future competitions and lucky scrambles

[sqAree]

I totally agree and had this idea myself a while ago.

Since for the 2x2x2 and 3x3x3 all possible positions are known meanwhile, it could be accomplished to only allow scrambles that lead to positions that CANNOT be solved with (a) skip(s) by any method available. It could just be looked up from a database that contains only non-lucky-solvable scrambles.

I find this really is of high importance since, especially for the 2x2x2, the current records are almost irrelevant for cubing expertise. Even for the 3x3x3, it's obviously important since there's quite a large gap between the single-WR and the ao5-WR. Almost all single-WRs were lucky solves with move counts around 45 whereas CFOP, the exclusive method for WRs lately (or even ever since?) is ~ 55.

But we can't really force PLL skips to not happen as we don't know how different cubers solve their F2L or which OLL algs they use etc. What if someone knows all ZBLLs and skips PLL every time? I remember a thread some while ago where someone wrote code that produces natural looking CFOP solutions to any scrambles that always result in LL skip.

Ok, I probably just missed that you were being sarcastic. ._.

 

[Torch]

I totally agree and had this idea myself a while ago.

Since for the 2x2x2 and 3x3x3 all possible positions are known meanwhile, it could be accomplished to only allow scrambles that lead to positions that CANNOT be solved with (a) skip(s) by any method available. It could just be looked up from a database that contains only non-lucky-solvable scrambles.

I find this really is of high importance since, especially for the 2x2x2, the current records are almost irrelevant for cubing expertise. Even for the 3x3x3, it's obviously important since there's quite a large gap between the single-WR and the ao5-WR. Almost all single-WRs were lucky solves with move counts around 45 whereas CFOP, the exclusive method for WRs lately (or even ever since?) is ~ 55.

https://www.speedsolving.com/forum/threads/cfop-variation-cf.37349/

 

[DGCubes]

I totally agree and had this idea myself a while ago.
Since for the 2x2x2 and 3x3x3 all possible positions are known meanwhile, it could be accomplished to only allow scrambles that lead to positions that CANNOT be solved with (a) skip(s) by any method available. It could just be looked up from a database that contains only non-lucky-solvable scrambles.

I find this really is of high importance since, especially for the 2x2x2, the current records are almost irrelevant for cubing expertise. Even for the 3x3x3, it's obviously important since there's quite a large gap between the single-WR and the ao5-WR. Almost all single-WRs were lucky solves with move counts around 45 whereas CFOP, the exclusive method for WRs lately (or even ever since?) is ~ 55.

Um... how about a method like ZZ-CT, where every solve forces a last-layer skip? Or what if you use ZBLL or WV in a solve and force a "skip?" Or what if some guy didn't know any PLLs, so his method was to keep solving over and over until he got a PLL skip? Or what if you totally messed up and had to resolve F2L in an unconventional way that would lead to a skip? EVERY scramble can lead to a skip in EVERY method depending on how you solve it. Besides, WCA regulations absolutely must be method neutral, because methods change and evolve over time, and many methods have dozens of intricacies that make it hard to determine if the solver got a skip.

Say you use ZZ, and you only know COLL and EPLL algs for LL. There's a 1/12 chance of getting a PLL skip. Sure, you would be lucky if you got one, but making this illegal would make 1 in 12 of your solves also illegal (technically more if you also count OLL skips). That's just over one per competition if you compete in 3 rounds of 3x3. We already have movecount requirements that are relatively reasonable; it would be flat out impossible to illegalize skips.

 

[AlexMaass]

Since for the 2x2x2 and 3x3x3 all possible positions are known meanwhile, it could be accomplished to only allow scrambles that lead to positions that CANNOT be solved with (a) skip(s) by any method available. It could just be looked up from a database that contains only non-lucky-solvable scrambles.

No it can't, there was a thread in which someone demonstrated a program that would generate CFOP solutions to any scramble (https://www.speedsolving.com/forum/threads/cfop-variation-cf.37349/) and making it so that it would always end up with an LL skip, this is just not possible, you could do F2L differently and get a skip because solving the F2L differently can affect the last layer pieces. This is a really naive and dumb post IMO.

 

[YouCubing]

I totally agree and had this idea myself a while ago.

Since for the 2x2x2 and 3x3x3 all possible positions are known meanwhile, it could be accomplished to only allow scrambles that lead to positions that CANNOT be solved with (a) skip(s) by any method available. It could just be looked up from a database that contains only non-lucky-solvable scrambles.

I find this really is of high importance since, especially for the 2x2x2, the current records are almost irrelevant for cubing expertise. Even for the 3x3x3, it's obviously important since there's quite a large gap between the single-WR and the ao5-WR. Almost all single-WRs were lucky solves with move counts around 45 whereas CFOP, the exclusive method for WRs lately (or even ever since?) is ~ 55.

I don't understand you, your logic, or any way that would be possible

 

[Dash Lambda]

I find this really is of high importance since, especially for the 2x2x2, the current records are almost irrelevant for cubing expertise. Even for the 3x3x3, it's obviously important since there's quite a large gap between the single-WR and the ao5-WR. Almost all single-WRs were lucky solves with move counts around 45 whereas CFOP, the exclusive method for WRs lately (or even ever since?) is ~ 55.

You see, the thing about cubing, is that it's about three-quarters skill and one quarter probability. The average movecount for CFOP is 55, but they'll typically be anywhere from ~30-70 moves because that's how samples and averages work. If you don't like probability, then this really isn't the sport for you.

A good exercise I do is that, whenever there's a new WR, I get the scramble and do it myself before I watch the video. I think you'll find that, no-matter how lucky the solve, you probably can't do it as fast the guy with the record (except for 2x2 and pyra, but those are trivial).

