Blindfold method: Clarifications

Andrej Karpathy. Okay, in this video I will give some clarifications on the blindfold solving video that I made a while ago, and I'll answer some common questions about it. The video was quite successful and I had many positive comments about it, but still I think some things could have been explained a little better. So first of all, I'm somewhat surprised not too many people got confused when I talked about the parity issue. I thought that was poorly handled, so Here's the problem. You're given a scrambled cube, then you memorize it. You memorize all the edges and you memorize all the corners and how to do them. Now the problem becomes that if you do odd number of swaps to solve the edges, then recall we're always swapping these two corners as a side effect of solving the edges. Well, what happens is if you do this odd number of times, the corners are actually swapped from what you remember them as. So To fix that, you have to swap them one more time to get it to what you memorized them as. And we do that by swapping these two corners and swapping these two edges as well. Now you may ask, why should I swap these two edges? I just solved all the edges, like, why should I unsolve them? Well, if you recall, we always— we're going to do corners next, and we always shoot something from here to its correct place, and we always swap these two edges back and forth. By the end, you're guaranteed that these two edges will come to their correct positions, and that's just because if you do everything correctly, it's impossible to run into a case where you only have two edges unsolved. This is an impossible state. I had to disassemble the cube to get it to this state. You are guaranteed that everything will just work out with those two edges if you do everything else correctly. So in particular, what you have to note is that once you finish the edges, you have to do the parity fix before you move on to the corners because the corners are in their wrong positions. So the parity is done between edges and corners, it's not done at the end. Next, when I did the example solve, I had the first layer already finished. Now this confused some people because it seemed like I was doing the Fridrich method blind or something. That was somewhat of a mistake on my part. So the fact that the first layer happened to be solved was a complete coincidence. You must realize that this method has absolutely no connection whatsoever to the Fridrich method or any of the beginner's methods. There is simply no concept of first, second, or last layer. Every single corner or edge has a unique position on the cube where it needs to go. We simply solve one piece at a time by shooting them to their correct places from the buffer position, one by one, until we are done. The order in which we solve these pieces is completely determined by the random order in which they come into the buffer positions. We just solve whatever is in the buffer at any point. Now, I just wanted to give an example solve where many of the edges and corners were already solved so that it was quick. I don't exactly remember why I made the first layer solved in particular. It could have been anything that had many of the random edges or corners solved. For example, this could have been a better example because it's very short to do. Many of the edges and corners are already solved, right? So this would be very easy to solve as well using the blindfold method because we don't have to solve many pieces. That would have been a better example. Next. One person asked if there was a difference between green-orange and orange-green, and of course, yes, there is a big difference. So here in the buffer position, I have the orange-green edge. You always call the one on the top first. So this is the orange-green edge, this is the green-orange edge, and you solve them differently, of course, because remember that the T permutation swaps these two edges, but you also have to consider what stickers of those edges are solved. So this is the green-orange edge, and it has to come here. But in particular, this green sticker has to replace this orange sticker so that it's on the green side here. So when we move this edge to this position so that they're ready to be swapped, we must make sure that the orange sticker of this edge is on the top because we want the green to replace it. Remember that the T-permutation preserves orientation on top, so this sticker of this edge is swapped with this sticker of this edge, and these two stickers are swapped. So to solve this one, it's not easy. I mean, it's harder than solving this one. The setup move for this, when we have orange-green, orange has to replace the red part here, so that's the setup move. The T-permutations will swap them, and it will make sure that orange gets replaced with red. This is different because we have the green-orange, so if we did this like we did previously, the orange would— the edge would go in the wrong way. So you actually have to use a different setup move for this, and the setup move is, of course, you want to make sure that orange is on top so it's actually— oh my god, this is hard to show. It's that, that. So it's, uh, what is this, D' L, I think. So that edge moves there, but the orange is on top, that's the important part. You swap and then you undo setup move to solve it. So yes, there's a difference. One person also had a pretty good question. He asked, if an edge is in the correct position but it's incorrectly flipped, like this one here, does that count as a cycle? And the answer is yes, it does count as a cycle. And you have to be always on the lookout for cases like this. So actually, whenever I'm about to blindfold solve a cube, the first thing I check for are exactly these cases: when you have an edge in the correct place but incorrectly flipped, or when you have a corner in the correct place and it's incorrectly twisted. The reason for this is that it's very easy to get confused and you think you're done, but you're actually one off because it's hard to spot this. I mean, you already think that you've already memorized so much, you're already probably done by now, but actually you're not, you're one off, and it's usually pretty hard to tell. So you do have to break into it and you do have to solve it, and it's going to take you 2 shots to do it. So first you break into that cycle, so I'm doing the J permutation to get it into my buffer position, and then I shoot it to the correct position in the correct way. So you have to do 2 algorithms to solve these cases. Another issue that came up a couple of times was that people are confused about how this method relates to the M2R2 method. Both methods were created by Stephan Pochmann. He came up with this method first and then the M2R2 method later. This method is also sometimes referred to as the old Pochmann method, with the M2R2 being, I guess, the new method. Anyway, so here's the difference. M2R2 is basically a blindfold solving method that is very similar to this one, but it's just much faster. The downside is that it requires you to learn algorithms. There are roughly 10 of them. It is also a little more complicated, so if you don't feel comfortable using this method, you should probably not bother trying to get into M2R2. I will link you to the M2R2 method in the video description if you're interested in it. The main difference is it's the same kind of idea, but you don't use— you don't shoot edges from here to here using the T permutation, but you shoot edges from here here using M2, and you shoot, I think, corners from here to here, I think, using R2 or something like that. It's very strange. So basically, yeah, it's faster, but you have to learn stuff. Last question that I received a couple of times was how I keep track of what positions are solved and what positions aren't solved. Like, sometimes, you know, you have to break into a new cycle because you're not done, but you are just not sure where that cycle is. And you have to kind of go through your story and figure out what positions you have not shot to yet. So actually I had the exact same problem when I was starting out, but it got progressively better the more I practiced. But sort of what I kind of tried to do was, for example, I solve the orange-blue, it will come here, and I'll put my thumb on it to indicate that I already shot to it. And then the red-yellow— I mean, red-green, so I'll sort of put the thumb on that as well. And then we have the green-yellow, so I'll do that, and I sort of tried to do it like that. It sort of helped, but yeah, I don't really have a good answer for that. It just kind of comes with practice. Okay, so I hope that I answered some questions there, and I hope that the video was clear. I certainly wouldn't like to post a clarification video to a clarification video. And that's it. Bye-bye.