Kalman filtering (KF) has good potential in fast rotation of state of polarization (RSOP) tracking. Different measurement equations cause the diverse RSOP tracking performances. We compare the conventional KF (CKF) and the modified KF (MKF), which have different measurement equations. Semi-theoretical analysis indicates the lower conditional variances of measurement residuals and process noise of MKF. Compared with CKF, the MKF has > 3 dB optical signal-to-noise ratio (OSNR) improvement at the 10 MHz scrambling rate in simulation. For MKF, more significant tracking speed improvement exists for lower OSNR. MKF can be smoothly combined with an adaptive algorithm, which outperforms adaptive CKF throughout the simulations.