Something I notice on RTT, about its predictions for delayed trains (in practice this only really has relevance for freight).
Say the schedule has two timing points, A and B, and Z somewhere further down the line, and the timings look like this
A - 1200
B - 1205
...
Z - 1230
The train passes A at 1145, so appears to be running 15 minutes early, and an estimated passing time of 1150 is shown for B and 1215 is shown for Z.
But at 1200 it's still not passed B. You'd think the system would recognise that the earliest it could now possibly pass B is 1201 so it could predict this, and 1226 for Z. Or even assume in the absence of other information that it is just going to run in its pathed slot, and revert the predicted timings to the scheduled ones.
That's not what happens though - it'll carry on showing 1215 for Z even if it's still not shown up at B by 1205 and even if time ticks on past 1215 without it showing up at B!
What's the reasoning for this - is it that it might just not have registered at all timing points and therefore the fact that it's passed A is considered of greater predictive value than the fact that it hasn't registered at B?