The location, track, landfall point and intensity forecast error of tropical cyclones (TCs) over Western North Pacific in 2014 are evaluated on the basis of CMA/STI’s “Besttrack” dataset. The results show that the performance of TC positioning is a little worse than in 2013, with an average error by all methods being 25.3 km. The average track forecast errors of CMA subjective method are 84.3 km (24 h), 145.6 km (48 h) and 205.4 km (72 h), 280.2 km (96 h) and 415.3 km (120 h), respectively. A huge reduction on track forecast was made at large lead time compared to 2013. Meanwhile, global model’s average track forecast errors are 88.1 km (24 h), 159.6 km (48 h), 253.9 km (72 h), 393.6 km (96 h) and 572.1 km (120 h), and the average errors for regional models are 97.4 km (24 h), 188.2 km (48 h), and 302.7 km (72 h). The overall performance of statistical forecast method is still better than numerical prediction method in intensity forecast, while in the numerical prediction methods, the performance of regional models is slightly better than that of global models.