Unskilled Driving is one of the reasons for traffic accidents. The accidents could be reduced if drivers' skills are tested accurately. In this research, XuanAi QJ4B high fidelity and immersive driving simulator was used as the experimental platform that included testing roads and controlling software. 10 roads based on real roads in China were prepared to be the test ones that including five types of sections: highway, city, country, Mountain and their mixed one. For each road, there were two driving conditions: driving without traffic, driving with heavy traffic and emergencies that probably caused accidents. The software could record the situation of the vehicle and driver such as the time of finishing task, times of out of road, times of collision and so on. 19 drivers participated in the test, and then evaluated driving skills by each other. Comparing the driving situation with or without traffic and analyzing the roads we found out that the time of finishing task and times of out of road were sensitive to distinguish the drivers. In addition, under the situation with heavy traffic, comparing the behavior of accident drivers and behavior of drivers without accidents, the result showed that city road, country road and mountain road were sensitive to distinguish the drivers.