feat(web): one rating everywhere from the stored bayesian score #1545

Activity

1. JOY (JOY)

   added

   status:readyTask ready to be picked up by an AgentTask ready to be picked up by an Agent

   on Oct 31, 2026
2. JOY commented on Oct 31, 2026

   

   JOY (JOY)

   on Oct 31, 2026

   ContributorAuthor

   More actions

   Implementation plan (Web)

   I read the P2 migration in PR #1549 (Docs/Proposals/2026-10-31-dos-me-criteria-ratings.sql):

   • master_products.rating_score is numeric(3,2): Bayesian, k = 3, mu = 4.2, NULL when there are no reviews.
   • rating_weight_sum is also added.

   Every place that switches to rating_score (after P2 is applied)

   There is one reader change, and the rest follows from it:

   • resolveCompareEffectiveUserRating (lib/rate-domain.ts) reads rating_score first. Fallback to user_rating_weighted stays until the backfill is checked.
   • scoreMasterProductForLeaderboard userScore5 (lib/leaderboard-engine.ts) reads the same helper.

   Direct readers of user_rating_weighted or average_rating, which move to the helper:

   • Product page (smartphones/[slug]/page.tsx):
     • the headline computeBayesianRatingSummary reads the stored score, keeping the review list only for counts;
     • the peer cards.
   • Catalog cards: smartphones/page.tsx, rate-api.ts (the card value and the sort on user_rating_weighted).
   • Home: page.tsx, hero-products.ts, HeroSearchInspector.
   • Leaderboard: leaderboard/page.tsx.
   • Deals: server/rate/deals.ts.
   • Other callers of resolveCompareEffectiveUserRating: compare, watchlist, deep inspect, catalog toolbar. They follow automatically.
   • API: /api/v1/rate/* select lists add rating_score. Responses add rating_score_5, and the existing fields stay for one release. server/rate/openapi.ts documents it.

   Ranking follows D3:

   • The leaderboard ranks by rating_score. The composite only breaks ties.
   • Battle cards, leaderboard rows and the podium show rating_score at 1 decimal.
   • A persona or criterion view keeps its own score, labelled with the persona or criterion. It is not shown as "the rating".
   • display-truth tests follow.

   The three follow-ups from PR #1478

   1. Versus API tie rule. It can land before the migration. is_tie and overall_winner use the score at one decimal, like the page. The composite and price become the tiebreak only. margin_10 is the gap between the scores as shown, and the OpenAPI description follows. Switching the API's base score to rating_score waits for the column.
   2. Composite or persona scores at 2 decimals on battle cards, the leaderboard and deals. This waits for the column: showing a 1-decimal number that the list is not sorted by would make equal-looking scores rank differently.
   3. Small fixes:
      • "0.0" when neither model has a score. It can land before: the summary and verdict say "Chưa có đánh giá" like the stars.
      • Double rounding. I propose no code change before the column. The double rounding goes away when the screen reads rating_score directly. But the column itself is stored at 2 decimals, so a Bayesian 4.249 is stored as 4.25 and shown as 4.3. To round only once, P2 would store numeric(4,3) (or wider) and the screen would round to 1 decimal. A question for the P2 author (PR docs(proposals): dos-me migration for criteria ratings, pros and cons (p2) #1549), not a blocker: the error is at most 0.05 and only near x.x5.

   Order

   • PR 1 (now, no new column): follow-up 1 and the empty "0.0" state. I am claiming feat(web): one rating everywhere from the stored bayesian score #1545 for it.
   • PR 2 (after P2 is applied and V1 to V6 pass): the reader switch, the ranking, the API field and OpenAPI, and follow-up 2.
3. JOY (JOY)

   added

   status:in-progressTask currently claimed and being worked on by an AgentTask currently claimed and being worked on by an Agent

   agent:webClaimed by Agent: WebClaimed by Agent: Web

   and removed

   status:readyTask ready to be picked up by an AgentTask ready to be picked up by an Agent

   on Oct 31, 2026
4. JOY (JOY)

   self-assigned this

   on Oct 31, 2026
5. JOY (JOY)

   added a commit that references this issue on Oct 31, 2026

   fix(web): versus api ties at one decimal and unrated pairs say so (#1551

   ...

   eaf6cf7