[SunrisesIllNeverSee]

What

compareTally() counted 10 contests, with all 4 raw pillars as more-wins, including fresh input. Fresh input is the denominator of Yield (Υ = cache_read × output / input²), so using less of it for the same output is what the product rewards. Scoring more input as a win contradicted "Volume isn't Yield".

• lib/analytics/compare-facts.ts: fresh input is now LOWER-wins. Output, cache read and cache write stay higher-wins (volume).
• components/compare/CompareLedger.tsx: the Input row uses lower-wins for its winner and bar. The Raw band is labelled "Raw · Flow · Total".
• components/share/CompareMatchupCard.tsx: its private copy of the tally (which would have kept the old rule) is replaced by the shared compareTally.
• CompareMatchup, the LEADING chip, the page score, CompareMobileScore and the Context panel already read compareTally, so they follow automatically. deriveFacts has no input winner logic. CompareVersus is unused and has no input axis.
• No metric definitions changed. Canon context loaded; the change only flips which direction wins on a raw row.

MO§ES vs Kabir (live numbers from the task)

Input 25,766,329 vs 387,900; output 199.3M vs 192.0M; cache read 51,791.3M vs 50,412.5M; cache write 2,737.8M vs 939.2M; Kabir wins all 6 derived metrics.

• Before: MO§ES 4, Kabir 6 (MO§ES "won" input, output, cache read, cache write).
• After: MO§ES 3, Kabir 7 (Kabir wins the 6 derived metrics plus input).

Tests

• New __tests__/board/compare-tally-input.test.ts (3 cases, including the MO§ES vs Kabir shape).
• npx tsc --noEmit clean, eslint clean on changed files, npm run test:ui 397 passed, npm test 695 passed (4 skipped).
• Local visual check on /compare (H82 vs Kabir, local data): ledger, chip, Context panel and page all show 1–9. Not checked on a phone viewport, since the mobile score reads the same function.

Not done

• Did not touch lib/board/queries.ts (PR fix(board): profile + share cards show the board's Yield and rank #322). Not merged (merging deploys).

🤖 Generated with Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 62853b7e-d315-4504-b77e-961f53acafa4

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}