Skip to content
powercubers

How to get faster at the Rubik's cube

There is no trick that takes seconds off overnight. Cubers get faster by moving through a few methods in a common order, and by practising the part of the solve where they lose time. This page goes through those stages, where the time usually goes at each one, how to practise, and how to see whether it is working.

The stages most cubers go through

A beginner's method

Most people learn first to solve the cube one layer at a time, with a handful of algorithms that keep the solved part in place. With practice it can bring you to around a minute.

CFOP with intuitive F2L and a two-look last layer

CFOP solves a cross on one face, then the first two layers (F2L) by pairing each corner with its edge and putting the pair in at once, then the last layer in two steps: OLL turns every piece of it the right way up and PLL puts the pieces in their places.

At first F2L is done by understanding, not by algorithms: you learn how a pair comes together and work out each case. The last layer can be done in two looks for each step, with 16 algorithms in all. Expect to be slower for a while when you switch: pairing feels awkward before it feels quick.

Full OLL and PLL

The full sets solve each step in one look, with 57 algorithms for OLL and 21 for PLL. You can learn them a few at a time, using two looks for the cases you do not know yet. Full OLL, the longer list, is usually the last part of CFOP people learn.

Roux and ZZ

They are the two other methods common among fast solvers. Roux builds a block of three by two by one on each side, solves the corners of the top layer and finishes the last six edges with turns of the top and the middle slice, with little to memorise and almost no rotations. ZZ starts by orienting every edge while it builds the cross, then solves the first two layers and the last layer; having the edges oriented from the start saves moves later. Any of the three works; what matters is that you stay with one long enough to get good at it.

Where a solve loses time

Speed comes from three things: how fast you turn, how well you look ahead, and how good your way through each step is. Turning faster is the one people try first. Pauses, the moments when your hands stop because you are looking for the next piece, are often the bigger loss, and turning faster does nothing about them.

  • With a beginner's method: stopping to remember an algorithm, or doing it with awkward turns. Drill each one until your hands do it without thinking.
  • In the cross: working it out move by move during the solve. A cross never needs more than 8 moves, so with practice you can plan all of it in the 15 seconds of inspection. Planning it is a skill worth practising on its own.
  • In F2L, the longest step: stopping after each pair to search for the next one, and rotating the cube to find pieces. Look-ahead is the fix: while you put one pair in, your eyes are already tracking the pieces of the next.
  • In the last layer: the second it takes to recognise the case, and the algorithms you still have to think about. Drill those until they run on their own, like the beginner ones.

How to practise

  • Slow solves. Turn slowly enough that you never stop. It feels backwards, but it is how you learn to look ahead: once the pauses are gone at a slow pace, you raise the pace.
  • One step at a time. Pick the step that costs you most and drill it: crosses planned in inspection, F2L cases you solve badly, the algorithms you still hesitate on.
  • Untimed work. Solve without the timer and look for shorter ways through F2L. Fewer moves for the same pair is time you save without turning any faster.
  • Timed sessions. Then time yourself with inspection on, as in a competition, to see whether the practice carried over.

How to measure it with this timer

Practice works best when you can see what changed. What the timer has today:

  • Phases. In the timer section of the settings, set phases to the CFOP preset (Cross, F2L, OLL, PLL) or the Roux preset (FB, SB, CMLL, LSE: first block, second block, corners and last six edges), or pick 2 to 6 phases with your own names. Press Space at the end of each step; the last press stops the timer. The time of each step shows under the time, and each solve keeps its splits with the share of the solve each step took.
  • Stats. The stats page reads your history for you, for the last 7, 30 or 90 days or all of it: whether your ao12 is going down and how fast (from 50 solves), whether your recent solves are more or less consistent than before, whether your first solves of the day are slower (warm-up), whether you slow down in long sessions (fatigue), and at what time of day you are fastest. Each reading appears once you have enough solves for it.
  • Tags and labels. Mark solves with a tag, such as a cube or a condition like tired, and filter by it; with an account you also get labels. The stats page compares the solves of each tag or label with the rest, so you can see whether a cube or a habit makes a difference.
  • Daily challenges. With an account, every event has a daily challenge, open once you have 15 solves in the event: one scramble for everyone, one attempt with WCA inspection and a leaderboard. It is practice at doing well on the one solve that counts.
  • Averages. Watch ao5 and ao12 rather than single solves. The guide to ao5 and ao12 explains why.

New to reading algorithms and scrambles? Start with cube notation. And when you want to try your times against others in person, see your first WCA competition.

Questions

Do I need to learn full OLL and PLL?

Not to start. Two-look OLL and two-look PLL need 16 algorithms between them and solve every last layer. The full sets, 57 and 21 algorithms, save a look each and are usually learned a few at a time.

Is turning faster the best way to get faster?

It is one of three parts, with look-ahead and how you solve each step. Pauses to look for the next piece add up quickly, and turning faster does not remove them. Learning to see the next step while you do the current one does.

Should I time every solve?

No. Slow, untimed solves are where you practise looking ahead and finding shorter ways through F2L. Timed solves tell you whether it worked.

Should I switch from CFOP to Roux or ZZ?

You do not need to. CFOP, Roux and ZZ are all used by fast solvers. Switching means relearning the start of every solve, so it is worth it when you like how the other method works, not as a fix for a plateau.

How long does it take to get sub 20?

It varies too much from person to person to give a number. What it usually comes down to is a speed method such as CFOP, an F2L without long pauses, and the last layer algorithms you use done from memory.