Abstract: Classical least-square clock model has been applied extensively in the area of navigation and positioning. However, it cannot reach a better result caused by some blunders and clock jumps. Actually, blunders and clock jumps have occurred so much that cannot be ignored in clock error series. General pre-processing methods, as transfer between phase and frequency, plot show, blunder detection and et al, cannot be satisfied with real-time clock error predicting. So, a new clock error method is proposed based on adaptively robust sequential adjustment with classified adaptive factors and open windows. Main ideas as follows: firstly, clock error series is opened windows with a better size; secondly, blunders in every window are processed using robust estimation; thirdly, clock jumps are eliminated by adaptive least-square between different windows. In a word, Observation anomaly and state anomaly can be impaired availably by above-mentioned steps. In addition, because different clock parameters depict different clock characteristic, classified adaptive factors are proposed to reject outlier in the clock data. Analysis results indicate: comparing with adaptively sequential adjustment, new method’s precision improved 78.9% and 60.4% in the fact of estimate and predict satellite clock error. Also, new method’s estimate and predict precision, because of classified adaptive factors, improved about 4.3% and 29.2% comparing with adaptively robust sequential adjustment. Furthermore, new method is usually fit for other clock models as AR model, Grey model and et al.