[WCA Regulations 2014] Scramble Filtering (new poll)

[Methuselah96]

Here is what I agree and disagree with:


BLD events: Adds yet another element of luck. We don't need that.
2x2: Agreed. This also means the single can still be broken.
Pyraminx: Limiting solutions to 7 moves when God's number is 11 is too high. Maybe something like 5.
Skewb: No idea
Sq-1: No idea
5x5, 6x6, 7x7: Kind of goes without saying.

Pyraminx: I still don't know if they include tips as moves (which I'm assuming right now that they do), but if they do God's number is 15.
Skewb: God's number for Skewb is also 11. It was meant to match Pyraminx somewhat.

 

[tseitsei]

I am against scramble filtering.

Mainly because I don't think good "limits" can be assigned for each puzzle.
Obviously some scrambles will always be easier than other and some scrambles will be totally ridiculous. And to be honest I think that for example a 4 move scramble for 2x2 is (almost) as ridiculous as 2 move scramble and wr single is impossible to beat with a normal or even a good scramble either way... The scramble has to be ridiculous to even have a chance at wr...

Also the length of the scramble isn't the only criteria to block all possible lol scrambles:

a) now scramble your 3x3 cube with 14 random moves (or take a random state scramble that's ~14 moves long if you can find one somewhere that's even better).
Scramble probably looks and feels completely legit if you solve it.

b) now do an e-perm. Optimal e-perm is also 14 moves long so a random state scrambler with requirement for >13 move states could give you a simple pll as a scramble. Which would obviously result to an "unbeatable WR-single with any normal scramble"

Also just had another thought: if we want to filter out the "easy" short scrambles, then shouldn't we also filter out the long 19 or 20 move "hard" scrambles?
Just to be fair... It wouldn't be fair if someone got a really really hard scramble just like it wouldn't be fair if someone got a really really easy scramble...

 

[Daniel Wu]

Pyra though... Why so high?

 

[qqwref]

I think if we are going to filter scrambles we should only filter out the most extremely easy scrambles, so easy that you are unlikely to even see such a scramble in an average of 100 solves. I'm thinking on the order of one in a thousand or one in ten thousand (or less). 2.6% (for Skewb) is way too high.

For reference, here are some God's Algorithm tables for 2x2x2, Pyraminx, Skewb, and Square-1 (and part of the 3x3x3 one):

Spoiler

[B]2x2x2 (HTM):[/B]
Moves  Positions
0              1
1              9
2             54
3            321
4           1847
5           9992
6          50136
7         227536
8         870072
9        1887748
10        623800
11          2644
Total    3674160

[B]Pyraminx (HTM, no tips):[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5           9896
6          51808
7         220111
8         480467
9         166276
10          2457
11            32
Total     933120

[B]Pyraminx (HTM, with tips):[/B]
Moves  Positions
0              1
1             16
2            136
3            896
4           5456
5          32296
6         182432
7         931983
8        3829067
9       11108868
10      20736353
11      22907032
12      13067528
13       2739808
14         40336
15           512
Total   75582720

[B]Skewb[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5          10248
6          59304
7         315198
8        1225483
9        1455856
10         81028
11            90
Total    3149280

[B]Square-1 (twist metric)[/B]
Moves    Positions
0                1
1               64
2             1153
3            17050
4           235144
5          3091458
6         38893230
7        452031138
8       4459167504
9      33671064770
10    149502310936
11    183662070768
12     63945120032
13       157452752
Total 435891456000

[B]3x3x3 (HTM)[/B]
Moves        Positions
0                    1
1                   18
2                  243
3                 3240
4                43239
5               574908
6              7618438
7            100803036
8           1332343288
9          17596479795
10        232248063316
11       3063288809012
12      40374425656248
13     531653418284628
14    6989320578825358
15   91365146187124313
...
= 43252003274489856000

 

[Methuselah96]

That pyra filter rules out about 6.4% of all possible positions. I don't like that.

It is not 6.4% for Pyraminx. It is 0.29270314696%. It would be around 6% if you didn't include tips.

