Lane change in dense traffic is considered a challenging problem that typically requires the recognization of an opportune and appropriate time for maneuvers. In this work, we propose a chance-aware lane-change strategy with high-level model predictive control (MPC) through curriculum reinforcement learning (CRL). The embodied high-level MPC in our proposed framework is parameterized with augmented decision variables, where full-state references and regulatory factors concerning their importance are introduced. In this sense, improved adaptiveness to dense and dynamic environments with high complexity is exhibited. Furthermore, to improve the convergence speed and ensure a high-quality policy, effective curriculum design is integrated into the reinforcement learning (RL) framework with policy transfer and enhancement. With comprehensive experiments towards the chance-aware lane-change scenario, accelerated convergence speed and improved reward performance are demonstrated through comparisons with representative baseline methods. It is noteworthy that, given a narrow chance in the dense and dynamic traffic flow, the proposed approach generates high-quality lane-change maneuvers such that the vehicle merges into the traffic flow with a high success rate.