The MTMD LC 21-102 test method is an accelerated polishing by projection test, used in Quebec (Canada) to simulate the wear of coarse aggregates used in pavement surface layers, regarding skid resistance. Their suitability for use in construction is then evaluated by measuring their residual friction coefficient using a British pendulum. This study aims to investigate the effect of polishing time, better understand the wear mechanisms through microtexture evolution, identify the intrinsic properties of aggregates that govern polishing resistance and develop predictive models based on these influential properties. The research was carried out in two phases. The first focused on four aggregates and involved monitoring microtexture and friction parameters at incremental polishing stages. The second involved eighteen aggregates and focused on identifying the most influential properties using a broader dataset. The friction and microtexture parameters were measured before and after polishing. Mineralogical, petrographic and mechanical properties were assessed through optical microscopy, X-ray diffraction (XRD), Los Angeles and Micro-Deval tests. A 3D laser microprofilometer was used to capture the surface relief of aggregate particles and to determine their microtexture parameters such as peak density, shape and height. A British pendulum was used to measure the friction coefficient. The results show that the current standard polishing time may underestimate aggregate wear: friction continued to decrease beyond the specified duration, suggesting the need to extend the polishing time to reach maximum wear. Polishing by projection also revealed distinct wear mechanisms compared to other well-known methods: it acts more by indentation, digging into the aggregates surface and generating a new microtexture with, on average, less dense but taller and sharper peaks. Hard aggregates such as granites, greywackes and gneiss showed superior polishing resistance and better microtexture regeneration compared to softer materials like basalts, dolostones and limestones. The most influential properties were relative hardness (RHD), differential hardness (DH), mineral contents (quartz, feldspars, calcite), average mineral size (φm) and grain size distribution parameters (coefficient of variation CV and Gini-style index GSI), which showed the highest Pearson correlation coefficients with the final British Pendulum Number (BPNf). Predictive models developed with these properties achieved high accuracy (R² = 0.89 to 0.94), offering valuable tools for selecting aggregates with adequate skid resistance without performing polishing tests.