PandaChessBoardsthe ranking room
0 calls

Every pairing, one at a time

How good is this board?

Ninety-six boards — every palette wearing every piece set. Give each one a number from 1 to 10. The scores are kept locally and turned into a preferences file the agent can read, so next time it already knows which board to reach for.

Board 1 of 96

—

Wouldn't ship itShip it today

1–9, 0 for ten · S skip · Y ship it · U undo

0 of 96 judged—

Two boards enter

Which one would you actually use?

Numbers drift; a straight choice doesn't. Pick a side — the winner takes the round, and the result is folded into the same score.

0 rounds

← left · → right · ↓ tie

The full inventory

Every board in the system.

Sixteen palettes across six piece sets. Click any one to judge it on the spot.

Where things stand

The order of preference.

Scores blend your 1–10 calls with face-off results, held toward a neutral 5 until enough evidence comes in — so one lucky rating can't crown a board.

Piece sets, ranked

Average across every board that set appears on.

What your ratings say

Traits that move the score, versus the collection average.

The record

Everything you told it.

Every call is stored in this browser — nothing leaves the page unless you point it somewhere. Export the log as data, or take the distilled ruleset straight to CLAUDE.md.

CLAUDE-chessboards.md — generated from your ratings 

  


  

    
    
    
  

  
Each call is POSTed as one JSON object — board, set, score,
    how long you looked, and the design code. Leave it blank and nothing is sent.