Formation rock’s anti-drilling property is the most important factor affecting the performance of drill bit, but most bit type selection methods only qualitatively rather than quantitatively analyze the relationship between geologic rock’ s anti-drilling property and the bit type. On the basis of existing bit type selection methods, a bit type selection method by BP neural network based on rock’ s anti-drilling property is proposed, which has been applied in Dagang Oilfield. At first we set up Dagang Oilfield formation rock’ s anti-drilling property evaluation model according to logging data and laboratory experiments, at the same time the rock’ s mechanics are calculated by this model. Along with abundant field bit job data they have been encoded into neural network training material. Then BP neural network model for the formation rock anti-drilling property and its corresponding optimal bit type is set up. Finally learning samples are input into neural network to complete its training. By predicted rock mechanics data, after calculation by the neural network, the model will output the optimal bit type. This bit type selection method works well in Well Gangshen 18-17, compared with the offset well, the average ROP is improved about  25% by the chosen bit inShaYi layer in Well Gangshen 18-17.