Wetten, dass..? vom 5. November

Several people have asked me recently how to solve the Megaminx faster and I decided to write a little page about it instead of repeating myself in emails further.

Erik Akkersdijk is now the world's second fastest megaminxer very close behind me and he has a very good megaminx method page, too.

First of all you need a good Megaminx. I use the beautiful tiled Meffert one. Sadly they have too strong and long springs which make the puzzle very stiff and can sometimes even drive out the screw and center piece during normal play. I replaced the original springs with shorter and softer ones with ends ground flat and also added some washers to adjust the tension (I have about six washers in each center). First I stole the springs from two cheap 3x3 clones, later I ordered some from a spring manufacturer. Oh, and of course you should also lubricate your megaminx.

Here's how to open the Megaminx. The mechanism ist just like 3x3 but more of everything. First take out an edge, use a screwdriver if necessary.


When you have good access to a center, remove the cap with a thin blade. There's probably not much glue so you don't need much force. Still, be careful.


This is how the center and its cap look like, just so you know how deep you can stick the blade in without damaging anything.


When I'm done I drive in the screw until it's flush with the top of the center, i.e. if I'm looking flat over the center top (here I'm at an angle) then I make the screw just disappear.


Corners and centers are pretty safe from lockups but edges can cause some if you don't align the layers properly. Of course you should always over-/underturn the layers appropriately for the following twist, but...


... you can also sand the inside edges of the edge pieces more round to protect you from lockups a bit.


Now about the actual solving. I pretty much use Grant Tregay's method. At the bottom of his page you can see some statistics. I've done some today, too, solved three times at almost full speed so I could count the moves. If you use the same method, maybe these stats can help you determine where you waste the most moves/time. In the table you see #moves/seconds pairs.

Step Solve 1 Solve 2 Solve 3 Average Average in %
Face 1 47 / 30.99 49 / 32.71 41 / 23.68 45.7 / 29.13 24.9% / 30.33%
Face 2 24 / 14.29 25 / 16.17 30 / 19.12 26.3 / 16.53 14.4% / 17.21%
Face 3 18 / 8.72 17 / 9.78 19 / 9.9 18.0 / 9.47 9.8% / 9.86%
Face 4 19 / 8.08 15 / 10.09 21 / 9.44 18.3 / 9.20 10.0% / 9.58%
Faces 5-6 30 / 14.23 33 / 16.53 26 / 11.84 29.7 / 14.20 16.2% / 14.79%
Last Face 44 / 17.17 47 / 17.59 45 / 17.78 45.3 / 17.51 24.7% / 18.24%
Total 182 / 93.48 186 / 102.87 182 / 91.76 183.3 / 96.04 100.0% / 100.00%

I twist exclusively with my left hand, the right hand only holds and rotates the puzzle as a whole which you can see in my video of a 1 minute 21.67 seconds solve. For that reason, my last layer algs are mirrors of Grant's.

I always solve the faces in the same order, i.e. always white first, then turquoise, etc. During one face I solve the star edges in the order I see them and then also the corner/edge pairs in the order I see them.

Notice Grant doesn't say so, but the last layer algorithms for OE and PE are very easy to understand, just like the PC approach. No need to learn those algs by rote. The OE flips the edges one by one, first flipping one, then replacing it by another which gets flipped on the way back. For PE, rotate the last layer so that the UF edge is solved and the UR edge isn't. The first R+ brings a solved non-last-layer corner/edge pair next to the solved UF edge and together they build a triple that sticks together during the whole algorithm until it comes apart by the R- at the very end. Notice the first R+ also pushes the unsolved UR edge out of the last layer, then the last layer gets rotated and an R- brings the edge back into the last layer "where it belongs" (relative to the edge in the triple). Try to understand how PE works this way and you don't need to learn algs by rote. Also, you can align the last layer and puzzle so that you always have a 3-cycle except the one double-swap case Grant shows.

OC is a bunch of (some of them conjugated) commutators for corner-3-cycles, I actually invented equivalent algs myself to fit my twisting style better. If you see them this way maybe it's easier for you to learn them or you can make up your own algs, too. Except in rare cases in some steps I could tell you the exact purpose of every single twist I make, that's btw one reason I love the Megaminx and the reason I don't like Square-1 much. I don't want to learn many algs that I don't understand in order to solve fast.

Stefan Pochmann
Last modified: March 20 2007, 21:35:15