Scramble Length

Tim (2011-11-13 20:28:57 +0000)
After a very short square-1 scramble came up at Cachan Open and a very short (world record) scramble came up at Trentin Open, in my opinion it is time to start filtering scrambles. The 2x2 and pyraminx world record singles were set on scrambles that were absurdly easy and short. I don't think that the World Record should depend so heavily on luck. As a delegate, I have no idea how to decide whether or not a scramble is "too easy". It seems that 4-move scrambles are acceptable, but Bob was instructed to remove a 2-move scramble that he generated. I don't normally read all of the scrambles that I generate, because I will be competing with most of those scrambles, and so I try to avoid looking at the scrambles as much as possible. Therefore, I have no way of telling what to do if an easy scramble came up, and I would likely not notice. At this point we have software to optimally solve 2x2, 3x3, pyraminx, and square-1, all in reasonable amounts of time. I don't honestly think that we will be able to come up with a description of what an "easy scramble" is, without disadvantaging one method or another. The best idea I have, which is completely method-neutral (please tell me if you disagree with this statement), is to filter by distance from solved. I'm arbitrarily proposing the following numbers: Square-1: minimum 9 in twist metric 2x2: minimum 7 HTM 3x3: minimum 16 HTM Pyraminx: minimum 7 turns These numbers are chosen to make roughly 95% of scrambles acceptable. Yes, I know that these numbers are arbitrary. Yes, I know that some easy scrambles (like R U2 R' U' R U' R' on 2x2) are valid by my definitions. I also realize that not all scrambles under the depths I proposed are easy. But I think that the current 2x2 world record is now at a point where only a small fraction of scrambles could be solved fast enough to beat it, and I don't think that luck should make it impossible for nearly all competitors to have a shot at the world record, while making it very easy for others to do so. So at this point, I think that any rule of this nature is better than no rule. Finally, I would like to argue against Skewb being made an official event, at least unless a suitable solution to this problem is found. We now have two events in which the single solve is absurdly fast and in which lucky scrambles are virtually required to set the world record. I don't think we should add a third.
Sebastien (2011-11-14 07:18:29 +0000)
What is missing in your suggestion is how to handle the current WRs. If nothing is done, then especially for 222 the current WR would stand forever. I think that the WCA must decide between 2 philosophies: - we accept that these singles are mostly a matter of luck. Following this we should [u:c9jdfroq]never[/u:c9jdfroq] remove any generated scramble, not even if it is 1 move. - we decide that we don't want to list lucky WRs, restrict the scramble similar to what Tim suggested and simply remove the single rankings for both events from the website. I'm quite indifferent between those for the moment.
StefanPochmann (2011-11-14 19:32:28 +0000)
There's a third option between those two: Set the limits so that the current records can still be beaten, for example four moves minimum for 2x2. Then we're safe against something as stupid as a two moves solution, and there's no reason to invalidate previous records. I used to be indifferent as well, though I'd say I'm now in favor of such limits mainly because I'd like to avoid the inevitable outrage that would follow a two moves solution or so, especially from others who had a similarly easy scramble removed before. A secondary reason is that scrambles already have been invalidated by some delegates and I don't like this being decided differently by different delegates. Implementing such limits should be done by the scramble programs, not by humans. And if we'd generally rather have let's say a seven moves minimum for 2x2, the program could allow one 4-mover and otherwise only 7-movers. Or something gradual, like requiring the sorted distances of the five scrambles to be at least 4, 5, 6, 7 and 8 moves. That would avoid very easy averages (like with several 4-movers).