viewtopic.php?p=1458#p1458

 

[Carrot]

I don't see you complaining about all the other 2/5-turns, so why about these?

For 2/5-turns in general, here's a bit:
https://www.worldcubeassociation.org/forum/viewtopic.php?p=1458#p1458

Mainly because I see the R, D part as the "scrambling" part and the U's as an adjustment in order to break up the pairs that can't be broken up just by the R, D part, so I don't see them as being related. The best argument I can see for for only using 2/5 turns is that you can reach longer and since all permutation groups are even the quality is not decreased.However, I still like the 1/5 turns in the end of half of the lines.

(thanks for the link, but I had already skimmed that a few years back)

 

[sk8erman41]

I've been using the "Carrot" scramble in CSTimer and gotta say it is waayyyyyy easier, I like it a lot well done