CFOP: the road to sub-20
For someone who knows CFOP and solves in about 25 s: what to practise and how.
Solve breakdown
| Stage | At 25 s | Sub-20 |
|---|---|---|
| Cross | 3–4 s | 2–3 s |
| F2L | 14–16 s | 10–11 s |
| OLL | 3–4 s | 2–3 s |
| PLL | 3–4 s | 2–3 s |
Approximate values. Everyone loses time somewhere else.
Where you lose timeAt 25 s the most can be gained in F2L, then in the cross and the last layer.
A sub-20 solve is roughly: cross 2–3 s, F2L 10–11 s, OLL and PLL 2–3 s each. At 25 s it is usually F2L that lags: pauses between pairs, searching for pieces and cube rotations eat up 4–5 s.
Before you practise everything at once, check where your loss is. Time a few cross-only solves in the timer’s “Cross” mode and compare them with whole solves.
- F2L without pauses gives the most, not faster moves.
- Full PLL is usually 1–2 s less than 2-look.
- A cross over 8 moves or longer than 3 s is a sign you aren’t planning it as a whole.
Cross during inspectionPlan the whole cross during the 15 s inspection and solve it in 8 moves or fewer.
Every cross can be solved in at most 8 moves, and about 6 is enough on average. If your cross takes 10–12 moves, you lose a second or more on it.
Plan the edges relative to each other, not to the centres right away: two edges that are next to each other in the right order you insert together. At the end, one bottom-layer move aligns the whole cross with the centres.
Execute the cross from memory, without looking at it. Your eyes are already looking for the first pair.
- First find all four white edges, then the order.
- Avoid cube rotations during the cross; solve the cross on the bottom.
- After each solve in the “Cross” mode compare your moves with the optimal ones.
The cross of the example scramble after a z2 rotation: white on the bottom.
Cross+1 and cross+nXCross is a cross solved together with the first pair. A two-step plan in one inspection.
Cross+1 (XCross) is a cross solved together with one F2L pair. Instead of a cross in 6 moves and a pair in 7, you often do the whole thing in 8–9, with no pause after the cross.
Don’t force it. It pays off when the pair’s corner and edge are close to each other or already connected, when the pair’s slot is empty during the cross, or when one cross move connects the pair along the way.
How to learn: during inspection first plan the cross, then see where the first pair will be after it. When that works regularly, look for a cross that doesn’t break that pair. Cross+n (cross+2 and beyond) means tracking further pairs in your head, a level below 15 s.
Pseudo pairs are pairs inserted into “someone else’s” slot with the cross shifted by one bottom-layer move. At 25 s it’s enough to know they exist.
- To start: the cross plus awareness of where the first pair is (so-called cross+1 in your head).
- An empty slot during the cross is a chance for the pair’s corner to “drop in” along the way.
- Cross+1 mode shows the shortest XCross in every slot: check which slot was the easiest.
The same scramble: the cross and the green-orange pair at once.
F2L without pausesFewer cube rotations, inserting into back slots and picking the easiest pair.
At 25 s F2L is usually 14–16 s. Most of it goes on pauses to search for pieces and on cube rotations (y), after which you have to look around again.
Insert back slots without a rotation, with R′ and L moves. Before inserting a pair, use a U move to set it where the insertion is shortest.
Always take the easiest pair, not “the next one in order”. You insert short cases (3–4 moves) without thinking, and meanwhile you look for the next one.
- Insertion into the back right slot: R′ U′ R instead of a y rotation and R U R′.
- Learn first the cases where the pair is already connected in the top layer.
- The full 41 F2L cases aren’t needed for sub-20; intuitive F2L without pauses is enough.
A pair connected above the slot.
The same situation at the back, without y.
LookaheadSolve more slowly, but without stopping. Look for the next pair while inserting.
Lookahead means looking further ahead than what you’re doing right now. While you insert a pair, your eyes are already looking for the next one, so there’s no pause after inserting.
The best exercise: a few very slow solves a day, but without a single pause. You can turn on a metronome (e.g. 1–2 moves per second) and do one move per beat. When it works without stops, speed up a little.
Fast moves with pauses give a worse time than an even, calm pace. Hand speed comes later by itself.
- While inserting a pair, don’t look at it: you know the moves, look elsewhere.
- Fewer regrips mean fewer moments when you lose sight of the cube.
- Practise lookahead on short, familiar F2L cases.
PredictionYou know what the algorithm does to the pieces, so you see the next step before the algorithm ends.
Prediction means foreseeing the state after an algorithm. When you know an algorithm well enough to know where each piece will go, you can recognize the next case while it’s still being executed.
Start with simple things: at the last F2L pair watch how the top layer edges will end up (OLL). After PLL predict the AUF by tracking one sticker instead of looking at the cube when you finish.
Exercise: turn on the algorithm animation, pick one piece and tell yourself where it will end up. Then check. First the sexy move and Sune, then the algorithms you use most.
Advanced: predicting PLL already after OLL (e.g. after a Sune with visible headlights). That’s a level below 15 s; at 25 s the AUF and OLL edges are enough.
Track the front right corner: where will it end up and how twisted?
Which two corners and which two edges swap places?
Last layerFull PLL first, then OLL. Practise recognition with flashcards, without a cube.
From 2-look move to full PLL first: 21 algorithms, and the gain is 1–2 s in every solve, because you do one algorithm instead of two. Learn full OLL (57) afterwards, starting with the cases with a cross on top (OCLL), which you already know from 2-look.
Recognition from two sides: headlights, blocks and bars on the sides tell you which PLL it is, without rotating the cube. The flashcards in the app practise exactly that, with a random AUF.
Introduce new algorithms 2–3 a day and repeat the old ones. Training in the app picks the ones you have trouble with more often.
Color neutralityA cross from white or yellow gives shorter crosses, but at 25 s it’s not a priority.
Color neutral means solving the cross from any colour, dual (white/yellow) from white or yellow. With two colours to choose from the cross is shorter on average, and good XCrosses are easier to find.
At 25 s the gain is small compared with F2L without pauses. If you want to start, solve a whole cross from yellow every now and then so that recognizing pairs isn’t unfamiliar.
4-week planAbout 20 minutes a day: cross, F2L without pauses, new PLLs and one ao12 at the end of the week.
Week 1: 10 crosses a day in the “Cross” mode (goal: ≤8 moves, compare with the optimal one), 2 new PLLs a day and PLL flashcards.
Week 2: 5 slow solves without pauses a day (lookahead), more new PLLs, flashcards from the daily challenge.
Week 3: 5 solves a day in the “Cross+1” mode, F2L without cube rotations, a review of all PLL in Training.
Week 4: your first OLLs (start with OCLL), AUF prediction after PLL, and an ao12 in the timer at the end of every week to see your progress.