The Efficacy of ZBLL on Averages

[obelisk477]

I've seen a lot of people (CFOPers) saying they wanted to, or have started to, learn ZBLL on the forums, and it got me wondering on how many you actually have to learn to affect an average. The first thing you have to get numbers for is how often you get a cross on top in speedsolves. Normally this would be 25%, but since most people use some sort of edge control, I did an ao100 and found that it is more like ~28%.

Average of 5 (1 solve out of 5 being a ZBLL)
(20%/28%)*493 = 352 ZBLLs. This is how many you would need to learn to expect to maybe get a good single with a ZBLL in an official competition

Average of 12 (2 out of 12 being a ZBLL)
(17%/28%)*493 = 293 ZBLLs. This is how many you would need to learn to expect get a good single, and a counting single as well to potentially lower an average of 12.

I think these are interesting results, as you see people learning a few ZBLLs here and there in hopes of getting a good single. If you see how many you actually need to learn to consistently affect an average of 5 our 12, you start to notice that you have to commit to learning about 300 to start to really affect your times.

 

[OLLiver]

what if you use ZZ or ZB?

 

[obelisk477]

Right, I should've said on CFOP. But obviously for ZZ or full ZB it's 100%

 

[mark49152]

to expect to maybe get a good single

What does that mean? Better than 50/50?

 

[4Chan]

you have to commit to learning about 300 to start to really affect your times.

I respectfully disagree for the following reasons:

When high level CFOP users use ZBLLs, they're using algs with two main attributes:

1. Easy and obvious recognition due to distinct colour blocks.
2. Nice algs, since it turns out cases with blocks tend to be nicer than cases without blocks.


Therefore, by learning just 50-ishhh algs which appear 3-5% of the time, those cases drop the times for those singles moreso than traditional OLL/PLL.
Outside of those nice obvious cases, there's diminishing returns, which is why a lot of the high level people haven't done the full thing.
If they actually wanted to, they totally could.

EDIT: If you redo what you did and try to find statistics for "nice" cases with a big sample size, it'd be a pretty nice contribution for people who wanted to learn ancillary ZBLLs for CFOP! (But that's a lot of work LOLOL)

 

[Bindedsa]

Normally this would be 25%, but since most people use some sort of edge control, I did an ao100 and found that it is more like ~28%.

Unless I'm mistaken that number is 12.5% Where did you get 25% from? Also, 100 solves is a very small sample size and you should probably clarify how much edge control you use. Other than that I agree, you shouldn't learn a few ZBLLs and expect averages to drop, but there's are dozens of cases that are really easy to recognize and really fast that are worth learning no matter how many you know.

 

[obelisk477]

Unless I'm mistaken that number is 12.5% Where did you get 25% from? Also, 100 solves is a very small sample size and you should probably clarify how much edge control you use. Other than that I agree, you shouldn't learn a few ZBLLs and expect averages to drop, but there's are dozens of cases that are really easy to recognize and really fast that are worth learning no matter how many you know.

Yeah maybe the number is 12.5 instead of 25, I couldn't decide which at the time of writing, but either way I got 28% in practice, so it still works. And I think the ao100 was representative enough, I can't imagine getting wildly different numbers than that otherwise, although you seem to be implying that 28% sounds high (maybe you aren't). I sledgehammered all of the time when UF was flipped, used U2 R U' R' U' sledge maybe 40% of the time when UF and UB are flipped, always U' sledged when UB was flipped and there was a 2x2 block on U with the pair, and used only a few ZBLS that are easy and that I normally use in speedsolves.

Also do you know the most likely ZBLLs? I simplified and assumed all of them had equal probability for the thought experiment, but of course there are some with higher occurences than others.

 

[AlphaSheep]

Also do you know the most likely ZBLLs? I simplified and assumed all of them had equal probability for the thought experiment, but of course there are some with higher occurences than others.

I once wrote a program to work out the probability of each ZBLL case. Here are the results: http://sylvermyst.com/zbll/
There's a few symmetric cases that are less likely but the vast majority are equal.

 

[OLLiver]

I once wrote a program to work out the probability of each ZBLL case. Here are the results: http://sylvermyst.com/zbll/
There's a few symmetric cases that are less likely but the vast majority are equal.

This is actually beautiful