The traditional rock burst estimation method is usually based on the σc (rock strength) in practice, while the actual occurrence of rock burst depends more on the structure and strength of the rock mass. In this work, the actual rock bursts occurred in a railway tunnel project in Western China, and the σcm (rock mass strength) was calculated by the generalized Hoek–Brown criterion. According to the actual situation of rock bursts, a modified rock burst estimation criterion using the ratio of σcm to σmax (maximum geostress) was proposed. The influence of randomness on the reliability of rock burst estimation criterion was considered. The estimation results based on the traditional and modified method were furthermore compared with those of the actual rock bursts. The results show that σcm calculated by the generalized Hoek–Brown criterion may be considered well in the rock type and strength, construction condition, and structure features of the rock mass; the estimation results of rock burst using the ratio of σc to σmax are quite different from the actual situation, while those using the ratio of the σcm to σmax coincided relatively with the actual rock bursts; the ratios of σcm to σmax, which are greater than 0.167, 0.066 to 0.167, 0.012 to 0.066, and less than 0.012, are corresponded to the slight, medium, strong, and violent grades of the rock bursts, respectively; the randomness of data selection has certain influence on the rock burst estimation criterion, but the variation range is small; the modified estimation criterion of rock burst proposed in this work has a good reliability. The results presented herein are important for tunnel construction and the prevention of rock burst in the high geostress areas.