Model-Based Difficulty Estimation of Indoor Bouldering Routes

Abstract

In this thesis, we describe an application-specific model to calculate the difficulty of a bouldering route. Current bouldering routes are graded by the route setters themselves which can introduce a bias. Our model, on the contrary, uses a set of tailormade rules to find the least-cost path for a given route, and extracts the difficulty from this path. This thesis introduces a novel data capture method, with which routes at an indoor bouldering gym can be recorded accurately. We evaluate our model by supplying these captured routes of varying difficulty to our model and comparing the actual order of grades with the order of predicted difficulty values. Several models of increasing complexity are tested for an increase in the accuracy of the difficulty estimation. Our results showed that a model that incorporates the type of each hold was able to predict the relative order of difficulty for each pair of routes in 70% of the cases.