Andrej Karpathy shares 57 OLL algorithms for Rubik's Cube, grouped by trigger patterns and color-coded on his webpage. He recommends learning the shortest cases first and offers two advanced tips: partial edge control during F2L to orient top edges, and using only R and U moves on the final pair to land easier OLL cases.
Chapters
Hi guys. Ever since I mentioned that Cubiz OLLs are not exactly great, you have all been bugging me to tell you which OLLs I use instead. I was not able to point any of you to some page because the truth is that my current OLLs are a blend from about 5 websites. So over the past few days I've been putting together a webpage where I list all the OLLs that I currently use, and I'll show it to you in this video.
So anyway, here is the webpage. Okay, so here we go. On the very top here you have the printable page link and the notation link. So do not ask me what xy or small r is, check the notation page. And this is the listing of all the OLLs. Now the listing is pretty long because there is a lot of OLLs, there's 57 OLLs. So you probably know many of these already though. So the way I group them is by triggers. So the R U R' U' trigger comes up in many, many of the OLLs, so I grouped all those OLLs into the same group, and you can see that it's color-coded, so it comes up all the time. Second, you have the R' F R F' trigger, also comes up all the time. The R U R' U trigger, all the algorithms that abuse it. Then you have the sune and its derivatives. Sune is basically this algorithm here, it solves this case, and these are all the algorithms that are like it. It's a very short group, I'm not sure why I made it a group by itself, but whatever. And here's all the miscellaneous algorithms that do not fit with what I've been doing above. There's 18 more of those, so these are the algorithms that just do not have some very common trigger in them, but they're very, very quick. All of these algorithms are basically the quickest way I can make do each case. So now, the order that you should learn these in, I would suggest that you just go down the list here, basically, and just learn all of those like that. And these miscellaneous algorithms, just learn them in between, you know, just whenever you get frustrated with this case coming up all the time, just learn it. It's pretty short, pretty quick. You should learn all the algorithms that just look very easy, learn all of those first because as you can see here, this is the chance of occurrence. Most of these algorithms have the exact same chance of occurrence, 1 out of 54 or about 2%. So all of the algorithms come up about the exact same time, so why learn long algorithms when you can learn the short ones first and just know most of them all soon? I also assigned some names here. I made up many of those names. Had some fun doing that. So you've got names like the Vbar, Blok, Soon, Andysoon, Chameleon, Checker, Tractor. I don't know, I just made these things up. And of course a comment. And that's pretty much it. In the top here you can also suggest an algorithm to me.
If I like it, I may add it here.
You can see that in some cases I list two algorithms, and that's basically because I believe it's a matter of personal preference for many of these cases. I always say which algorithm I use, but I list some variations which you may want to use.
There's one more tip that I would like to share with you. As you can see from the OLLs, all of the cases that have none of the edges oriented are very hard to solve, they take a lot of moves to solve, so usually you want to avoid these cases. There's a method that kind of does that in some cases, and it's called the partial edge control. So usually what happens when you finish up F2L, you have all the pairs inserted, and the last pair, you're about to insert it, and you insert it like that. Okay, so the problem with that insertion like that is that it orients none of the edges on top. So just above, just before you insert this, you can see that none of the edges are oriented. If you just insert like this, they will all stay unoriented. There is one more insertion that actually orients two edges, and it's the R' F R F' insertion. It orients two edges already for you. So when you see that, when you're just about to finish your F2L and you see that the edges are unoriented, don't insert your pair like that. Because you'll run into a bad case. Try to insert it with the R' F R' insertion, and then you'll get a much easier case. For example, in this case, it's just 6 moves. So that's partial edge control. Similarly, you can notice that all the cases where all the edges are oriented are pretty easy to do.
They're fast algorithms.
So sometimes when you're finishing your F2L, you can notice on the, on the top you can see that your edges are oriented like I have here. So usually when you're about to insert your last pair and you see that those edges are oriented, you want to keep that orientation so that when you get to the OLL part, you get one of the easy cases where all the edges are oriented. So if you were to normally do this case, which many of you probably do it this way, so that you set up for this kind of insertion here, So that unoriented 2 edges, actually. So now you only have 2 edges. The trick that you need to know is that if you only turn 2 layers, for example R and U as it usually is, you can never unorient an edge. In this case, I used F layer, U layer, and R layer, and I ended up unorienting 2 edges. But another way to solve this only using R and U would be This. As you can see, it was a little bit more moves, but it's very quick because it's R and U only. I only used 2 layers, so none of the edges got misoriented, and now I have a very easy case, right? So when you see that your edges are unoriented— are oriented on the top, then try to finish the F2L only using 2 layers. 2 layers and you will get an easy case at the end.
