Built this: Orbit-Wars strategy agent

On 24 April 2026 I pushed Orbit-Wars, an agent for the Kaggle Orbit Wars environment, written in Python.

How the agent plays

From the README:

• Economic priority scoring: rank target planets by production value, distance and current ship count, so high-production planets are taken first.
• Threat assessment and “fortress” logic: find planets that are about to fall and reinforce them from nearby bases.
• Fleet sizing: send a margin above what’s needed (the README gives the formula 2.5x + 20) for a likely capture without stripping defences.
• Sun bypass: detect paths that would hit the sun and steer around it.
• Predictive intercepts: compute where an orbiting planet will be at the fleet’s arrival time.
• It also accounts for fleet speed depending on fleet size, and chases high-value comets.

test_agent.py runs a match against a random opponent and saves an HTML replay.

Where it connects

This is a hand-designed agent rather than a learned one, a useful contrast for my agentic AI notes, where “agent” usually means an LLM choosing actions.