Player-adaptive Spelunky level generation

Abstract

Procedural Content Generation (PCG) is nowadays widely applied to many different aspects of computer games. However, it can do more than to assist level designers during game creation. It can generate personalized levels according to the tastes and abilities of players online. This has already been demonstrated for (largely 1D) scrolling games and we show in this work how personalized, difficulty-adjusted levels can be generated for the more complex 2D platformer Spelunky. As direct and indirect player feedback is taken into account, the method may be filed under the Experience-Driven PCG approach. Our approach is based on a rather generic rule set that may also be transferred to similar games. We also present a user study showing that most users appreciate the online adaptation but are especially critical about making the game easier to play at any time.