Overtakes and dwell time delays for Japanese commuter trains

Reducing train delays important in many countries, even in those like Japan, where punctuality is already high. There is a clear pattern across the literature that the delays typically occur at stations and are recovered on line sections. Previous work has shown that one explanation for this is that trains interact at stations. When trains have different speeds or stopping patterns, overtakes are important even on double-track lines. The latter is often the case in Japanese railways, and we can better understand their railway operations and delays by explicitly studying the way trains overtake each other. This paper uses historical train traffic records from three Japanese railway companies, in total 88 million observations, and finds both that most of the overtakes occur at a small subset of stations, and that only about seven percent of overtakes were executed as scheduled. We also found that the combined dwell time delays decreased in these rare, successful cases but increased in the other scenarios, with a high degree of statistical significance. Looking at the interactions and the resulting dwell time delays, it is also possible to show and evaluate the actions of dispatchers. We found that they often reduced the delays somewhat by shifting the location of overtakes between trains that were either early or delayed. Finally, we suggest that interactions like overtakes can be used to help calibrate and validate simulation models, as they provide another meaningful and quantifiable way to describe the performance of railways, much like delay distributions and punctuality.