Wubi (五笔) is a way of typing Chinese by shape rather than by sound: each character is broken into a few component shapes, and each shape lives on a key. The character 汉 is typed I C Y. The I key carries the water radical 氵 on the left of the character, and the C key carries 又 on the right. Y is not part of the character at all. It is a one-key summary of how the character ends and how it is laid out, and that third key is the idea the rest of Wubi is built on.
Wubi is a shape-based input method invented by Wang Yongmin (王永民) in the 1980s, and it remains a mainland Chinese method: typists in Taiwan more often use Zhuyin, and in Hong Kong Cangjie or its simplified form, Quick (速成). This guide works through eight characters by hand in the 1986 edition (86版), the version most tutorials teach, then counts how often two characters end up with the same code. Every code below was checked in two independent Wubi code tables, and you can look up any character in our Wubi code lookup.
Five zones, grouped by first stroke
Wubi breaks characters into about 130 building blocks, called 字根 (roughly, "character roots"). Wang's own teaching text, as excerpted, speaks of 125 kinds; the usual figure of 130 counts some variant shapes separately. The roots are spread over 25 letter keys — Z is left free — in five zones, and a root's zone is set by its first stroke.
| Zone | First stroke | Keys, in order |
|---|---|---|
| 1 | 横, horizontal | G F D S A |
| 2 | 竖, vertical | H J K L M |
| 3 | 撇, falling left | T R E W Q |
| 4 | 捺 or 点, falling right or dot | Y U I O P |
| 5 | 折, turning | N B V C X |
Each key has a number from its zone and position: G is 11, H is 21, T is 31, Y is 41, N is 51. Those numbers matter in a moment.
Working 汉 through, one key at a time
The rule for a character's full code depends on how many roots it breaks into. With four or more, you type the first, second, third and last roots and stop. With fewer than four, you type every root and then add one more key: the 末笔字型识别码, the last-stroke-and-layout identifier.
汉 has two roots, 氵 and 又, so it takes the identifier. Work it out in two parts.
- The last stroke. The character's final stroke is the falling-right stroke that ends 又. That is stroke type 4.
- The layout. 氵 sits to the left of 又, so the character is left-right. Wubi numbers layouts 1 for left-right, 2 for top-bottom and 3 for everything else.
The two digits—4, then 1—give the key number 41, which is Y. The full code for 汉 is therefore I C Y. The identifier is always a single key; in this case, it brings the code to three keys instead of four.
The full table of identifiers is small enough to learn:
| Last stroke | Left-right | Top-bottom | Other |
|---|---|---|---|
| 横 | G | F | D |
| 竖 | H | J | K |
| 撇 | T | R | E |
| 捺 | Y | U | I |
| 折 | N | B | V |
Seven more, each adding a rule
字 is P B F. 宀 is on P, 子 on B. The final stroke of 子 is a rising stroke, which Wubi counts as horizontal, and the character is top-bottom, so the identifier is 12: F.
困 is L S I. 囗 is on L and 木 on S. For an enclosed character, the last stroke is taken from what is inside the box, not the stroke that closes it. 木 ends in a falling-right stroke, and enclosure counts as "other", so the identifier is 43: I.
国 is L G Y I. Three roots — 囗, 王 and the dot — so it takes an identifier. The enclosed part ends in a dot, the layout is "other", and the identifier is again I.
仇 is W V N. 亻 is on W and 九 on V. Some roots have fixed last-stroke conventions: 九, 力, 刀, 匕 and 七 always count as ending in a turning stroke. Turning plus left-right gives 51: N.
键 is Q V F P. This one breaks into five roots: 钅, 彐, 二, 丨 and 廴. Four is the limit, so you take the first three and the last, and the fourth root, 丨, is simply dropped.
金 is Q Q Q Q. Each key has one "key name" character — the one its roots are named after — and it is typed by pressing that key four times. 金 is Q's key name.
五 is G G H G. 五 is itself one of the roots, so it follows a different rule: type the key it lives on, then the keys for its first, second and last strokes — horizontal, vertical, horizontal.
How often two characters share a code
A code of at most four letters from 25 keys has room for hundreds of thousands of combinations, but Wubi's rules do not spread characters evenly across them. When two characters produce the same code, they are called 重码, duplicates, and the typist has to choose from a list.
Counted across all 6,763 characters of the GB2312 standard, using the full code of each: 510 characters, 7.5%, share their full code with at least one other character, spread over 249 codes. Some groups pair an everyday character with components nobody types on their own: 了 shares B N H with 凵, 卩 and 阝, although 了 is normally typed with the single key B. Others set ordinary words against each other. 云 ("cloud") and 去 ("go") are both F C U, and 万 ("ten thousand"), 尢 and 尤 are all D N V.
That figure sits beside a claim often made for Wubi — that a fluent typist rarely has to pick from a list — and both are true, for two reasons. First, input software ranks duplicates by frequency, so the more common character is offered first and a single press of the space bar accepts it. Second, Wubi gives many common characters shorter, unique codes. Of the 510 duplicated characters, 142 have a shorter code that no other character shares; 支, for instance, can be typed F C and is never confused with 云 or 去. The 7.5% figure, then, describes the code table itself, not what a typist encounters when writing ordinary text.
How it differs from Cangjie
Cangjie, widely used in Hong Kong, also takes characters apart by shape, but it handles short characters differently. Where Cangjie stops when it runs out of components to encode, Wubi adds the identifier. This gives a two-root character a third key that describes the character’s overall structure, a piece of information Cangjie doesn’t encode. We took Cangjie apart in the same way, one character at a time, in a guide to how Cangjie turns a character into five keystrokes.
Wubi's editions differ too. The 1998 edition reorganised the roots and changed some codes — 我 is T R N T in the 1986 edition and T R N Y in 1998 — and later editions widened the character set. The codes here use 1986 because it remains the edition most tutorials teach.
How the codes were checked
We checked every code in this guide against two independent, openly published Wubi 86 code tables. Each example follows the rules for key layout and identifiers as implemented in those tables. The duplicate count was run on 21 September 2026 against the code table for our lookup tool, using the longest code for each of the GB2312 characters. A different edition or a larger character set would produce different numbers.
View author profile → · Editorial policy
