[Stanback]

Problem

On the converter's Compare methods card, each spectrum pin centred its name and time under its dot. When two methods convert to nearby times, the labels printed over each other. With women's 1650 SCY 17:36.20 → 1500 SCM, NCAA (17:33.04) and Equivalence (17:45.52) are about 10% of the track apart, so the card showed "NCAAEquivalence" and "17:33.0417:45.52".

Fix (src/lib/MethodComparison.svelte)

• Each dot stays on a zero-width anchor at its true position, the same anchor idea as the MCP card.
• The track and labels are measured with Svelte bind: bindings. Each label centres on its pin when it fits, is clamped inside the track, and moves to the first lower row where it touches no other label, with a leader tick back to its dot. Rows where no leader runs through a label are preferred. Labels have an opaque surface background, so an unavoidable leader passes behind text, not over it.
• The row pitch comes from the measured label height, so enlarged text still separates. Before measuring, labels are centred under their pins with a rem fallback pitch.
• The label is now the <button>, so its focus ring and hit area match what is drawn.
• No new dependencies, and the common cases look the same as before (e.g. 100 free → LCM, where all three labels fit on one row).

Test

New Playwright case method spectrum keeps close method labels apart and inside the card. It uses the 1650 example at desktop and 390px, in light and dark, and asserts that:

• no two name/time boxes overlap
• every label stays inside the card
• dots keep their true order and relative spacing

It fails on main with "NCAA overlaps Equivalence".

Other surfaces

• iOS / Android: these show only a "Method spread (all 3)" text row, with no spectrum, so they're unaffected.
• MCP card: has the same collision (at 390px "NCAA" spans x 76–109 and "Equivalence" x 91–159). It's left unchanged here because the fix isn't small. The card builds its DOM imperatively, so it needs post-layout measurement, a resize observer and harness changes. It's best as a follow-up.

Verification

npm run check, npm test (187), npx playwright test (57 passed, 1 pre-existing mobile-only skip), node --test scripts/ci-changes.test.mjs scripts/native-test-support.test.mjs (84).

🤖 Generated with Claude Code