ADAPTIVE THEORY OF GAMES AND MILITARY STRATEGY,
Abstract
Game theory provides a way of finding a so-called optimal strategy, that is, a method by which a player can maximize his average gain in a competition knowing nothing of his opponent's strategy. In actual games or military campaigns the opponent will play with some bias, either consciously or subconsciously. Under such conditions it would be possible to adapt one's strategy according to the opponent's play. It is the intent of this paper to prove that such an adaptive strategy results in a higher average gain than could be expected from a fixed optimal strategy. Computers were used to simulate battles and matrix games with human opponents at remote consoles. Of particular interest was a program that simulated a tank-to-tank battle. A display driven by a paper tape punched on line was constructed to show tank positions. Each tank could maneuver in various directions to avoid enemy fire while aiming and firing at the other tank. It was hoped that such a war game would closely approach combat conditions. The computers played by both optimal strategy and adaptive strategy. In the latter case statistical biasing was used to weight the computer's play according to its opponent's strategy. In the majority of the games played the computer's average payoff was significantly higher when an adaptive strategy was used. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1969
- Accession Number
- AD0694406
Entities
People
- Craig A. Decker Jr
Organizations
- Harry Diamond Laboratories