On the Complexity of General Game Playing

The Game Description Language (GDL) used in General Game Playing (GGP) competitions provides a compact way to express multi-agents systems. Multiple features of GDL contribute to making it a convenient tool to describe multi-agent systems. We study the computational complexity of reasoning in GGP using various combinations of these features. Our analysis offers a complexity landscape for GGP with fragments ranging from np to EXPSPACE in the single-agent case, and from pspace to 2EXPTIME in the multi-agent case.