三个棱块朝同一方向循环错位。一个棱块已还原。
PLL —— 21 个算法
完整的魔方 PLL 算法参考。学习适合初学者的两步 PLL,或掌握全部 21 个算法实现完整一步 PLL。
21 种情况——完整参考
The home stretch
OLL already turned the whole top face one solid color. PLL has exactly one job left: slide every last-layer piece into its correct position without touching its orientation. When the last piece clicks home, the cube is solved — provided your F2L stayed intact.
PLL algorithms look long, and that is on purpose. By this stage every piece is already oriented, so an algorithm has the least freedom to move things around: it must cycle the pieces you want while returning every other piece untouched. That politeness is what makes the sequences long — not difficulty.
21 beats 57. Deciding which full step to learn first? PLL is the smaller mountain — only 21 cases against OLL’s 57. Front-load it.
Climb the ladder: 4 looks → 3 looks → 2 looks
You do not need all 57 OLL cases before you start using full PLL. Harris lays out an explicit path from the beginner method to one-look CFOP:
- 4-look (beginner)
Edge cross, then orient corners, then position corners, then position edges — four small algorithm families. This is the beginner last layer.
- 3-look (the bridge)
Learn just the 7 OLL cases where every edge is already flipped to orient the corners in one look, then do full PLL in one look. Cross → 7-case OLL → PLL. This is the rung most worth climbing next.
- 2-look (CFOP)
Full 57-case OLL in one look, then full 21-case PLL in one look. The expert last layer.
Recognition: line up first, then look
Harris’s recognition advice is an action, not a static glance: rotate U to bring as many already-solved pieces home as you can, then read what is left over. Pre-aligning makes the remaining mismatch jump out — and it is the only way to tell mirror cases (Ua vs Ub, Ja vs Jb, the four G-perms) apart reliably.
Headlights → adjacent swap
Two matching corners with a different center color between them (“headlights”) are a solved pair. The case is an adjacent-swap family: T, J, or R.
No headlights → diagonal swap
No headlights anywhere means a diagonal swap — Y, V, or N — or a G/E-perm. Diagonal swaps never show headlights.
See the two you’ll use most
Watch the two most common PLLs run on a real cube. The T-perm is the single most-used case and the backbone of 2-look PLL; the Ua-perm is the classic edge 3-cycle. The top face is already one color — PLL only slides the pieces into place.
T-perm — the adjacent-swap workhorse
Ua-perm — the edge 3-cycle
Learn it by family, not by flashcard
Harris does not present 21 loose pictures — he buckets them into seven bite-size families that share a shape and a feel. Learn one member, internalize the trigger, then derive its mirror. The families are the study plan.
Edge perms (Ua, Ub, H, Z)
All four corners home; only edges cycle. The fastest family — done as M-slice flicks.
M2UM2U2M2UM2Corner perms (Aa, Ab)
All four edges home; a single diagonal corner 3-cycle. Built on the R' U R' D2 trigger.
R-perms (Ra, Rb)
One adjacent corner swap plus an edge 3-cycle. A solved block on one side; the mirror is the other.
J-perms (Ja, Jb)
An adjacent corner pair and adjacent edge pair swap on one face — the most beginner-friendly combined perm.
RUR'F'RUR'U'R'FR2U'R'U'N-perms (Na, Nb)
Two diagonal swaps crossing — the hardest pair. Long by design: two whole columns trade.
G-perms (Ga–Gd)
A 3-corner cycle paired with a 3-edge cycle. Four mirror-confusable cousins, all driven by a wide-U turn plus a re-grip rotation.
The rest (T, F, V, Y)
The high-value standalones — including the T-perm, the most-used PLL of all.
RUR'U'R'FR2U'R'U'RUR'F'All 21 verified cases. Filter by family. Tap a card to copy the algorithm.
三个棱块朝与 Ua 相反的方向循环错位。一个棱块已还原。
四个棱块全部错位。相对的棱块互换。魔方从四面看都呈对称图案。
四个棱块全部错位。相邻棱块以 Z 图案互换。
三个角块朝一个方向循环,而所有棱块保持还原。
三个角块朝与 Aa 相反的方向循环,而所有棱块保持还原。
四个角块以对角配对全部错位。所有棱块已还原。
两个相邻角块需要互换。前-右与前-左角块对调,加上两个棱块。
互为对角的两个角块需要与两个棱块一起互换。没有任何相邻块匹配。
前-右与前-左角块互换,前/右面上的两个棱块互换。
互为对角的两个角块互换,外加两个相邻棱块互换。复杂图案,没有相邻块匹配。
前-右角块与后-右角块互换,前棱与右棱互换。
前-右角块与后-右角块互换,前棱与后棱互换。
两个角块和三个棱块以右侧循环错位。一个棱块在正确位置。
Ra 的镜像。两个角块和三个棱块错位,一个棱块正确。
两对对角角块一次性互换。四个角块全部错位,四个棱块也全部错位。
类似 Na 但镜像。所有块以反射图案错位。
三个角块和三个棱块都以复杂的循环图案错位。
三个角块和三个棱块错位。Ga 图案的镜像。
三个角块和三个棱块错位。从不同角度看类似 Ga。
三个角块和三个棱块错位。从不同角度看类似 Gb。
Common mistakes
Finger tricks that make PLL fast
M-slice flicks drive the edge perms. The H-perm (M2 U M2 U2 M2 U M2) and the U-perms are meant to be flicked with the slice, not turned face by face.
Wide-U plus a re-grip is the signature of every G-perm: a double-layer u turn cycles the layer as one fluid motion, and the baked-in rotation re-presents a comfortable R/U grip mid-algorithm.
Rotations are part of the trigger. When an algorithm contains a cube rotation, treat it as one continuous motion with the moves around it — not a separate step. And memorize with your hands: PLLs decompose into familiar triggers (the sexy move R U R' U', the R U R' F' opener, the F … F' sandwich), so the hand chunks them rather than reciting them.
Going further
Once full PLL is fluent, Harris points straight at the expert horizon:
- COLL — orient and permute the corners together in one look when edges are already oriented, so PLL collapses to a pure edge case.
- ZBLL — orient and permute the entire last layer in a single algorithm, given oriented edges. The far horizon.
- VH / ZB during F2L — influence the last layer while inserting the final F2L pair so you arrive at a known, smaller LL subset.
OLL 算法
在 PLL 之前回顾顶面朝向
CFOP 入门
了解 PLL 如何融入完整的 CFOP 方法