I think if we are going to filter scrambles we should only filter out the most extremely easy scrambles, so easy that you are unlikely to even see such a scramble in an average of 100 solves. I'm thinking on the order of one in a thousand or one in ten thousand (or less). 2.6% (for Skewb) is way too high.

For reference, here are some God's Algorithm tables for 2x2x2, Pyraminx, Skewb, and Square-1 (and part of the 3x3x3 one):

Spoiler

[B]2x2x2 (HTM):[/B]
Moves  Positions
0              1
1              9
2             54
3            321
4           1847
5           9992
6          50136
7         227536
8         870072
9        1887748
10        623800
11          2644
Total    3674160

[B]Pyraminx (HTM, no tips):[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5           9896
6          51808
7         220111
8         480467
9         166276
10          2457
11            32
Total     933120

[B]Pyraminx (HTM, with tips):[/B]
Moves  Positions
0              1
1             16
2            136
3            896
4           5456
5          32296
6         182432
7         931983
8        3829067
9       11108868
10      20736353
11      22907032
12      13067528
13       2739808
14         40336
15           512
Total   75582720

[B]Skewb[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5          10248
6          59304
7         315198
8        1225483
9        1455856
10         81028
11            90
Total    3149280

[B]Square-1 (twist metric)[/B]
Moves    Positions
0                1
1               64
2             1153
3            17050
4           235144
5          3091458
6         38893230
7        452031138
8       4459167504
9      33671064770
10    149502310936
11    183662070768
12     63945120032
13       157452752
Total 435891456000

[B]3x3x3 (HTM)[/B]
Moves        Positions
0                    1
1                   18
2                  243
3                 3240
4                43239
5               574908
6              7618438
7            100803036
8           1332343288
9          17596479795
10        232248063316
11       3063288809012
12      40374425656248
13     531653418284628
14    6989320578825358
15   91365146187124313
...
= 43252003274489856000

Thanks for the tables. I had just finished figuring out the pyraminx with tips. It was worth figuring it out myself anyway .

 

[Kirjava]

I've always been against scramble filtering, but a 7 move limit for Skewb is just nuts.

 

[Username]

It is not 6.4% for Pyraminx. It is 0.29270314696%. It would be around 6% if you didn't include tips.

Yeah I realized that. I feel like an idiot. Last time I'm argumenting against something ever.

 

[KongShou]

Why all the unnecessary regulations all of the sudden?

 

[Methuselah96]

If want something done about this we should come up with a list of clear arguments and also decide exactly what we're proposing so that we can e-mail the WCA board about it requesting that they consider our proposal.

 

[Owen]

Pretty bad idea. I think luck is an important part of the game.

 

[Stefan]

This scrambling filter has been in effect for a long time

Just to clarify: The Skewb filter is new and the 11 moves for Square-1 were introduced only recently (previously the general 2 moves).

if we want to filter out the "easy" short scrambles, then shouldn't we also filter out the long 19 or 20 move "hard" scrambles?

No. The point is to have proper scrambles. A one or two moves "scramble" isn't really a "scramble" at all. The other extreme (19 or 20 moves) doesn't have that problem.

 

[Methuselah96]

I think if we are going to filter scrambles we should only filter out the most extremely easy scrambles, so easy that you are unlikely to even see such a scramble in an average of 100 solves. I'm thinking on the order of one in a thousand or one in ten thousand (or less). 2.6% (for Skewb) is way too high.

For reference, here are some God's Algorithm tables for 2x2x2, Pyraminx, Skewb, and Square-1 (and part of the 3x3x3 one):

Spoiler

[B]2x2x2 (HTM):[/B]
Moves  Positions
0              1
1              9
2             54
3            321
4           1847
5           9992
6          50136
7         227536
8         870072
9        1887748
10        623800
11          2644
Total    3674160

[B]Pyraminx (HTM, no tips):[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5           9896
6          51808
7         220111
8         480467
9         166276
10          2457
11            32
Total     933120

[B]Pyraminx (HTM, with tips):[/B]
Moves  Positions
0              1
1             16
2            136
3            896
4           5456
5          32296
6         182432
7         931983
8        3829067
9       11108868
10      20736353
11      22907032
12      13067528
13       2739808
14         40336
15           512
Total   75582720

