# Scalable Board Expansion within a General Game System Source: https://www.youtube.com/watch?v=8D0xEbBdtKo Recap page: https://rapidrecap.app/video/8D0xEbBdtKo Generated: 2026-01-26T23:04:36.449+00:00 --- ## Quick Overview The research paper proposes a dynamic board expansion strategy for general game systems, specifically for games like Ludo or Centenno, which overcomes the limitations of fixed grid sizes by dynamically resizing the board and recalculating indices upon expansion, leading to significantly better performance and efficiency compared to traditional fixed-grid approaches. **Key Points:** - The paper addresses scalable board expansion in general game systems, citing examples like Ludo and Centenno. - The existing method of using a fixed 41x41 grid resulted in a performance bottleneck (0.34 playouts per second) due to wasted space. - The proposed dynamic approach involves creating a new, larger ring of cells around the existing board when expansion is needed. - When the perimeter expands, the system performs a mass update, shifting all existing indices and creating a new set of valid indices for the newly added cells. - This dynamic mapping is inspired by undo logic in game engines, ensuring that the state remains correct without full recalculation. - The dynamic perimeter strategy achieved a performance of 10.18 moves per second in the worst-case straight-line expansion test, a significant improvement over the fixed grid's 0.34 moves/sec. - The core concept is managing sparse data efficiently by only updating coordinates relevant to the expansion, avoiding computationally expensive full-board reprocessing. ![Screenshot at 00:05: The main problem being addressed: fixed-size data structures \(like grids\) stunt the ability of AI to master games with naturally expanding play areas, leading to performance bottlenecks.](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-00-05.jpg) **Context:** The video discusses a research paper by Clementine Sacré and supervised by Eric Piet that introduces an alternative to fixed-size board representations in general game systems, aiming to improve performance and scalability when the game board needs to grow. Traditional methods, often relying on static, pre-allocated data structures like fixed grids (e.g., 41x41), prove inefficient when the game state requires expansion, leading to computational waste and slow performance. ## Detailed Analysis The researchers introduce a method to allow AI agents to handle board expansion in general game systems, a concept exemplified by games like Ludo or Centenno where the playing area grows dynamically. The core issue with traditional computer science data structures, like fixed-size arrays or grids (e.g., 41x41), is that they are inefficient for sparse, growing environments because they require constant updates or massive memory allocation for empty spaces. The paper proposes a dynamic approach, inspired by undo logic in game engines, which avoids this inefficiency. When expansion is required, the system adds a new ring of cells around the perimeter and recalculates all coordinates relative to the new size. This means that if a piece was at index 0, it might move to index 7 (or 0-7 in the new coordinate system). The paper shows that the perimeter expansion strategy, which dynamically adjusts the coordinate system, significantly outperforms the static grid approach. For instance, in a worst-case straight-line expansion test, the dynamic method achieved 10.18 moves per second, whereas the fixed grid method only managed 0.34 moves per second. The key advantage is that the dynamic system avoids computationally exhaustive recalculations or resets, making it fundamentally more efficient for managing growing game states. ### Problem Identification - Fixed data structures (like 41x41 grids) cause performance bottlenecks (0.34 playouts/sec) in games requiring board expansion due to wasted space and inefficient updates. ### Proposed Solution - Dynamic board expansion using a perimeter approach, inspired by undo logic, where the coordinate system is shifted and re-indexed only along the expanding boundary. ### Performance Comparison - The dynamic perimeter strategy achieved 10.18 moves per second in worst-case scenarios, vastly outperforming the fixed grid's performance (0.34 moves/sec). ### Implementation Detail - When the board expands, the system calculates an offset translation vector to remap all existing piece coordinates to the new, larger coordinate system. ### Future Implications - This method provides a blueprint for efficiently managing dynamic growth in AI systems across various domains, such as robotics mapping or general AI decision-making, by processing only necessary changes. ![Screenshot at 00:00: Title card and introductory graphic featuring two podcasters, promoting membership.](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-00-00.jpg) ![Screenshot at 00:21: On-screen text displaying the paper's title: "Scalable Board Expansion within a General Game System".](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-00-21.jpg) ![Screenshot at 00:54: Visual representation of the fixed grid structure \(the basis of the inefficient method being critiqued\).](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-00-54.jpg) ![Screenshot at 02:24: Visual demonstration of the dynamic grid allocating a massive, empty ring of cells around the existing structure.](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-02-24.jpg) ![Screenshot at 04:44: Speaker discussing the two main ways the board can physically grow: perimeter approach versus stacking/multiplying existing cells.](https://ss.rapidrecap.app/screens/8D0xEbBdtKo/00-04-44.jpg)