From the perspective of passengers, a railway timetable can be called better than another if its expected passenger time is lower in practice. So, the authors constructed an analytical function that evaluates a timetable on this criterion: total expected passenger time in practice. Other methods to evaluate timetables invariably describe different performance indicators: realisability, conflict- freeness, stability, efficiency, robustness, resilience, but mostly do not indicate how to score and weigh these different performance indicators. This means that when comparing two timetables, deciding which one is preferable remains hard. The authors' objective of expected passenger time in practice resolves these issues. Also, compared to a simulation approach, the authors' analytical stochastic approach has a major practical advantage. It decouples all actions (ride, dwell, transfer, knock-on) in the timetable and in doing so, can evaluate the expected time in every action separately and simply add all expected times of composing actions afterwards. So the exponential amount of combinations of primary delays over all actions that standard simulation packages explicitly iterate over is dealt with implicitly and much more efficiently. This means that the evaluation made by the authors' method requires less time. The authors' method is applied to two timetables of all passenger trains in Belgium. Both timetables were manually planned and then put into operation in practice. With the authors' method, one can conclude that one timetable has considerably lower total expected passenger time in practice than the other one. The authors also show that this is caused mainly by better passenger transfer planning, but also partly by a changed line planning. Comparison of the reported results for both timetables also suggests that advantages of each could maybe be combined.