<p>Earth-rock dams play an indispensable role in the national water network and sustainable water resource utilization. However, their service performance has become increasingly characterized by uncertainty and variability due to climate change and aging infrastructure, making reliable evaluation essential in hydraulic engineering risk management. Despite this need, most existing evaluation models are limited in addressing the compound effects of data randomness and conceptual fuzziness, which reduces the reliability of their evaluation outcomes. To bridge this gap, this study develops a comprehensive evaluation approach for earth-rock dam performance using a variable weight extension cloud model. First, a representative indicator system was derived based on relevant technical standards and guidelines, together with statistical analysis of numerous dam failure cases. Then, game theory was employed to integrate subjective expert judgment with objective data information, forming a balanced weighting scheme. Finally, the extension matter-element model was coupled with the normal cloud model to effectively capture these compound uncertainties. The effectiveness of the proposed model is demonstrated through a case study of a gravel soil core rockfill dam in southwest China. Compared with existing models, the proposed approach produces more dispersed and informative grade memberships, particularly in adjacent grades, and the results provide clearer quantitative evidence of its superior capability in representing uncertainty.</p>