Battery aging remains a critical challenge and state-of-heath (SOH) estimation is a key task for battery management system (BMS). Adaptive model-based methods have been extensively studied, and conventional applications of extended Kalman filter (EKF) are mostly limited to only one electrode, in order to address the weak observability of battery system due to the absence of a reference electrode. Moreover, the state-of-art is dominated by improving model precision and algorithm efficiency, while the significance of measurement data for SOH estimation is often neglected. This study applies EKF to both electrodes and guarantees the estimation accuracy through the optimized initialization. Effective cyclable lithium, Δn_{Li, avg}, is proposed and validated with experimental data as a reliable aging parameter to interpret the long-term evolution of battery degradation. Data of different operating conditions (charging, discharging, charge depleting), different state-of-charge (SOC) sections and different rates of measurement update, are tested for their impacts on aging estimation. It can be concluded that a sufficiently wide SOC range from 1C charging is comparatively the optimal condition for estimating Δn_{Li, avg}. Within the bond of required estimation precision, a slower measurement update proves to be a preferred solution due to its equivalent estimation accuracy and less demanding computational resource.