Committing to full CFOP
What seventy-eight algorithms actually buy you, how long they take, and why they are learnt in shape groups rather than in numerical order.
By the end you will be able to
- Say what full OLL and full PLL are worth in seconds, against the two-look versions
- Work out how long the set will take at a rate you can keep to
- Choose whether to learn PLL or OLL first, for a reason rather than by default
- Explain why the cases are learnt in shape groups and not by number
Two-look last layer takes sixteen algorithms and finishes the cube in four steps: orient the edges, orient the corners, place the corners, place the edges. Full CFOP replaces those four with two. One algorithm makes the whole top face yellow. One algorithm finishes the cube.
There are fifty-seven of the first and twenty-one of the second. Seventy-eight algorithms is the largest piece of memorisation in ordinary speedcubing, and it is worth being clear-eyed about what it is for before starting.
The bill
You are not starting from nothing. The seven corner algorithms from two-look OLL are OLL cases 21 to 27, kept unchanged, and every algorithm in two-look PLL is one of the twenty-one. So the real bill is around sixty-four new algorithms — fifty OLL and fourteen PLL.
| Step | Two-look | One-look | Saving |
|---|---|---|---|
| OLL | 10 algorithms, about 15 turns, two looks | 57 algorithms, about 10 turns, one look | roughly a second |
| PLL | 6 algorithms, about 25 turns, two looks | 21 algorithms, about 13 turns, one look | two seconds, sometimes three |
How long it takes
Take a rate you can keep to rather than the rate you manage in the first fortnight. Two new algorithms a week, drilled properly and still there a month later, is a good honest pace. Sixty-four algorithms at that rate is about eight months.
- Four a week finishes in four months, and is achievable if you already drill daily and have good finger habits.
- Ten a week is the rate people announce on the first Monday. Retention falls off a cliff somewhere around the fourth case of the week, and the cases you half-learn are worse than the ones you have not met, because you hesitate over them in solves.
- One a week is not too slow. It is a year, and at the end of it you have full CFOP, which you did not have at the start.
PLL first
If you do one set and not the other, do PLL. The arithmetic is not close. Sixteen of the twenty-one permutation cases come up about once in every eighteen solves; a typical orientation case comes up about once in fifty-four. Each PLL you learn therefore pays out three times as often as each OLL, and there are twenty-one of them rather than fifty-seven.
There is a second reason. PLL is the last thing you do, so a slow one is dead time at the end of a solve where nothing else is competing for your attention. An OLL you fumble at least happens while you still have a permutation case to read afterwards.
Groups, not numbers
The OLL numbering is historical. It sorts nothing you care about. Learning cases 1 to 10 in order gives you four dots, two squares, two lightning bolts and two fish: four unrelated recognition patterns, starting with the eight cases that are both the longest to execute and the rarest to meet.
The library groups the fifty-seven by shape instead — the pattern the yellow stickers make on the top face. Cases in a group look alike, which means you learn their recognition together and, more to the point, you learn the differences between them together. That is where the recognition time actually goes.
What "learnt" means
Worth fixing before you start, because the honest definition is stricter than the comfortable one and it is what stops half-learnt cases piling up.
- You can name the case in about a second, from the two faces you can see, without turning the cube round.
- Your hands run it without your attention, at your normal turning speed, with the same fingers every time.
- It survives the night. Drill it, sleep, and do it cold the next morning before you touch anything else.
- It survives a mixed drill — the whole group shuffled together, in random order.
- Using it in a solve does not cost you your lookahead. If it does, it is not finished.