The subsets, and what they cost
COLL, Winter Variation and ZBLL weighed against the seconds they actually save. Most people should learn COLL and stop.
By the end you will be able to
- Say what each of the main last-layer subsets does and when it applies
- Work out the return on an algorithm set before you commit to learning it
- Decide honestly whether ZBLL is worth the next year of your practice
- Tell whether the last layer is where your time is going at all
Full CFOP finishes the last layer in two algorithms, around twenty-two turns and two recognition pauses. Everything past this point is an attempt to buy away one of those pauses, and every attempt is priced in algorithms.
The sets are worth knowing about even if you learn none of them, because the shape of the trade is the same each time and it is the trade rather than the algorithms that tells you what to do next.
What influencing means
Every subset works by using something you already knew earlier in the solve to narrow the case you meet later. The commonest thing to know is whether the last-layer edges came out oriented — yellow facing up on all four — when F2L finished.
COLL
COLL solves the last-layer corners completely — orientation and position together — while leaving the edges oriented. You reach for it on the one solve in eight where the edges are already oriented, and it replaces OLL and the corner half of PLL with a single algorithm.
It is forty-two algorithms: seven corner-orientation shapes, six ways the corners can be arranged within each. Three more cases exist where the corners are already oriented and only need moving, and you know all three of those already — they are the two A permutations and the E permutation.
Afterwards the corners are done and the edges can only be in one of four states, which the next lesson covers. One time in twelve they are already right and the cube is finished.
Winter Variation
Winter Variation orients the last-layer corners while you insert the final F2L pair. Given oriented edges, that removes OLL from the solve altogether — you finish F2L and go straight to PLL. Twenty-seven algorithms, none of them long.
ZBLL
ZBLL finishes an edge-oriented last layer in one algorithm — orientation and permutation of everything, in one look. It is four hundred and ninety-three algorithms, give or take a handful depending on how you count the symmetric cases.
In CFOP it applies on that same one solve in eight. Its natural home is ZZ, where edge orientation is guaranteed by the first step and ZBLL therefore applies to every solve you do.
The arithmetic nobody does
| Set | Algorithms | When it applies | Roughly what it saves |
|---|---|---|---|
| PLL, in one look | 21 | Every solve | About two seconds a solve. This was the good deal, and you have already taken it. |
| COLL | 42 (35 new) | 1 solve in 8 | A second or so on those solves — call it a tenth of a second on average |
| Winter Variation | 27 | 1 in 8, and only when the last pair cooperates | A second on perhaps one solve in twenty |
| ZBLL | 493 | 1 solve in 8 with CFOP, every solve with ZZ | Around a second, on the solves where it lands |
What most people should do
Learn COLL, and stop. It is the smallest set with a genuine payoff, the recognition it teaches you transfers to everything after it, and the same forty-two cases are the corner step of Roux, so nothing is wasted if you ever wander in that direction.
- Learn COLL if your F2L is smooth, your PLL recognition is instant, and you want a set that makes the last layer feel different rather than marginally quicker.
- Skip all of it if you still pause between F2L pairs. The pause is worth ten times the subset.
- Learn ZBLL if you are already comfortably under twelve seconds, or you solve ZZ, or you find learning algorithms genuinely enjoyable. Those are the three honest reasons and the third is the commonest.