This directory contains a random agent baseline for Super Mario Bros, providing a performance reference point for evaluating trained agents.
The baseline runs a completely random agent that selects actions uniformly at random from the available action set. This establishes the minimum performance level that any learning agent should exceed.
- Random Actions: Takes random actions from the complex action set (12 actions)
- Performance Measurement: Tracks scores, stages reached, and survival time
- Video Recording: Optional gameplay recording for qualitative analysis
- Statistical Analysis: Collects performance statistics over many episodes