Train Prediction Data Source Improvements

Updated on 09/27/2017

Starting October 8, 2017, you will notice various changes on the PIDS at the stations, predictions displayed on the public website, and eventually, the predictions retrieved through the WMATA API.

None of these changes will affect your existing applications, as our response formats do not change, but are worth mentioning in case you wish to enhance your existing applications to take advantage of the new responses in the near future.

  1. Unknown prediction ETAs become DLY (signs + website, eventually to the API)

    When the prediction system isn't confident enough to display a numeric value (or ARR/BRD), it will return DLY for the train's ETA. Despite the connotation, it does not necessarily mean the train is delayed, just that the prediction system can't predict the arrival time confidently due to held trains, trains waiting to start trips, or during single tracking.

    For the time being, the API will continue to return empty data for the ETA (as it does today) in place of DLY to insulate existing apps. We encourage you to modify your applications to handle the DLY response, but there is no immediate rush. Note that a future update to the API will incorporate DLY as part of the possible values returned for the ETA, in addition to some enhanced attributes.

  2. Uncertain number of cars

    You may see more instances of "-" for car counts when the system isn't certain of which train will service a particular trip. This value will of course change to its correct numeric value as soon as the system is more confident of which train will be servicing the trip. This differs from the current prediction system which simply reads the scheduled number of cars when it's uncertain.

  3. Terminal station improvements

    Terminal station predictions will be better (yay!). You will still see empty predictions, but should see more predictions for terminal stations than you do now.

    In addition, the prediction system will be able to display different predictions for Group 1 and Group 2 when two trains are waiting to depart from a terminal station. This is different than the current scenario where terminal station predictions only display predictions for the next train leaving.

  4. Better overall predictions

    General improvements to the prediction system mean that there should be:

    • Fewer instances of "--" Trains.
    • Fewer predictions for "ghost trains" (i.e.: predictions based on temporary data anomalies rather than real trains).
    • Fewer blank predictions across the system.
    • Faster removal of BRD predictions after a train has left the station.
    • During single-tracking, more faithful prediction of trains on the platforms they are actually going to service.