Counting Kangxi Strokes: Why 氵 Counts as Four

The first step in reading a name by the Five Grids is counting strokes. It is also the step most often got wrong, because name reading does not count modern strokes — it counts Kangxi dictionary strokes.

The two can differ by several strokes, and one wrong count throws off all five grids.

Where they differ

One: a radical counts as its full form

The Kangxi dictionary files characters by radical, and a radical is counted by its full character, not by the abbreviated shape it takes as a component:

• 氵 is the Water radical and counts as 4 (水), not 3. So 江 is 4 + 3 = 7 strokes, not 6
• 艹 is the Grass radical and counts as 6 (艸), not 3. So 范 is 6 + 5 = 11
• 月 as the flesh radical is 肉 and counts as 6. So 胡 is 6 + 5 = 11
• 扌 counts as 4 (手), 忄 as 4 (心), 犭 as 4 (犬)
• 辶 counts as 7 (辵)
• 阝 on the left is 阜 and counts 8; on the right it is 邑 and counts 7 — the same shape, two values

So 郭 is the right-hand 阝 (邑, 7) plus 享 (8), giving 15.

Two: simplified characters count as traditional

A simplified character is counted by its traditional form:

• 刘 counts as 劉, 15
• 张 counts as 張, 11
• 罗 counts as 羅, 20
• 叶 counts as 葉, 15

But there is a trap: not every simplified-looking character should be converted. 于, 台, 余, 范, 谷, 卜, 干, 松, 沈 and 姜 are traditional characters in their own right and must not become 於, 臺 or 餘. Treating them as simplified is a common source of wrong counts.

A reproducible stroke table

Many online tools carry hand-copied stroke tables, and one transcription error propagates forever. This site generates its table from Unicode's Unihan database instead:

Take the kRSUnicode field — radical and residual strokes — and add the stroke count of the radical's own character. A radical's own character already carries the Kangxi count, so every special case in the first section above falls out automatically. Not one of them needs a hand-written rule.

For the simplified-versus-traditional decision, the deciding field is kBigFive: if a character appears in Big5, a traditional-only character set, it already is a traditional form and is left alone; only characters outside it are converted. That single test sorts 于, 台, 余, 范 and 谷 onto one side and 刘, 黄, 孙, 罗 and 郑 onto the other, correctly.

There is also a self-correcting trick. The 214 radicals are ordered by stroke count, so the sequence must never decrease. Applying that rule automatically fixed four radicals — 瓦, 禸, 鬼 and 鼎 — whose modern counts are one stroke short of the Kangxi ones.

The resulting table is 21KB, and the counts for 89 common surnames were checked one by one against standard name-reading references.

Known deviations

Stating the limits plainly: Unihan analyses radicals against modern glyph shapes, and a few characters end up one stroke away from the Kangxi dictionary's own analysis — 戴 (Kangxi files it under 戈 with 14 residual, giving 18, while Unihan yields 17) and 萬 among them. These are corrected one at a time in the generator script.

Characters like 叶, where the simplified and traditional forms coincide, are genuinely ambiguous: 叶 is a variant of 协 as well as the simplification of 葉. This site follows surname convention and reads it as 葉, 15 strokes.

See it for yourself

Enter a name and every character's Kangxi count is shown alongside the grids it produces. The generator script is public too, in the project repository under scripts/.