[Help Thread] ZZ and ZB Discussion

[aznanimedude]

don't know why i never did it that way, but that's sexy and i shall begin using that

 

[nickvu2]

I found this proposal for CFOP training very interesting.

http://www.speedsolving.com/forum/s...bing-training-that-yields-systematic-progress

You read at least the OP, right? Ok, then let's continue. How could we apply it to ZZ? He took data from hundreds of sub-10 cubing videos. We don't have that. Perhaps we could self report splits. Also, there are so many last slot and LL methods that people are using. How do we take that into consideration?

In my opinion, looking for trends in small data sets is better than looking for trends with no data at all. Maybe we should all do weekly splits and identify our approach on each step. Eg. EO+Line vs. EOLine; full left block + full right block vs. freestyle blocks vs. cross + slots. Throw it in a database and run queries to find out what our weaknesses are, or look at any other trends for that matter. Even data from sub-optimal sub-methods such as EOCross are worth collecting. We could show just how ineffective the sub-method is...or hell, maybe find out our assumptions were wrong.

Anyway, what do you all think? Is this worth the effort? Any other ideas on how to carry out this approach?

 

[uvafan]

I found this proposal for CFOP training very interesting.

http://www.speedsolving.com/forum/s...bing-training-that-yields-systematic-progress

You read at least the OP, right? Ok, then let's continue. How could we apply it to ZZ? He took data from hundreds of sub-10 cubing videos. We don't have that. Perhaps we could self report splits. Also, there are so many last slot and LL methods that people are using. How do we take that into consideration?

In my opinion, looking for trends in small data sets is better than looking for trends with no data at all. Maybe we should all do weekly splits and identify our approach on each step. Eg. EO+Line vs. EOLine; full left block + full right block vs. freestyle blocks vs. cross + slots. Throw it in a database and run queries to find out what our weaknesses are, or look at any other trends for that matter. Even data from sub-optimal sub-methods such as EOCross are worth collecting. We could show just how ineffective the sub-method is...or hell, maybe find out our assumptions were wrong.

Anyway, what do you all think? Is this worth the effort? Any other ideas on how to carry out this approach?

Sounds good, except for the fact that he took data from reconstructions - they are more accurate then just regularly timing yourself. How about we find someone who wants some reconstructing experience to reconstruct all sub25 ZZ people for like 5 solves, then see what all the splits are. I think that would be very interesting.

 

[Petro Leum]

Sounds good, except for the fact that he took data from reconstructions - they are more accurate then just regularly timing yourself. How about we find someone who wants some reconstructing experience to reconstruct all sub25 ZZ people for like 5 solves, then see what all the splits are. I think that would be very interesting.

id love to, but i am soooo busy atm, i woudl start around christmas, so you might as well let someone else do it.... i will give out some vids though

 

[aznanimedude]

yay i'd be useful then if it's sub25 solves :3, but then you'd all see how bad my eol splits really are >:X

 

[mDiPalma]

if you throw a few well recorded solves my way at any time, i wouldn't mind reconstructing them.

just send me a link to the video, the scrambles, and maybe include your generalized approach (coll, zzll, eo+line) just so i know how to label things.

 

[ThomasJE]

Yeah, I wouldn't mind reconstructing a few solves here and there. I did reconstruct asmallkitten's 9.85 Ao5; that may help. (reconstructions are in post #10)

 

[nickvu2]

Are reconstructions done for move counts, or just to pinpoint exactly when one step ends and another begins? Why are they needed? I don't see a reference to reconstructions in the OP.

 

[uvafan]

Are reconstructions done for move counts, or just to pinpoint exactly when one step ends and another begins? Why are they needed? I don't see a reference to reconstructions in the OP.

I was thinking to pinpoint exactly when one step ends and another step begins, which is very hard to do without reconstructions; the movecounts could be useful to, though - in the other thread, I believe he mentions using reconstructions of sub10 solvers.

EDIT: maybe we should just time splits and post them. Then we could calculate the average percentages and see what we need to improve on.

for splits, we could use eoline, eoline + 1 block, eoline + f2l, LL.

 

[ThomasJE]

for splits, we could use eoline, eoline + 1 block, eoline + f2l, LL.

Usually, I do EOL+1 block (usually 1x2x2), F2L (inc. EOLine), LL.

 

[nickvu2]

EDIT: maybe we should just time splits and post them. Then we could calculate the average percentages and see what we need to improve on.

for splits, we could use eoline, eoline + 1 block, eoline + f2l, LL.

Even though reconstructions sound more accurate, it does seem super tedious for whoever is doing them. My one concern with splits is that it doesn't account for recognition time when you do a step in isolation. Look ahead serves as inspection time, and some steps are harder to look ahead on. The LL split, for example, should there be inspection time? Maybe instead of just LL, do a full solve and subtract eoline+f2l. Or allow maybe 1 second inspection before LL to simulate look ahead.

I'd personally like to add a last slot split. I think a lot of people either already do, or are interested in a last slot method. So what about: eoline, eoline + 1 block, eoline + f2l - 1, eoline + f2l, eoline + f2l + LL (full solve)?

For 1st block, is there anyone who does 1x2x3 all at once, or is it safe to assume people go for 1x2x2 first?

