How Ketroc's Banshee Micro Works

In this session of Postbot, Drekken sits down with @Ketroc to discuss his bot’s advanced micro-management, particularly focusing on Banshee units and their interaction with enemy forces. Ketroc delves into the mechanics behind his bot’s decision-making processes, pathfinding techniques, and overall strategy.

Highlights

  1. Banshee Micro-management:
  • Ketrock explains the intricate micro-management of Banshees, showcasing how they interact with enemy units and navigate the map.

  • He demonstrates the use of a threat map with color codes to represent different levels of threat and detection.

  • The Banshee units use a custom pathfinding algorithm, likened to a “blind person with a stick,” to move around the map and avoid danger.

  1. Code Reusability and Flexibility:
  • Ketrock emphasizes the importance of having a basic micro class that can be extended for specific units and scenarios.

  • This approach allows for easy customization and reusability of code, making it efficient to manage different unit types.

  1. Advanced Targeting and Threat Assessment:
  • The bot includes advanced features like assessing whether a Banshee should cloak based on the surrounding threats and its health.

  • Ketrock also discusses how his bot detects upgrades like Blink for stalkers by observing their movement patterns.

  1. Strategic Decision-Making:
  • Ketrock’s bot can dynamically reassign units to different tasks based on the current game state, such as releasing a unit from one task to handle a more urgent threat.

  • The bot can also prioritize targets, like preferring to attack a War Prism over other nearby units, based on the current strategy.

  1. Influence Mapping and Pathfinding:
  • Ketrock explains the use of influence maps to determine safe and dangerous areas, taking into account enemy vision and terrain levels.

  • The bot’s custom pathfinding algorithm helps navigate these maps efficiently, despite not using traditional pathfinding methods.

If you want to present what you’ve been working on reply to this thread!