This study investigates the climbing dynamics of learning on a long-time scale, by using Drifting Markov models. Climbing constitutes a complex decision-making task that requires effective visual-motor coordination and exploration of the environment. Drifting Markov models, is a class of constrained heterogeneous Markov processes that allow the modeling of data that exhibit heterogeneity. By applying the later models on real-world visual motor skill data, we aim to uncover the persistent dynamics of learning in climbing. To that end a real case study is conducted based on an experiment, with results that (a) help in the understanding of skill acquisition in physically demanding environments; and (b) provide insights into the role of exploration and visual-motor coordination in learning.