Against the backdrop of continuously evolving policies and highly volatile spot electricity prices, electricity retailers face dual challenges of revenue uncertainty and risk control, necessitating decision-making models that balance profit and risk. This paper develops a bi-level leader-follower game optimization framework based on conditional value-at-risk (CVaR) constraints. In the upper level, the retailer formulates long-term contract procurement plans and tiered retail pricing strategies, while in the lower level, users respond based on a cost-minimization principle, achieving coordinated optimization of profit and risk in electricity trading. The model generates multi-scenario data based on historical electricity price distributions. After linearizing the CVaR constraints, an alternating iterative best-response approach is adopted to solve for the Stackelberg game equilibrium. The proposed method is further validated through comparison with traditional particle swarm optimization and CVaR-constrained PSO methods. Results show that, under the 80% green electricity consumption constraint on the user side, the proposed method improves average profit by approximately 7%~9%, while reducing the CVaR metric by up to 96% at a confidence level of α=0.95. This significantly enhances revenue robustness and risk adaptability, providing an effective reference for electricity retailers in optimizing strategies under highly volatile electricity markets.