mark49152
Premium Member
EDIT: Having come back and looked at this three years later, it is clearly over-complicated, probably because I spent so long immersed in the details when I originally wrote it! 
So, TL;DR version: Here is a list of all the OLL shapes, ordered by the average number of moves they save compared to solving the same cases with 2-look OLL. Learning full OLL in this order is a good idea because the ones nearer the top will cut more time off your solves, on average.
In fact, learning just the top 15 OLL algs delivers 42% of the total move savings of learning full OLL, so if you only want to learn some of it, those are the ones to go for. Here's a condensed guide: View attachment ollTopThird.pdf
------ Original Post ------
Two of the questions that come up regularly on the forums are "when should I learn full OLL?" and "what order should I learn the OLLs in?".
Well, obviously you should learn them when you think they will benefit your solving more than anything else you could spend your practice time on. Then learn them in an order that starts with those that will benefit your solving most. The question of course is when and by how much they will benefit you.
The attached guide should help you decide this for yourself. It lists all 49 OLL cases that could be learned on top of EOLL & OCLL, ranked in order of move savings. That means it compares the number of moves taken to solve each case using 2-look OLL with the number of moves taken to solve the same case with 1-look OLL, and ranks the OLLs according to how much shorter the 1-look algorithm is. The logic is that you should learn the OLLs in order of potential move savings.
Here it is: View attachment oll1look.pdf
This guide assumes that you know 2-look OLL already, meaning the usual EO/OCLL approach (there are other ways of doing 2-look but we'll ignore that). If you don't know 2-look OLL, meaning all 3 EOLL algs and all 7 OCLL algs, learn that first. Also, full PLL is more beneficial than full OLL, so if you don't know full PLL, learn that first.
Guide to the table
The table is split vertically into three sections.
The blue section shows the OLL cases. Wiki# is the reference number of the case on the Speedsolving Wiki. EO and U patt give the edge pattern and the OLL U-face pattern, and the fourth column shows the probability of the case occurring. This is important, because clearly it's less useful to learn a case with probability 1/216 than one with probability 1/54.
The orange section shows the OCLL cases that occur when using 2-look OLL on each case. There are four positions relative to the picture shown in the wiki - unrotated, U, U2 and U'. The four columns show which OCLL you will get depending on which position you are in when you start EOLL, and also which AUF move is required between EOLL and OCLL. Note that for OLLs with "L" shape edges, the starting positions are relative to the position with the L at back left, rather than the wiki picture, to save having to look those cases up, and different algs (see OLL-44) are used for each L position.
The last two columns in the orange section show the worst-case and average move count for EOLL/AUF/OCLL. Worst case means the longest sequence from all possible starting positions, per case. Move counts are in QTM but with double moves (R2, U2 etc.) counted as 1.5 and cube rotations counted as 2 moves. This gives better proportionality between move count and execution time (at least for me!) although it is still only an approximation and actual execution time obviously depends also on how easy or awkward the flow of the alg is.
The green section shows the full OLL algs and their move counts, in the same metric. The final column shows the number of moves saved, relative to the average 2-look move count. The table is ranked by order of move savings; OLLs with the same move saving are ranked by length of alg on the basis that longer solutions are more in need of shortening. OLLs that have a probability less than 1/54 are weighted accordingly, i.e. a case with probability 1/108 has its score halved before ranking.
Underneath is a second table that shows scores for OLL groups. These are groups of cases with the same U-face pattern, as listed on the SS wiki, and the score given is the average of all cases in that group. Some of the groups have been modified slightly.
Recommendations
Based on this analysis, some OLLs have the potential to save as many as 8-10 moves and are certainly worth learning earlier in your progression. That many moves will make a difference well before you get sub-20. On the other hand, some OLLs save so few moves that they would offer only minimal benefit compared to fast and well-practiced 2-look algs, even if you are sub-20.
My plan is as follows. Obviously you can make your own mind up what will work for you.
First, I will learn a selection of OLLs that save 6+ moves, by selecting U-pattern groups. Learning in groups reduces recognition time because you can make a fast decision based on U face only as to whether you will do 1-look or 2-look. Learning only part of a group slows down the 2-look cases because you pause to eliminate the 1-look possibilities.
I selected 15 cases in 5 groups and I call these the "top third" OLLs. These account for 42% of the total moves saved by learning full OLL, with an average score of 8 moves per case, compared to only 3 moves per case for the worst 15 OLLs. Some of them are really easy, like learning which cases are solved directly by the EOLL algs OLL-2 and OLL-45 and making sure to execute them from the right starting position. Most of the dot OLLs are included, which makes sense as EOLL and the 2-look solution is longer.
