Columns for Big Cubes

As far as I can tell, this method dates to mid 2007, when I posted it on the TwistyPuzzles forum (although I've made some small changes since then). I didn't realize it at the time, but this method has some interesting properties. It's a direct solve method, which means that all steps solve pieces directly, rather than reducing ororienting them and then fixing that later. It uses very few slice moves (almost all of them are at the end in the commutator stage). And, for most of the solve, the bottom center stays on the bottom of the cube, like the Fridrich method for 3x3x3.

This is a pretty simple step: simply build one center. You'll want to use the color you plan to put on the bottom for the rest of the solve (and so I'll call this the "D color" for the rest of the solve). I don't suggest being color neutral because the only option there is which center to start with, and that would only save a few seconds at best, at the expense of recognition on the rest of the solve. It will help to choose a color that's bright and easy to spot for this center, because in the next few steps you will be looking for edge pieces with this color on them, and that will be harder if this color blends in with the rest.

This step is only necessary for the 5x5x5, so if you're solving the 4x4x4 you should just skip it. Here, you want to pair up four inner columns; each one contains a middle edge and three centers, although the middle center is fixed in place, so there are really 3 movable pieces to place for each column. For reference, these columns will have three centers of the same color, and then an edge with both that color and the D color.

There are a few possible ways to do this, and I'm not sure which is best. Personally, I put the center from Step 1 on the left, and then pair up the two movable centers first using r, l, and U moves, and only rotations along the R-L axis. I try to do one block at a time, and once I've finished one I will rotate it so it won't be affected by r and l moves. This can usually be done efficiently and quickly, but sometimes there won't be enough pieces of the right color on the four available centers and you'll have to do extra rotations. I've calculated that this happens about 5.4% of the time, so it will usually not be a problem. When you're done with the center parts, find the four middle edges, and for each center block pair up the edge and then put the column in the right place.

From now on, you're going to have the Step 1 center on the bottom of the cube. In this step, you build columns of 4 pieces: three centers of the same color and one edge with both that color and the D color. There are 8 columns to build - two of each color. Note that there are two edges for each color, and one can only be on the left and the other can only be on the right. Because each edge can be paired up with any center block you've made, I suggest pairing up an entire column first, and then inserting the whole thing in the proper slot afterwards.

In general, you'll be using r Ux r' type triggers to both pair up and insert the blocks of centers. It's your choice how to add in the edges, but there are actually a lot of different ways to efficiently move pieces around and pair up the blocks. I could list several of them, but if you play around with it yourself you'll end up with a much better intuitive understanding of this step. Avoid slice moves unless you have to, though. On the 4x4x4, you can actually insert the first few centers even more efficiently, since you don't have the inner columns from Step 2 in your way, but you'll want to play around with this yourself to see what options you have. I'll just point out that on the 5x5x5 you need to build the columns on top, but on the 4x4x4 you don't necessarily have that restriction.

You are now going to finish the first 4 layers (F4L), and after a lot of experimentation I've found that it's easiest to do it in two separate steps. So, the first part is to put in the first layer corners, and (if you're solving the 5x5x5) the middle edges too. It might not seem obvious how to do this at first, but the truth is that if you use the outer layers only this is pretty much the same as the F2L step of Fridrich. Find the edge and corner, and then pair up and insert. If you're solving the 4x4x4, you only really need to insert the corner, but in practice I find it valuable to try to insert an additional edge when possible, again using tricks from F2L. As a final note, recognition for this step will be tricky at first, but you'll find that it improves quickly with practice, especially if your D color is clearly distinct from the rest.

This step is basically the same as the F3L step of Thom Barlow's K4. It tends to take a lot of moves, but the lookahead is really good. He's done a really great job of explaining how this works on his website, so check it out for sure. I just want to point out that this can be done without any rotations at all, and in fact should be, unless you see a very nice shortcut that would be uncomfortable to execute without rotating.

In this step, you will be solving the corners of the last layer, and the middle edges too if you're doing the 5x5x5. Like Step 4, this is reminiscent of 3x3x3 methods, and like Step 4 the recognition can be pretty tricky.

For the 5x5x5, I think the fastest way to do this part is OLL/PLL, like in Fridrich. For the 4x4x4 the situation is more complex. If you know full CLL or COLL, you should probably do that. If you can only do it in two steps, though, you can actually optimize enough that I'm not sure it's even slower. The reason is that if you use something like OLL/PLL you have the opportunity to solve a few single edges along with the corners. If you are lucky enough to get an edge pair, you can always solve it. So that approach can often shave a lot of time off the next step.

Only 8 pieces left! This step is really the same in 4x4x4 and 5x5x5, and is the same as the last step of K4 too. An efficient solver will want to average about two pieces solved per algorithm here, and there are two basic ways to do this: using simple commutators to do 3-cycles, and solving entire edge pairs one by one. You should definitely check out the linked K4 page - Thom gives enough 3-cycles to be able to learn the first approach, but he also includes a lot more algs, so that you can learn to do the second approach as well. He recommends the second approach, and I do too, although it's definitely more algorithm-heavy.