Measuring risk in stealth games

Level design for stealth games requires the ability to explore and understand the possible paths players may take through a given scenario and how they are impacted by different design choices. Good tool support can help by demonstrating the existence of such paths, but for rapid, interactive design, the relative difficulty of possible solutions also needs to be quantified, in a way that correlates well with human perception of risk. Here we propose and evaluate three different metrics for defining and quantifying the risk of stealthy paths. We validate and compare these measures through a small human study, showing that a simple path-distance measure correlates best with human judgement. An evaluation of a non-trivial stealth scenario demonstrates the practicality of our approach, and shows how such measures can be useful in understanding a level design.