[Unofficial] Mike Hughey megaminx BLD 1:18:00.87

[Mike Hughey]

Memorization time was 36:30. I say that at the beginning, in case you don't feel like reading it.

Video is sped up 8x so it will fit. Sorry for the borders on the side - my youngest daughter was apparently playing with the camera again, and I didn't notice until this morning.

I'm using a chinaminx - it's well-broken-in, but even so it's pretty bad for BLD, since I do a lot of "D" moves (as used in official scrambles) (by using these as my first setup move, I can move a whole set of pieces to a place where I know the setup moves for each possibility, allowing me to reduce 30 setup possibilities to just 10 or so), and it's hard to keep a finger on the piece I'm working on while turning the whole bottom of the puzzle, and it often locks up a bit in the process. A Meffert's would probably help a lot, but I'm too cheap to get one.

I'll probably keep trying this until I get sub-hour. Then I'll write up my whole method for doing it and post it here. It's a pretty weak method - I have to alternate between two locations to shoot from and perform an inverse algorithm every other time for edges - like M2 but without the advantage of the piece to shoot from moving to the same starting location each time. But it does work!

 

[Samlambert]

Wow that was awesome, congratulations! =)

 

[Escher]

i love your laugh at the end, it makes me happy

and you are extremely impressive!

 

[Derrick Eide17]

Mike is getting too awesome!
Just great

 

[Stefan]

Frack. I better undust my pyraminx crystal right away.

And don't call that a weak method, it's what I'd use for megaminx as well and what I've been sightedly practicing for pyraminx crystal.

 

[Mike Hughey]

Frack. I better undust my pyraminx crystal right away.

And don't call that a weak method, it's what I'd use for megaminx as well and what I've been sightedly practicing for pyraminx crystal.

Thanks - I was looking forward to hearing from you. I just assumed you had probably come up with a better way using some smarter algorithm. But I've thought about it a bit and I'm not sure I can see an easy way to come up with a smarter algorithm. (I still think it might exist, but I think it's also possible it might not exist.)

I use a 12-move algorithm for 3-cycling the edges all on the top face, and then the basic corner permuter that I learned from Erik Akkersdijk's website for permuting corners at the end of the last layer for 3-cycling the corners. For the corners, I use 2 of the top corners as my fixed buffers and then freely set up to whichever of the other 3 top corners is most convenient, but for the edges, I always set up to the same place.

You don't have to worry about much from me for pyraminx crystal - I have no immediate plans to get one. But if someone wants to give me one..... (Honestly, I'm just kidding - please nobody offer me one! I can go find and buy one if I really want to try it.)

 

[masterofthebass]

I just thought about a method for this. You can do the Y perm method quite well on megaminx. A Y perm effects 5 adj pieces on the U face and then 1 separate corner. If you use the standard Y perm alg and the inverse, you can just use that like you would on 3x3 . Now I have to come up with a way for edges, and MAYBE I'll try this.

 

[Stefan]

I think by setting up to the same layer as the two buffers you're using unnecessarily many moves. Both setup moves as well as swap alg moves. I might use eight moves commutators for corners and 10 moves commutators for edges.

For corners you could also use a double swap instead, let's say (R U R' U')*3 or (R U' R' U)*3. That's fast and you don't need to alternate between forwards and backwards. Main effect would be swapping two corners, side effect swapping to others.

 

[blah]

I think by setting up to the same layer as the two buffers you're using unnecessarily many moves. Both setup moves as well as swap alg moves. I might use eight moves commutators for corners and 10 moves commutators for edges.

For corners you could also use a double swap instead, let's say (R U R' U')*3 or (R U' R' U)*3. That's fast and you don't need to alternate between forwards and backwards. Main effect would be swapping two corners, side effect swapping to others.

Didn't you mention this elsewhere?

 

[Dene]

WOW! That was amazing, like, the fact that you even tried is crazy, but actually finishing is insane!

 

[tim]

Awesome Mike! Somehow i wish i would own a megaminx.

Was this your first attempt?

 

[Mike Hughey]

Awesome Mike! Somehow i wish i would own a megaminx.

Was this your first attempt?

Thanks!

It was my second try. The first one had 9 edges wrong and 4 corners in the correct place but misoriented. I practiced a couple of solves sighted between the first and second attempts, which made the second attempt much easier and better.

