Based on hindcast data of the National Climate Centre second generation monthly Dynamic Extended Range Forecast operational system (DERF2.0) from 1983 to 2015, prediction performance of the system for western Pacific subtropical high (WPSH) is evaluated by using time correlation coefficient, standardized root mean square error, anomaly sign consistency rate and Taylor diagrams. The results show that the model can predict spatial distribution of 588 dagpm contour lines and zonal wind shear position, but system bias is large and strong. The prediction performance for interannual variability of the 500 hPa height field is better than for zonal wind field. Climatological annual cycle information of WPSH are well represented, but there are obvious systematic deviations. The prediction is larger for WPSH area, stronger for intensity, by north for ridge line, and by west for west ridge point than the observation. The prediction performance for interannual variability of WPSH is good. The best is for intensity, secondly for area, and the poorest for ridge line. The prediction performance depends on the lead time. With time approaching, the prediction performance shows an increasing trend. Updating forecast is needed in time for rolling correction in operation. The performance of model WPSH prediction for area ahead of 0-20 days, for intensity ahead of 0-2 days, for western ridge point ahead of 0-5 days, for ridge line ahead of 0-7 days is better than persistence forecast. This can be used as a reference in forecasting operation.