The paper presents an extended Kalman filter-based vehicle sideslip angle estimator, designed by using single-track vehicle dynamics model. Within the vehicle dynamics model, the tire forces have been modeled as random walk stochastic sate variables. The accuracy of such estimator methodology has been investigated by running the estimator algorithm off-line on sets of experimental data recorded on the test vehicle. These data sets include those acquired from a high-precision inertial measurement unit and standard vehicle dynamics sensors. Moreover, major sources of the estimation errors have been identified and practical error compensation procedures have been proposed to significantly improve the sideslip angle estimation accuracy in the case of using the standard production vehicle sensors. In particular, the influence of vehicle roll and road bank angles is analyzed and compensated for.