[B]Skewb[/B]
Moves  Positions
0              1
1              8
2             48
3            288
4           1728
5          10248
6          59304
7         315198
8        1225483
9        1455856
10         81028
11            90
Total    3149280

[B]Square-1 (twist metric)[/B]
Moves    Positions
0                1
1               64
2             1153
3            17050
4           235144
5          3091458
6         38893230
7        452031138
8       4459167504
9      33671064770
10    149502310936
11    183662070768
12     63945120032
13       157452752
Total 435891456000

[B]3x3x3 (HTM)[/B]
Moves        Positions
0                    1
1                   18
2                  243
3                 3240
4                43239
5               574908
6              7618438
7            100803036
8           1332343288
9          17596479795
10        232248063316
11       3063288809012
12      40374425656248
13     531653418284628
14    6989320578825358
15   91365146187124313
...
= 43252003274489856000

Do you have the stats for Square-1 if you count (x, y) as a move and / as a move. Would that be just doubling the twist metric table?

 

[tx789]

7 moves for pyraminx is high but does that include tips? But Skewb 7 moves is very very high since ~2% of scrambles are now unusable in comp. Filtering scribbles is gone for completely stupid 1 move scrambles or something like that.

 

[XTowncuber]

errrm, odder's WR pyra single is 6 moves...I don't think I should have to be less lucky than him to get a WR. 5-6 moves limit seems more reasonable.

 

[Pro94]

I agree with scramble filtering but I think regulations need some changes:
I'd change Pyraminx 7 moves for two reasons:
1. WR single had a 6 moves solution (without tips);
2. Almost 0.3% filtering is quite high in my opinion.
So 6 moves for Pyraminx should be good.

Same thing for Skewb: 2.2% is too high. If what Sarah said is right, 6 moves (0.04%) would be much better.

 

[Erik]

Like I said in the other thread: we have averages of 5 to filter skill. Luck is part of the cube and it should not be influenced.

Imho a 1 move scramble and solution is no worse than a 4 move scramble and solution. Both extremely lucky and unbeatable on a 'harder' scramble. Why filter one of them? I still see 0 reason for filtering anything.

The "problem" of easy and lucky scrambles does not exist in my opinion. Also the "solution" filtering, does not solve your so-called problem either.

 

[DavidCip86]

Why does it specify that 5, 6 and 7 should be solvable in 2 or more moves?

 

[Stefan]

Imho a 1 move scramble and solution is no worse than a 4 move scramble and solution. Both extremely lucky and unbeatable on a 'harder' scramble. Why filter one of them? I still see 0 reason for filtering anything.

A one move "scramble" is considered solved with a time penalty (+2 seconds), so because of that, it's kinda officially not scrambled. And our puzzles should get scrambled.

Why does it specify that 5, 6 and 7 should be solvable in 2 or more moves?

It doesn't. It says "at least two moves to solve". Reason is above.

 

[qqwref]

Do you have the stats for Square-1 if you count (x, y) as a move and / as a move. Would that be just doubling the twist metric table?

No, unfortunately I don't have this, and it may not be known. It is not as simple as doubling the twist metric table because (x,y) is counted as 1 if one of them is zero and 2 otherwise. So the table would have to be computed from scratch, and from what I've heard, generating the original table took a year (in 2005).

 

[Methuselah96]

I'm probably not understanding this problem correctly, but why do 2x2, pyraminx, square-1, and skewb have such a high filter as compared to 3x3. Actually, my real question is why doesn't 3x3 have a higher filter? Isn't the purpose to not let extremely easy (2 moves away) cases happen? Why is 3x3 allowed 2 moves, but other puzzles are not?

No, unfortunately I don't have this, and it may not be known. It is not as simple as doubling the twist metric table because (x,y) is counted as 1 if one of them is zero and 2 otherwise. So the table would have to be computed from scratch, and from what I've heard, generating the original table took a year (in 2005).

That's what I thought. Thanks anyway. There's talk about changing the metric for Square-1 in any case.