Here is a cheat-sheet for these 15 cases/algs: View attachment ollTopThird.pdf
Next, mid-table OLLs. In general, I would not bother to learn these except for algs that are really easy and fast, or where I don’t like the OCLL case. A good example is OLL-54, which is an easy double wide Sune, and where 2-look always results in a Pi OCLL which I’m not keen on. I won't learn any case unless I believe I can recognise it fast enough not to slow down 2-look for other cases in the same group.
Bottom of the league OLLs that only save 1-2 moves I probably wouldn’t even consider provided that I know what OCLL is coming and can do the 2-look without pause. The prize for the most useless OLL goes to OLL-41, for which the "optimum" 1-look alg scores worse than 2-look (due to the rotations).
Finally, there’s another option for OLLs that have a relatively small move saving, say <6. That is to force a preferred OCLL case by choosing the right starting position. For example, OLL-14. The 1-look OLL alg has a rotation, and I don’t like Pi OCLL either, but I know that if I start 2-look with the L at back left then I will get a Bowtie with no AUF. What’s even better is that I can run the F' at the end of the EOLL immediately into the F' at the start of the Bowtie.
Caveats and disclaimers
Move count is a useful metric for estimating alg execution time, but it's nowhere near 100% reliable. It's just an approximation. A 12-move alg that flows well can be faster than a 9-move alg that doesn't, so it's quite possible that an OLL that the table says will save you 10 moves will make less of a difference to your times than another alg that only saves you 7 moves. This table is intended as just a helpful resource, not a definitive account of the relative speed of OLLs.
Also there is more than one possible alg that can solve each OLL. The rankings in the table are based on the algs shown, but there might be other algs out there that are shorter or that are longer but faster. It's up to you to find those. If this guide says learning a case has little benefit, that doesn't mean there isn't some other, better alg out there that makes learning it worthwhile. The algs here are mostly from Badmephisto, with some from the Speedsolving wiki, and are the ones that work best for me personally.
If you read this far, I hope it helped! Happy OLLing!
So, TL;DR version: Here is a list of all the OLL shapes, ordered by the average number of moves they save compared to solving the same cases with 2-look OLL. Learning full OLL in this order is a good idea because the ones nearer the top will cut more time off your solves, on average.
Code:
10.00 Stealth
8.50 Squares
8.13 Dots (1/54)
7.50 T shapes
7.31 Small lightning bolts
6.67 P shapes
6.00 Kites
5.88 C shapes
5.58 L shapes
5.56 Knight moves
5.38 Fishes (excluding kites)
5.25 Big lightning bolts
4.25 H
4.13 I shapes
3.81 Dots (rare)
3.44 Awkward
2.63 W shapesIn fact, learning just the top 15 OLL algs delivers 42% of the total move savings of learning full OLL, so if you only want to learn some of it, those are the ones to go for. Here's a condensed guide: View attachment ollTopThird.pdf
------ Original Post ------
Two of the questions that come up regularly on the forums are "when should I learn full OLL?" and "what order should I learn the OLLs in?".
Well, obviously you should learn them when you think they will benefit your solving more than anything else you could spend your practice time on. Then learn them in an order that starts with those that will benefit your solving most. The question of course is when and by how much they will benefit you.
The attached guide should help you decide this for yourself. It lists all 49 OLL cases that could be learned on top of EOLL & OCLL, ranked in order of move savings. That means it compares the number of moves taken to solve each case using 2-look OLL with the number of moves taken to solve the same case with 1-look OLL, and ranks the OLLs according to how much shorter the 1-look algorithm is. The logic is that you should learn the OLLs in order of potential move savings.
Here it is: View attachment oll1look.pdf
This guide assumes that you know 2-look OLL already, meaning the usual EO/OCLL approach (there are other ways of doing 2-look but we'll ignore that). If you don't know 2-look OLL, meaning all 3 EOLL algs and all 7 OCLL algs, learn that first. Also, full PLL is more beneficial than full OLL, so if you don't know full PLL, learn that first.
Guide to the table
The table is split vertically into three sections.
The blue section shows the OLL cases. Wiki# is the reference number of the case on the Speedsolving Wiki. EO and U patt give the edge pattern and the OLL U-face pattern, and the fourth column shows the probability of the case occurring. This is important, because clearly it's less useful to learn a case with probability 1/216 than one with probability 1/54.
