What do you think is the absolute human limit of 3BLD?

[Marcell]

If we could memo piece positions instead of tracing cycles, we wouldn't have to constantly turn the cube around while memoing, which could save seconds.

Also, Kirjava's idea sounds very interesting.

 

[Noahaha]

I'm suprised you think 25, I mean, we're almost there now!

I mean global average for one person, not WR single.

 

[antoineccantin]

Who knows. Someone could just be really good at tracking pieces and be able to sub-10 speedBLD memo, then use a speed efficient method for execution.

 

[DrKorbin]

<M,U>-gen 5-cycles

 

[Ollie]

<M,U>-gen 5-cycles

This is awesome, but what's the probability of getting a 5-cycle from that list in a typical 3BLD solve?

 

[A Leman]

<M,U>-gen 5-cycles

Wow. thanks. I may go through some of that for fun sometime.

I personally can't judge what a limit would be. I think 3style is already very good and differences now could just come from an easy scramble, solving isolated cycles, or just being really comfortable with a memo method.

 

[DrKorbin]

This is awesome, but what's the probability of getting a 5-cycle from that list in a typical 3BLD solve?

The probabilty that one 5-cycle is MU-generated is 1/66 = (10*8*6*4) / (22*20*18*16). You can also do setups to these cases.
On the other hand, 3-cycles are mostly intuitive and easy to learning, but I see no logic in these algs for 5-cycles.

 

[CubeRoots]

I mean global average for one person, not WR single.

so do I! Marcell appears to be low 30s maybe sub 30 global

 

[irontwig]

I've had thoughts about a reduction based system. Something like;

Reduce to some given state via some set of steps -> use translation matrix based on exec during steps taken (this table will probably be huge and the biggest issue) to convert initial memo into current state memo. -> X-look reduced state.

You are solving more things then usual with each alg. The matricies are pre traced so to not add an overhead to memo.

This is 1 phase reduction - 2 phase may be more feasible.


Solving methods are so good these days that ideas for completely new ones that could be better become quite abstract.

Sorry, but I can't see how this would be feasible at all.

 

[Kirjava]

why?

 

[A Leman]

why?

How would this be faster than cycles? It would require pauses during execution and a serious amount of thinking. I have thought about this for a couple hours and I am also wondering how such a method could be learned intuitively. I can memo edges faster now than I could prepare something like a 2 look lse. Also, How would a reduction method like this be memorized? As you can probably notice, I am very skeptical about your Idea, but your ideas are normally very interesting(even if they will never happen).

 

[Kirjava]

How would this be faster than cycles?

Lower movecount, potentially less memo.

It would require pauses during execution and a serious amount of thinking.

Why would it require pauses during execution? Translation matrix ensures minimal thinking required.

Think before you conject.

I have thought about this for a couple hours and I am also wondering how such a method could be learned intuitively.

It wouldn't be.

I can memo edges faster now than I could prepare something like a 2 look lse.

Fail to see the relevance.

Please be aware that this is purely suppositional.

 

[blah]

Ville once fit in M' U M U' into a solve.

I remember this.

 

[Cubo largo]

I think it's possible with insane memo and an absurd TPS

 

[blackzabbathfan]

I said sub-13. It might be a while but 4 second memo with sub-9 second execution might just be possible.

 

[irontwig]

why?

I just can't see a way to use a reduction based approach that would be both "light weight" enough and faster than 3-style.

 

[nlCuber22]

description of redux based system

An incredibly sweet idea, but will anybody take action and actually see if it has potential to be a better method than the ones we already have? (inb4 y not u feg)

Also know that I think it definitely could be, based on what little I know about 3bld.

 

[bobthegiraffemonkey]

Reduction sounds interesting, any specific ideas you have thought of that might work? I think this thread might turn into a discussion on new theoretical BLD methods.

I looked at 5-cycles before briefly, but I had more important things to work on. I was planning to get back to it soon. I was planning to learn the cases of DF and the U layer edges, which have the first or last two targets on the UF and UB pieces (but probably not those with both targets on U, that's a really easy 3-cycle case while the others are considerably more awkward). Might help with some awkward cases and I think they will be the easiest to consider setups to as well.

More on topic, I'd guess the limit would be low 20s average.

 

[evogler]

I imagine the very, very fastest people would have an advantage of a few seconds due to TPS? Not that the top bld people are slow by any means (Mr. Zalewski has #93 3x3 average), but if Marcell (best 3x3 average 13.29) and Feliks could have a baby, that person might do some damage...

 

[Noahaha]

I imagine the very, very fastest people would have an advantage of a few seconds due to TPS? Not that the top bld people are slow by any means (Mr. Zalewski has #93 3x3 average), but if Marcell (best 3x3 average 13.29) and Feliks could have a baby, that person might do some damage...

I agree. I think a lot of people get into BLD because they are not so fast at 3x3.