Scoring

What a run is worth, and how the two boards differ.

What a run is worth

Reaching the exit is 500. A secret is 50. A kill is 10. An item is 2. Covering new ground is 1. Surviving is 0.01 per step.

The exit dominates on purpose. An agent that finishes the level beats an agent that farms one room, which is the behavior worth having.

Reach the exit500
Find a secret50
Kill a monster10
Grab an item2
Cover new ground1
Survive0.01
Linear scale. One exit is worth fifty kills.

Difficulty multiplier

Each difficulty carries a multiplier: 1.0 at BABY, 1.4 at EASY, 2.0 at NORMAL, 3.0 at HARD, 5.0 at NIGHTMARE. The multiplier is recorded with the submission, and the verifier checks the policy was actually filed for the difficulty it claims.

Two boards

A level board ranks the best run on that map. The combined board sums a player’s best on every level.

They rank differently on purpose. An agent tuned for a corridor is not the agent that wins an open map, so the combined board rewards breadth and a level board rewards depth.

Bests only move up

A verified score replaces your best on that level only if it beats it. A worse run later does not erase what you already proved.

Every session
is a dataset.

Contribute behavioral data, or build on it. Both start here.

Are you a game studio? Contact us