Home → Guides → Learning OLL and PLL

Learning OLL and PLL without burning out

Guide · about 9 minutes

There are 57 OLL cases and 21 PLL cases. Most people who set out to learn all 78 stall somewhere around case thirty, and the reason is almost never memory.

The mistake almost everyone makes

The instinct is to treat this as a memorisation problem: 78 sequences, learn them, become fast. So people work through a list, learn ten algorithms in a week, feel productive — and then find their times have not improved, or have got worse.

The reason is that knowing an algorithm and using an algorithm are different skills, and the second one is the bottleneck.

In a real solve you have to look at a scrambled last layer, work out which of 57 cases you are looking at, from any of four angles, in well under a second, and then produce the right sequence without deliberating. Reciting an algorithm on demand is not that. It is barely related to that.

Recognition is the hard part. Execution is the easy part. Almost all learning material is organised around execution, which is why so many people learn a lot of algorithms and get no faster.

Do 2-look first, properly

2-look OLL and 2-look PLL solve the whole last layer with about ten algorithms instead of 78. Do them first, and do not skip ahead.

2-look OLL splits the step: orient the edges (a handful of cases), then orient the corners (seven cases). 2-look PLL splits similarly: permute the corners, then the edges. You take an extra look and an extra algorithm, but you have removed roughly 68 things to memorise.

People treat this as a stepping stone to be passed through quickly. It is not. A solver with fluent 2-look and genuinely good F2L is comfortably sub-20, and often sub-15. Meanwhile a solver with full OLL, full PLL and slow F2L is stuck around 25 and cannot work out why.

PLL before OLL

When you do move to full sets, learn PLL first. Three reasons:

It is a fifth of the work. 21 cases against 57. You will finish it, and finishing a set is worth a lot — it makes the next one feel possible rather than endless.

The payoff is bigger per algorithm. 2-look PLL frequently costs you two full algorithms plus an AUF plus a second recognition. Collapsing that into one saves more time per case than most single OLLs will.

Recognition is more learnable. PLL is recognised from headlights, blocks and bars on the side faces — visual patterns with names people already use. OLL recognition is a subtler business of counting oriented stickers, and it is easier once you have already trained the habit of recognising a case rather than reading a cube.

A workable PLL order

Start with the ones you will meet constantly and that are short: the two adjacent and diagonal corner swaps you may already know from 2-look, then the U perms, then H and Z, then T, then J and R, then the rest. By the time you reach the awkward G perms you will have a solid base, and the G perms are where most people stall — leave them last so stalling there does not mean stalling on everything.

The four G perms deserve a note: they look alike, they are recognised late, and everybody finds them irritating. Do them as a block of four, at the end, together, so you learn to distinguish them from each other rather than one at a time.

Then OLL, in groups that look alike

The standard mistake with OLL is learning it in numerical order from a list. The numbering is arbitrary, and learning case 14 then case 15 teaches you nothing about telling them apart.

Learn by shape group instead: all the dots together, all the crosses, the squares, the lightning bolts, the fish, the P shapes, the W shapes, the C shapes.

Within a group, cases share a look and differ in a detail. Learning them together forces you to find that detail, which is precisely the recognition skill you need. Learn them apart and you will know both algorithms and still hesitate every time one appears.

Start with the groups that already have edges oriented — the ones you meet after 2-look's first step — because those are the cases you can slot into your existing solve immediately, without waiting until the set is complete.

How many at a time

Two or three a week. Not ten.

This sounds slow. At three a week, full PLL takes seven weeks and full OLL takes around twenty. That is roughly six months for both, which is about how long it takes people who succeed. The people who try to do it in six weeks are the ones who quit.

The constraint is not how many you can memorise. It is how many you can drive into reflex before they interfere with each other. Learn ten in a week and you will half-know ten, confuse them under pressure, and in a real solve fall back on 2-look anyway — which means you did the work and got nothing.

The actual loop

1. Learn the finger tricks, not the letters. An algorithm you execute as a memorised move sequence is slow forever. Learn it as a physical motion — which finger does what, where the regrips are. Ten slow, deliberate repetitions beat fifty rushed ones.
2. Drill recognition separately from execution. Look at the case, name it, do not solve it. This is the skill that is actually missing, and it is the one nobody practises.
3. Drill from all four angles. A case you only recognise from the front is a case you will miss three times out of four. This is the single most-skipped step in learning OLL.
4. Use it in real solves the same day. An algorithm that never appears in a solve does not consolidate. Force it: do solves where you deliberately set up the case.
5. Review the old ones. Every session, run through cases you learned weeks ago. Without this, case 12 quietly evaporates while you are learning case 40.

Expect to get slower first

This is the part people are not warned about, and it is why they give up.

When you start using new algorithms in real solves, your times get worse. You hesitate over recognition. You half-remember the new case and fall back to 2-look mid-solve. You execute the new algorithm slowly because it is not in your fingers yet.

This is normal and it lasts weeks. A newly learned algorithm is slower than the 2-look pair it replaced until it is fluent, and fluency takes hundreds of repetitions. The curve goes down before it goes up, for everyone.

Knowing that in advance is most of the battle. People who quit at case thirty usually quit because their average went up and they concluded they were doing it wrong. They were not — they were in the middle of the dip.

Which algorithm to choose

For each case there are several published algorithms. Use these criteria, in order:

• Does it flow? Move count matters less than whether it runs in one continuous motion. A 13-move algorithm with no regrips beats an 11-move one with two.
• Does it fit your hands? If an algorithm is uncomfortable, pick another. There is no prize for the "standard" one, and you will be doing it thousands of times.
• Is it consistent with ones you know? Algorithms sharing a trigger are easier to retain together.
• Where does it leave the cube? For OLL, an algorithm that leaves a recognisable PLL angle saves you a look afterwards.

And then commit. Switching algorithms after learning one costs more than the small gain from a marginally better sequence — you have to unlearn a motor pattern, which is harder than learning one.

Drilling it

The algorithm database lets you browse the full OLL and PLL sets, and the trainer drills a chosen subset: it shows a case, times how long you take, and cycles through the ones you selected.

Use it on the two or three you are learning this week, plus a rotating sample of older ones. That combination — new cases in depth, old cases in review — is what makes a set stick rather than evaporate behind you.

Then take it into a full round. An algorithm you can execute while relaxed and cannot execute in an attempt that counts is not learned yet, and finding that out at home is considerably better than finding it out on the day.