Enhancing the Performance of Dynamic Scripting in Computer Games

Unsupervised online learning in commercial computer games allows computer-controlled opponents to adapt to the way the game is being played. As such it provides a mechanism to deal with weaknesses in the game AI and to respond to changes in human player tactics. In prior work we designed a novel technique called “dynamic scripting” that is able to create successful adaptive opponents. However, experimental evaluations indicated that, occasionally, the time needed for dynamic scripting to generate effective opponents becomes unacceptably long. We investigated two different countermeasures against these long adaptation times (which we call “outliers”), namely a better balance between rewards and penalties, and a history-fallback mechanism. Experimental results indicate that a combination of these two countermeasures is able to reduce the number of outliers significantly. We therefore conclude that the performance of dynamic scripting is enhanced by these countermeasures.