Chengyu are the fixed idioms of written Chinese, and the free dictionary CC-CEDICT carries thousands of the four-character kind. We counted 5,056 of them: the 5,045 that CC-CEDICT tags as idioms, plus 11 we added to the list. Of those, 4,645 use four different characters. Only 411 repeat one. Of the 411, 216 follow the ABAC pattern, as in 一丝一毫 (yī sī yī háo), the merest shred. They repeat 100 different characters between them, and the commonest frame, 不X不Y, accounts for 26. The doubled AABB shape holds 29. Typologies of these patterns exist, but we found no published share across a whole dictionary, and this is one dictionary's selection, released under CC BY-SA 4.0.
ABAC stamps out most of the repeats
The letters are placeholders for characters. In an ABAC pattern, the first and third characters match; in AABB, each of two characters is doubled. Shares are of the 5,056 distinct idioms, on simplified forms; no idiom changes shape in traditional script. English glosses throughout are ours.
| Pattern | Idioms | Share | Example |
|---|---|---|---|
| ABCD (no repeat) | 4,645 | 91.87% | 一丘之貉 |
| ABAC | 216 | 4.27% | 无影无踪 |
| AABC | 71 | 1.40% | 井井有条 |
| ABCC | 47 | 0.93% | 人心惶惶 |
| ABCB | 35 | 0.69% | 人云亦云 |
| AABB | 29 | 0.57% | 兢兢业业 |
| ABCA | 10 | 0.20% | 为所欲为 |
| ABAB | 2 | 0.04% | 彼此彼此 |
| ABBC | 1 | 0.02% | 心怦怦跳 |
ABAC is 52.6% of the repeats, on our arithmetic, but it is a position pattern and no single frame. 不 starts 26 of the 216. 一 and 无 start 15 each, and 有 starts 12. 无影无踪 (wú yǐng wú zōng) means vanished without a trace. The two ABAB idioms, 彼此彼此 and 悠哉悠哉, each say a two-character word twice.
Numeral characters press into nearly a tenth of the list
We searched for 一二三四五六七八九十百千万. 446 idioms contain at least one, 8.8% of the list. Adding 两 (liǎng, two of something) raises it to 462, or 9.1%. Counting every occurrence and including 两, 135 idioms hold two or more numerals, as 一举两得 (one move, two gains) does.
一 appears in 244 idioms, 259 times in all. 千 has 49, and 17 idioms pair it with 万, as 千军万马 does for a vast army. A matched character is not always a count. In 一丘之貉 (yī qiū zhī hé) the 一 means the same: raccoon dogs from one hill, birds of a feather.
Horse characters leave the most impressions among the animals
Our animal list has 40 characters: the 12 of the zodiac, 鼠牛虎兔龙蛇马羊猴鸡狗猪, and 28 others, 鹿狐狼熊象豹狮鱼鸟雀燕鹤鹰凤龟蛙蝉蚁犬驴猿鸭鹅鸦鸿蛟虫鹏. 309 idioms (6.1%) contain at least one of them, and 213 (4.2%) a zodiac animal.
The horse, 马, is in 67 idioms and occurs 69 times. The tiger follows with 38 idioms and the dragon with 30. 69 idioms hold two different animal characters. This is a character match and says nothing about meaning: 马马虎虎 (mǎ ma hū hū in CC-CEDICT) holds a horse and a tiger and means careless, just about passable.
The list casts 不 in every slot but the last
不 (bù), the negative, occurs 492 times, more than any other character. It is first 119 times, second 116 and third 257, and it never ends an idiom. 无 ends just two.
1,499 different characters open an idiom and 1,633 close one. 一 opens 148, and 人 (person) closes the most, 44. Of the 1,633 closing characters, 766 open no idiom in the list. In 成语接龙, the game of chaining idioms by their last character, those 766 are walls; the jielong guide maps the dead ends in a larger list.
Traditional pages forge their chain links from traditional characters
This site's chengyu finder plays that chain over the same 5,057 entries; 难兄难弟 has one for each of its two readings. Press Chain and it lists idioms opening on the final character, in the script the page shows, then two sound lists: same syllable and tone, and same syllable in another tone. On the English and simplified pages, across every entry, those two sound lists hold 128,110 suggestions: 32,825 (25.6%) share the tone and 95,285 (74.4%) do not.
To re-run one, open the traditional-script page, search 一言不發 and press 接龍, its label for Chain. Nine idioms open on 發. 髮指眥裂 (bristling with fury) shares the simplified form 发 but is read fà, so it sits in the other-tone list. On the simplified page, Chain on 一针见血, which ends on xiě, lists 12 idioms opening on 血, all read xuè. Of 25,389 same-character suggestions on those pages, 227 change the syllable like this, and about 490 more keep it and change only the tone. Sound matches use CC-CEDICT's citation tones, so the spoken tone of 一 never enters the chain.
Until a revision dated 7 November 2026, the finder's FAQ called its data public domain, though the list, pinyin and glosses are CC-CEDICT text under CC BY-SA 4.0. Its one same-sound list ignored tone, and traditional pages matched either script, giving 166 suggestions that opened on the wrong traditional character.
The list derives from CC-CEDICT (2026-05-24 release), used under CC BY-SA 4.0; rebuilt from that file, it matches byte for byte.
Taiwan's 成語典 hallmarks under half of a seeded sample as main entries
We drew ten idioms from the list with seed 213 and looked each up in Taiwan's Ministry of Education 《成語典》 on 7 November 2026, searching its main text and its appendix for the traditional form. Four of the ten are main entries: 唯唯諾諾、差強人意、花天酒地、拭目以待. Four appear only in the appendix: 精神恍惚、我醉欲眠、火上加油、吞吞吐吐. Two, 一切就緒 and 兼濟天下, were absent. Ten is a handful, so we give no rate.
An appendix hit is not a ruling that a phrase is no idiom. All four sit in the appendix list 「修訂本參考資料」. The proverbs guide explains how main entries divide.
Readings matched CC-CEDICT on three of the four main entries. For 唯唯諾諾 (yes-yes, meekly agreeing) the ministry gives 「ㄨㄟˇ ㄨㄟˇ ㄋㄨㄛˋ ㄋㄨㄛˋ」, wěi in our romanisation, where CC-CEDICT gives wéi. Outside the sample, it reads 誰 in 鹿死誰手 as 「ㄕㄨㄟˊ」 (our shuí), against CC-CEDICT's shéi. In this list CC-CEDICT prints 一 as yī all 259 times. For 一針見血 the ministry adds a changed reading, 「（變） ㄧˋ ㄓㄣ ㄐㄧㄢˋ ㄒㄧㄝˇ」, the tone heard in speech. Its licence page releases the dictionary under CC BY-ND 3.0 Taiwan. Credit: 中華民國教育部（Ministry of Education, R.O.C.）。《成語典》（版本編號：2020_20260929）網址：http://dict.idioms.moe.edu.tw/
Longer and shorter idioms break the four-character mould across CC-CEDICT
The counts above stop at four characters. CC-CEDICT tags 5,839 distinct headwords as idioms, and 5,045 of them, 86.4%, have four Han characters. Only 22 are shorter, the smallest at two characters. The other 772 are longer, and 303 of those run to eight characters or more. Four characters are its commonest idiom length, never its rule. The ministry's own definition says much the same of 成語 (chengyu) in general, and the proverbs guide quotes it. Any share across chengyu as a whole would have to start from a list that keeps the long ones.