The orange section shows the OCLL cases that occur when using 2-look OLL on each case. There are four positions relative to the picture shown in the wiki - unrotated, U, U2 and U'. The four columns show which OCLL you will get depending on which position you are in when you start EOLL, and also which AUF move is required between EOLL and OCLL. Note that for OLLs with "L" shape edges, the starting positions are relative to the position with the L at back left, rather than the wiki picture, to save having to look those cases up, and different algs (see OLL-44) are used for each L position.
The last two columns in the orange section show the worst-case and average move count for EOLL/AUF/OCLL. Worst case means the longest sequence from all possible starting positions, per case. Move counts are in QTM but with double moves (R2, U2 etc.) counted as 1.5 and cube rotations counted as 2 moves. This gives better proportionality between move count and execution time (at least for me!
) although it is still only an approximation and actual execution time obviously depends also on how easy or awkward the flow of the alg is.The green section shows the full OLL algs and their move counts, in the same metric. The final column shows the number of moves saved, relative to the average 2-look move count. The table is ranked by order of move savings; OLLs with the same move saving are ranked by length of alg on the basis that longer solutions are more in need of shortening. OLLs that have a probability less than 1/54 are weighted accordingly, i.e. a case with probability 1/108 has its score halved before ranking.
Underneath is a second table that shows scores for OLL groups. These are groups of cases with the same U-face pattern, as listed on the SS wiki, and the score given is the average of all cases in that group. Some of the groups have been modified slightly.
Recommendations
Based on this analysis, some OLLs have the potential to save as many as 8-10 moves and are certainly worth learning earlier in your progression. That many moves will make a difference well before you get sub-20. On the other hand, some OLLs save so few moves that they would offer only minimal benefit compared to fast and well-practiced 2-look algs, even if you are sub-20.
My plan is as follows. Obviously you can make your own mind up what will work for you.
First, I will learn a selection of OLLs that save 6+ moves, by selecting U-pattern groups. Learning in groups reduces recognition time because you can make a fast decision based on U face only as to whether you will do 1-look or 2-look. Learning only part of a group slows down the 2-look cases because you pause to eliminate the 1-look possibilities.
I selected 15 cases in 5 groups and I call these the "top third" OLLs. These account for 42% of the total moves saved by learning full OLL, with an average score of 8 moves per case, compared to only 3 moves per case for the worst 15 OLLs. Some of them are really easy, like learning which cases are solved directly by the EOLL algs OLL-2 and OLL-45 and making sure to execute them from the right starting position. Most of the dot OLLs are included, which makes sense as EOLL and the 2-look solution is longer.
Here is a cheat-sheet for these 15 cases/algs: View attachment ollTopThird.pdf
Next, mid-table OLLs. In general, I would not bother to learn these except for algs that are really easy and fast, or where I don’t like the OCLL case. A good example is OLL-54, which is an easy double wide Sune, and where 2-look always results in a Pi OCLL which I’m not keen on. I won't learn any case unless I believe I can recognise it fast enough not to slow down 2-look for other cases in the same group.
Bottom of the league OLLs that only save 1-2 moves I probably wouldn’t even consider provided that I know what OCLL is coming and can do the 2-look without pause. The prize for the most useless OLL goes to OLL-41, for which the "optimum" 1-look alg scores worse than 2-look (due to the rotations).
Finally, there’s another option for OLLs that have a relatively small move saving, say <6. That is to force a preferred OCLL case by choosing the right starting position. For example, OLL-14. The 1-look OLL alg has a rotation, and I don’t like Pi OCLL either, but I know that if I start 2-look with the L at back left then I will get a Bowtie with no AUF. What’s even better is that I can run the F' at the end of the EOLL immediately into the F' at the start of the Bowtie.
Caveats and disclaimers
Move count is a useful metric for estimating alg execution time, but it's nowhere near 100% reliable. It's just an approximation. A 12-move alg that flows well can be faster than a 9-move alg that doesn't, so it's quite possible that an OLL that the table says will save you 10 moves will make less of a difference to your times than another alg that only saves you 7 moves. This table is intended as just a helpful resource, not a definitive account of the relative speed of OLLs.
Also there is more than one possible alg that can solve each OLL. The rankings in the table are based on the algs shown, but there might be other algs out there that are shorter or that are longer but faster. It's up to you to find those. If this guide says learning a case has little benefit, that doesn't mean there isn't some other, better alg out there that makes learning it worthwhile. The algs here are mostly from Badmephisto, with some from the Speedsolving wiki, and are the ones that work best for me personally.
If you read this far, I hope it helped! Happy OLLing!
Last edited:
THANKS A BUNCH 
