These are counts in one KanjiVG snapshot. They do not say that a larger component is written first, last, or in between because of its size. They are not a rule of stroke order, and they make no claim about meaning.

On this page

What is counted

A component
One top-level slot of the character group, the slot the valence atlas counts. A run of bare strokes is one slot. Sibling groups that share kvg:element and kvg:number and carry kvg:part are one slot. A character is in this page when that count is 2 or more.
Area share
The axis-aligned box of the samples along that component’s strokes, divided by the same box of the whole character. A cubic or a quadratic is sampled as 24 equal-parameter chords. The share is stored in thousandths. The boxes may overlap, so the shares of one character need not add to 1.000. A share of 1.000 means that component’s box equals the character’s box.
Start rank
The order of the components’ first strokes. Rank 1 is the component whose smallest stroke number is the smallest. The stroke numbers partition the character, so the ranks run from 1 through the number of components.
First
One component has the greatest stored share, and its start rank is 1.
Last
One component has the greatest stored share, and its start rank equals the number of components.
In between
One component has the greatest stored share, and its start rank is neither 1 nor the last rank. A character of two components has no such rank.
Tie
Two or more components share the greatest stored share. The character is not also counted as first, last, or in between.
Position
kvg:position on the largest component, mapped as on Radical positions. left, right, top, bottom, kamae, tare, and nyo are へん, つくり, かんむり, あし, かまえ, たれ, and にょう. No position is なし. Any other stored value keeps its own row. On a tie, the row is that position only when every component at the greatest share has it. Otherwise the row is 複数.

Where the largest begins

One row is one number of top-level components. The character is counted once. The four place columns add to the character column. The total is the three rows.

Components Characters First Last In between Tie

Position of the largest

One row is the position of the largest component. The character is counted once. The four place columns add to the character column. The rows add to the total above.

Position Characters First Last In between Tie

Shares, by position

Each bar is that row’s characters, stacked to the same height. The table has the counts. The number under a bar is the character count.

Mean area share by start rank

One row is one start rank, over every counted character that has a component at that rank. The mean is of the stored shares.

Start rank Components Mean share

The same mean inside one arity

Rank 1 of a two-component character is not mixed here with rank 1 of a four-component character. Each block adds to the component count of that arity.

Components Start rank Components Mean share

Notable examples

Character Components Strokes Where the largest begins Position Shares, in start-rank order Gap

Plates

Cinnabar strokes are the largest component. On a tie, cinnabar is the tied component that starts earlier and indigo is the one that starts later. Click a plate to replay. With reduced motion the finished character stays still, and the highlight stays. The plate stays cream in either theme.

Largest begins first

Largest begins last

A tie

In between

What this shows

Stroke diagrams: KanjiVG, Ulrich Apel et al., CC BY-SA 3.0, from commit 70a0b7ae0c18ceb5cb358274b029cce0234a43bc of github.com/KanjiVG/kanjivg. The plate files are unmodified copies in site/assets/kanjivg/. Component slots: Valence atlas. Position families: Radical positions as ordering slots. Jōyō index: the list used on the Gallery.