Tim, you should go buy one of the cheap megaminxes - they're really quite inexpensive.

 

[Mike Hughey]

I think by setting up to the same layer as the two buffers you're using unnecessarily many moves. Both setup moves as well as swap alg moves. I might use eight moves commutators for corners and 10 moves commutators for edges.

For corners you could also use a double swap instead, let's say (R U R' U')*3 or (R U' R' U)*3. That's fast and you don't need to alternate between forwards and backwards. Main effect would be swapping two corners, side effect swapping to others.

I didn't want to respond to this until I got to go home and try some of it. I'm very stupid - I didn't know that 8 move commutators for corners were even possible, but when you mentioned it, I tried it and in just a few seconds discovered that obviously some that are similar to those on cubes work just fine. I can see where it makes a lot of sense to switch to that.

But I can't find any 10 move commutators for edges. I'm just not very good at finding commutators on a megaminx. Would you please give an example of a 10 move commutator for edges? It might be very helpful.

The 8 move commutator for corners would work very nice, and alternating forwards and backwards is easy because it's the same moves in a slightly different order. But I can also see where the double swap might work well; I'll experiment with both. But they won't speed up the solve all that much, because corners are already much easier and faster than edges. I mainly need to improve the edges - they're probably 2/3 to 3/4 of the solve.

 

[Stefan]

Sorry, my mistake, 12 moves for edges. I was thinking of this kind: [ L' R F2 R' L, U ]. Maybe there's something shorter but I don't know it.

"The" 8 move commutator sounds wrong. You could use at least these two (also work on 3x3x3):
[ R' D' R, U ]
[ F D F', U ]

They cycle the same two U stickers with two different stickers on the same corner in D. So you get easier setup moves because you cover two orientations instead of just one. If you add another alg to shoot to the third sticker on the same corner, you'd just need to setup a target into that corner position in any orientation and then apply the appropriate alg.

Btw, I thought about memorizing not which sticker to target, but which piece and what the orientation would be. So instead of memorizing let's say blue/orange(/white) meaning the blue sticker of that piece, memorize that you need this piece and which of the three algs to use. So if that piece were "K" for me, I'd memorize K1, K2, or K3. And then for let's say K2, during execution I'd "setup piece K" and apply algorithm 2. Reduces the number of setup algs for corners from 54 to 18 and for edges from 56 to 28. You have to take care of the orientations at some point, and this way you'd explicitly do it while you can still see the puzzle. On the other hand, you could also learn that blue/orange=K2 and orange/white=K3. So yeah... just a thought.

 

[Mike Hughey]

Sorry, my mistake, 12 moves for edges. I was thinking of this kind: [ L' R F2 R' L, U ]. Maybe there's something shorter but I don't know it.

Okay, that's not too different from what I was using.

"The" 8 move commutator sounds wrong. You could use at least these two (also work on 3x3x3):
[ R' D' R, U ]
[ F D F', U ]

They cycle the same two U stickers with two different stickers on the same corner in D. So you get easier setup moves because you cover two orientations instead of just one. If you add another alg to shoot to the third sticker on the same corner, you'd just need to setup a target into that corner position in any orientation and then apply the appropriate alg.

The first one is the one I found last night. The other is also pretty obvious, I guess. Yes, I can see where the setups are much easier. Not only are they shorter, but they're also a bit easier to see. I'm really probably going to switch to this next time.

Btw, I thought about memorizing not which sticker to target, but which piece and what the orientation would be. So instead of memorizing let's say blue/orange(/white) meaning the blue sticker of that piece, memorize that you need this piece and which of the three algs to use. So if that piece were "K" for me, I'd memorize K1, K2, or K3. And then for let's say K2, during execution I'd "setup piece K" and apply algorithm 2. Reduces the number of setup algs for corners from 54 to 18 and for edges from 56 to 28. You have to take care of the orientations at some point, and this way you'd explicitly do it while you can still see the puzzle. On the other hand, you could also learn that blue/orange=K2 and orange/white=K3. So yeah... just a thought.

Definitely an interesting idea! For edges this would be difficult on memorization, though, since it would leave slightly more than 26 pieces, so the alphabet doesn't work. Messes up my image memorization - I'd have to retool it a bit. But for corners it seems like it might be nice.