Visual odometry is extensively utilized in robotics, autonomous navigation, and augmented reality to estimate motion by analyzing sequential images. Provides high precision without relying on external sensors such as GPS. However, real-time visual odometry systems must strike a balance between computational correctness and speed, a critical yet often overlooked trade-off in existing evaluation methods. Traditional metrics such as ATE, RPE, RMSE, FPS, and RTF typically assess precision or efficiency in isolation and lack adaptability to real-time constraints. To address this limitation, we propose the Performance Efficiency Evaluation Ratio (PEER), a novel, adaptive, and lightweight metric that jointly evaluates algorithm performance based on both fidelity and computation time. PEER incorporates a tunable weighting parameter to prioritize performance, speed, or a balanced trade-off, and employs normalization techniques to ensure comparability across different algorithms and systems. We evaluated PEER using the KITTI dataset with various feature extraction algorithms (SIFT, ORB, BRISK, KAZE, AKAZE) and matching algorithms (BF, FLANN). The findings indicate that PEER not only distinguishes performance effectively across various scenarios but also shows strong alignment with established MCDM methods such as TOPSIS and VIKOR, all while maintaining lower computational complexity and greater suitability for real-time deployment. Overall, PEER provides a robust and flexible framework for optimizing visual odometry systems.