Scorecards - ARC-AGI-3 Docs

Scorecards aggregate the results from your agent’s game performance.

Ways to Get a Scorecard

Manually — See Full Play Test for details
Python Toolkit — See Get Scorecard guide
Swarm — Running a swarm will automatically open/close a scorecard for each agent

For game runs done via the API, scorecards can be viewed online at https://arcprize.org/scorecards and https://arcprize.org/scorecards/`scorecard_id`.

Scorecard Fields

Field	Description
`card_id`	Unique scorecard ID
`score`	Average of the best score for each environment
`environments`	Per-environment runs, scores, action counts, completed levels, and reset counts
`tags_scores`	Aggregate results grouped by game metadata tags
`total_environments_completed`	Number of environments with a winning run
`total_levels_completed`	Levels completed across the best environment runs
`total_actions`	Actions taken across all recorded runs
`tags`	Labels used to categorize and filter scorecards (e.g., `["experiment1", "v2.0", "test"]`)
`source_url`	Optional source or provenance URL
`opaque`	Optional JSON-serializable metadata
`competition_mode`	Whether competition restrictions apply to the scorecard

import arc_agi

arc = arc_agi.Arcade()

scorecard_id = arc.create_scorecard(
    tags=["experiment", "my-awesome-agent-v5-final-final"],
    source_url="https://github.com/arcprize/docs",
    opaque={"custom_field": "any data"}
)

For more information, see Create Scorecard. Scorecards are not public, however you can share replays from scorecards created via the API with others. Local scorecards cannot be shared.

Notes

Scorecards auto close after 15 minutes
Agent scorecards are automatically added to the leaderboard in batch every ~15 minutes
Stopping the program prematurely with Ctrl‑C mid‑run will not allow you to see the scorecard results