 

[aznanimedude]

pretty sure it's safe to assume that, i personally have done 1x2x3 blocks straight off, but only once in a blue moon when the stars and planets had lined up for me to do such, usually just a 1x2x2 into a 1x2x3

 

[mDiPalma]

Even though reconstructions sound more accurate, it does seem super tedious for whoever is doing them. My one concern with splits is that it doesn't account for recognition time when you do a step in isolation. Look ahead serves as inspection time, and some steps are harder to look ahead on. The LL split, for example, should there be inspection time? Maybe instead of just LL, do a full solve and subtract eoline+f2l. Or allow maybe 1 second inspection before LL to simulate look ahead.

...

For 1st block, is there anyone who does 1x2x3 all at once, or is it safe to assume people go for 1x2x2 first?

1) use the breakdowns tab on ctimer.co.uk to get accurate splits times.

2) you should build both 123 blocks at the same time (at once). otherwise, you are not taking full advantage of the method.

 

[ThomasJE]

Even though reconstructions sound more accurate, it does seem super tedious for whoever is doing them.

Can't we just take stats from reconstructions that have already been done?

Also, I think we should have the time as sub-20, rather than sub-25. It just seems to make more sense to me, as sub-20 is seen as more of a 'goal' than sub-25.

 

[nickvu2]

Can't we just take stats from reconstructions that have already been done?

Also, I think we should have the time as sub-20, rather than sub-25. It just seems to make more sense to me, as sub-20 is seen as more of a 'goal' than sub-25.

Didn't realize you had already timed the reconstruction splits! ...oh, and proportions too! Why did you do EOLine+1 instead of just EOLine?

I agree that faster averages are more important. Sub-20 makes sense. Or perhaps weighing the data based on speed so that fast averages are more influential in the global proportions.

There doesn't need to be consensus on the approach we use to get the data. People who are excited about reconstructions can do that, while people excited about timing splits can do there thing. There are pros and cons to each one. It would be interesting to see how the 2 data sets compare. Thomas, I encourage you to continue exploring reconstructions. Let me know how I can help you.

 

[ThomasJE]

Didn't realize you had already timed the reconstruction splits! ...oh, and proportions too! Why did you do EOLine+1 instead of just EOLine?

Well some solvers do a 'X-EOLine', so that is more used. Also, that's what Brest does, so I stole it from him.

I agree that faster averages are more important. Sub-20 makes sense. Or perhaps weighing the data based on speed so that fast averages are more influential in the global proportions.

That could be done... but it would require some complex formulae.

There doesn't need to be consensus on the approach we use to get the data. People who are excited about reconstructions can do that, while people excited about timing splits can do there thing. There are pros and cons to each one. It would be interesting to see how the 2 data sets compare. Thomas, I encourage you to continue exploring reconstructions. Let me know how I can help you.

I think data from reconstructions and from timing splits can remain separate, but we could combine them to make a total average percentage. I think we should only take sub-20 singles instead of averages, but if there's an average with at least one sub-20 single, we can reconstruct that single and use that data.

I have set up a spreadsheet where we can put all this data, and it's here. Post reconstructed solves/time splits here and I'll add them, and then the more data we get, the more accurate averages we get. I've started us off with asmallkitten's 9.85 Ao5 I reconstructed.

 

[uvafan]

My splits(ao12s):
EOL+1: 4.75
F2L: 9.52
LL: 4.53

F2L + LL = 14.05

EOL+1/F2L = 49.9%

EOL+1/Total = 33.8%

F2L/Total = 67.8%

LL/Total = 32.2%

Compared to asmallkitten's solves, I guess this means that I should work on F2L, but not EOLine?

Let me know if you would be willing to reconstruct some solves to get more accurate splits or something.

 

[CubeRoots]

My splits(ao12s):
EOL+1: 4.75
F2L: 9.52
LL: 4.53

F2L + LL = 14.05

EOL+1/F2L = 49.9%

EOL+1/Total = 33.8%

F2L/Total = 67.8%

LL/Total = 32.2%

Compared to asmallkitten's solves, I guess this means that I should work on F2L, but not EOLine?

Let me know if you would be willing to reconstruct some solves to get more accurate splits or something.

Do you really do XEOLINE every time?! , I rarely manage to see one... I think you need faster ZZF2L..

mine are
EOLine: 4
ZZF2L: 8
LL: 5.
solve: 18/19 (cause pauses/recog)

on average

ZZF2L is the best thing about ZZ, you should be burning that ****.

should try for 4 7 4

 

[uvafan]

Do you really do XEOLINE every time?! , I rarely manage to see one... I think you need faster ZZF2L..

mine are
EOLine: 4
ZZF2L: 8
LL: 5.
solve: 18/19 (cause pauses/recog)

on average

ZZF2L is the best thing about ZZ, you should be burning that ****.

should try for 4 7 4

No I don't. That was what we we're supposed to record - EOLine + 1

EDIT: Also, keep in mind my F2L splits include EOLine.

 

[CubeRoots]

No I don't. That was what we we're supposed to record - EOLine + 1

EDIT: Also, keep in mind my F2L splits include EOLine.

whoops