 

[turtwig]

Even if you exclude a certain set of scrambles, within the scrambles that you allow, there will still be luckier ones, and some records would still be very hard to beat. For example, if we exclude all possible skips (it's impossible, but just for the sake of argument), someone could still get a 1 move cross, 4 3-move pairs, sune and U perm, which would then become virtually unbeatable.

I personally don't think that we should filter any scrambles (except solved cubes and +2's, obviously), but filtering by minimum movecount is much better than filtering by arbitrary things like skips which is impossible and doesn't make any sense.

 

[biscuit]

*some 15 second solver gets an E1, gets a two move solution that no one else gets.*

*gets a record that will never be beaten*

That's the kind of scenario that makes me think no filtering is an awful idea, even if it is an arbitrary limit.

 

[DGCubes]

*some 15 second solver gets an E1, gets a two move solution that no one else gets.*

*gets a record that will never be beaten*

That's the kind of scenario that makes me think no filtering is an awful idea, even if it is an arbitrary limit.

Totally agree with this. The fact that 2 move scrambles are allowed at all (and that this is true all the way up to 7x7) just seems wrong, especially when 2x2 scrambles are filtered out at sub-4 moves.

 

[AlexMaass]

Totally agree with this. The fact that 2 move scrambles are allowed at all (and that this is true all the way up to 7x7) just seems wrong, especially when 2x2 scrambles are filtered out at sub-4 moves.

I don't think this is true though, can Lucas Garron verify?

 

[DGCubes]

I don't think this is true though, can Lucas Garron verify?

Yeah, idk what TNoodle actually filters out, but the regs say:

4b3) Specification for the scramble program: An official scramble sequence must produce a random state from those that require at least 2 moves to solve (equal probability for each state). The following additions/exceptions apply:

which applies for 3x3 and 4x4, and

4b3e) 5x5x5 Cube, 6x6x6 Cube, 7x7x7 Cube, and Megaminx: sufficiently many random moves (instead of random state), at least 2 moves to solve.

Although I would like some insight on how TNoodle handles this (especially the "equal probability for each state" part, which sounds like a 2-move solution should be included in the pool of possible scrambles).

 

[AlexMaass]

Yeah, idk what TNoodle actually filters out, but the regs say:

which applies for 3x3 and 4x4, and


Although I would like some insight on how TNoodle handles this (especially the "equal probability for each state" part, which sounds like a 2-move solution should be included in the pool of possible scrambles).

I've heard TNoodle does filtering above 2 moves for 3x3, idk what it is for other puzzles

misscrambles are allowed on 7x7 so a delegate who gave 0 ****s could just let a 2 mover happen

 

[Dash Lambda]

*some 15 second solver gets an E1, gets a two move solution that no one else gets.*

*gets a record that will never be beaten*

That's the kind of scenario that makes me think no filtering is an awful idea, even if it is an arbitrary limit.

I don't know about other people, but if I got that scramble, I'd just do a DNF. Maybe solve it, do a superflip, then stop the timer.

 

[WACWCA]

I don't know about other people, but if I got that scramble, I'd just do a DNF. Maybe solve it, do a superflip, then stop the timer.

I have a feeling like 60% of the people would think it was a mistake, and it wasnt scrambled, then would send it back to find out it was the real scramble and then get an extra scramble or a DNF, idk how that would work

 

[biscuit]

I have a feeling like 60% of the people would think it was a mistake, and it wasnt scrambled, then would send it back to find out it was the real scramble and then get an extra scramble or a DNF, idk how that would work

I'd probably think it's a mistake too. However, if you ever get something crazy like that, notify your judge that you think there's a mistake, solve it as normal, then go talk to the delegate.

 

[AlphaSheep]

To put this into perspective, a 2 move solve is so ridiculously incredibly unlikely, and we're not talking one in a billion unlikely. We're talking in the order of a billion times less likely than that. If you had a unique scramble for each star in the Milky Way galaxy, theres still a better than 999,999 in a million that you would not be able to solve any of them in 2 moves.

I do think the limit should be higher for 3x3 though, but more for thoroughness than I think it will ever actually happen.

Although I would like some insight on how TNoodle handles this (especially the "equal probability for each state" part, which sounds like a 2-move solution should be included in the pool of possible scrambles).

The cool thing is that TNoodle is open source, so if you know some programming, you can dig in the code and see exactly how it works. It's not the most straight forward code to read, but it's not impossible to decipher. To understand the scrambling, you'd want to look at the .java files in the scrambles/src/puzzle/ folder. Specifically, all puzzles with limits different from the default define a variable wcaMinScrambleDistance, in accordance with the limits in the regulations.

The only one that's hard to figure out is the limit for 3x3 because of the way it's set. Fortunately there's an automatically generated readme that displays the limits for all puzzles. When you run TNoodle, there's a question mark in the top right corner. If you click that, you'll see what limits it was using, and at the moment, it uses 2 moves for 3x3.

Regarding selecting states so that each state has equal probability, yes scrambles under the limit are originally included in the pool of possible scrambles. It essentially has a way of numbering every single possible unique state (including those that can be solved under the move limit, and yes, even the solved state) from 0 to n and picks a random number. Here's the code that generates the state for pyraminx. If it turns out the state it picked can be solved in fewer moves than the limit for that puzzle, it scraps that state, picks a new one and tries again.

That doesn't affect the relative probabilities of the valid states at all. Imagine if you roll a dice, and if you roll a 1, you get to roll again until you get something other than 1. The probabilities of rolling a 2, 3, 4, 5 or 6 are still all equal.

It is perfectly possible that the program will keep generating the scrambles under the limit which would basically cause the program to freeze. But that's so incredibly unlikely you can virtually guarantee that that won't happen. Your computer's far far more likely to be struck by lightning while generating the scrambles.