Train numbers make a difference
Deutsche Bahn's overall punctuality rate says little about your particular connection. The train number and station are what matter.
The current data window compares 1,201 ICE, IC and EC trains with at least 30 observed arrivals each. Their rates range from 0 to 100 percent.
IC 1884 typically arrived 34 minutes late; for IC 2568 it was 0 minutes. These are historical values across every evaluated stop for each train number.
The period runs from January 2026 to July 2026. Each detail page also gives its sample size, typical delay and cancellation rate.
Where the figures come from
The figures come from Deutsche Bahn's own published data. Its Timetables API reports the scheduled time, actual time and cancellations for every stop. The data is licensed under CC BY 4.0.
A stop counts as on time when the train arrives less than six minutes late. The railway uses the same threshold for its own figures, keeping the comparison consistent.
The period starts after the December timetable change. Before then, the same train numbers can represent different services, so a combined average would describe a train that no longer exists. Every figure is based on at least 30 observed stops.
For a connection's overall reliability score, Bahnsparer trains two compact tree models: one for all transfers and one for on-time arrival. The model learns from data since July 2024, is calibrated on a separate summer period and was most recently tested on June and July 2026. Those test months affect neither training nor model selection.
Data source: Deutsche Bahn Timetables API via the piebro/deutsche-bahn-data dataset, CC BY 4.0. Bahnsparer is independent and is not part of Deutsche Bahn AG.
What the figure cannot tell you
The value describes past months, not your specific travel day. It cannot predict engineering work, severe weather or current disruption.
For a journey with transfers, the reliability score combines two questions: how likely is every transfer to work, and how likely is an on-time arrival? The details show each included transfer separately. The result is a probability calibrated from past journeys, not a live forecast.
„Arrival“ adds arrival times from past journeys to that percentage.„usually“ combines the scheduled arrival with the historically typical delay. For a full timetable period, „9 out of 10 by“ also gives the time by which nine out of ten measured arrivals occurred. This is not a live forecast either.
Where possible, the app uses values for the specific long-distance train. Otherwise it shows the train-category average or regional-service values for that station. If data is missing for part of the journey, the gap is named beside the value. The app shows no score only when nothing can be measured.
Where it appears
In the day view, reliability and „Arrival“ sit directly beside fare and duration. Expanding a connection shows the destination, historically typical delay, cancellation rate and sample for each segment.
During the journey: DB time and 9 out of 10
Once your train is running, Bahnsparer labels „Current DB arrival“ DB's current forecast: the expected arrival time, any platform change and the reason given by the railway.
Between 15 minutes and six hours before the scheduled arrival, the app may add„For planning · 9 out of 10 by“ below it. This later time shows when nine out of ten journeys with a similar DB forecast actually arrived. It does not replace DB's time and does not control alerts, ticket restrictions or alternative searches.
If no current DB forecast is available, the label reads „Scheduled arrival“. If the destination stop is cancelled, it reads „Arrival cancelled“.