Passenger Trains > Transitdocs Calculation Question

How does transitdocs calculate estiated arrival time?  Especially for late trains.  Is this an Amtrak calculation?  Is there any human input?  I know there is considerable padding, but we also knoow "late trains get later."

#8 (27) is roughly 6 hours late at Cut Bank, MT 700 miles into a 2200 mile trip.  Estimated arrival Chicago is 2:17 late.  Do you really believe it will make up almost 4 hours?  Perhaps they could run a contest to see if this board could estimate better than they do.   Maybe we could find on line betting for arrival time estimates.  Is there an over/under for 2:17 late?

I think they just use the padding at stations.  A stop with 15 minutes of padding will show no time at the station if the train is running late.  

John

It won't make up 4 hours. In fact, it will lose more time since it has a freight motor. 



Edited 1 time(s). Last edit at 02/28/22 18:13 by sethamtrak.

Based on the times I have seen they use the normal length of time it takes to travel from that point to the station. If the train is 10 miles out on 60 mile per hour track the estimate would be "10 minutes". I have not seen the estimates take in to account the actual speed of the train (a train traveling 9 miles an hour or 20 miles per hour will still show "10 minutes" when it is 10 miles out). Consuming the "padding" makes sense for late trains. In general, I have found the estimates to be "opimistic" ... somehow the estimate assumes a train stopped or dragging along at 20 mph is suddenly going to jump up to it's regular track speed. That is optimism.

The estimated times shown in transitdocs are retrieved from Amtraks software, which makes predictions based on pure running times and scheduled dwells. Essentially it assumes no further delays. I've only seen this info overridden by humans a handful of times, and its only when an accurate late departure time can be predicted. For example, if a train is waiting out a MOW work window or waiting for a rested crew.

I rather guessed the transitdocs forecasts echoed Amtrak's forecasts, and Amtrak's forecasts are almost always ludicrously optimistic, even accounting for the padding in the trains' schedules.

Amtrak's API (if you can call it that) contains estimates for each train all the way to its destination, and I assume transitdocs et al use that information verbatim. I've dabbled with the data and was quite surprised at how... lengthy... it is. What I have noticed is that ETAs for the next station often don't alter at all, *except* for some stations. I know from a FOIA question I sent Amtrak years ago that they do have additional timing points between stations, though back then very few of them. For example, Houston to San Antonio had just one intermediate timing point (it's a lengthy run so it makes sense to have a fixed point between the two to get a better handle on its timekeeping). All the non-station timing points had an X prefixed station code, eg XAB between HOU and SAS (not the actual code; I lost them unfortunately).

I've also noticed that ETAs when approaching stations go into cloud cuckoo land if the train is delayed. For example, ETA 09:00 when the current time is 09:15. Same for departures when the train is experiencing a lengthy dwell. The fact that the train cannot mathematically arrive at the predicted time (ignoring the 09:00/09:15 example for the moment) does not seem to factor much in the calculations.

I did wonder whether I could make something that took historical running times, discarded the anomolies, and gave a better prediction of ETA. Like somebody said earlier, if the train is 10 miles away from its next station stop, and the track speed is max 60mph, then logically the best it can achieve it in is 10-11 minutes. Even if we don't know the track speed, we can take the average speed from the historical running times and use that instead. Again, it's not perfect by any means. Manually we could also apply a maximum speed to a train, such as that Southwest Chief the other day that was running at 70(?) maximum - we all knew it was only going to get later but Amtrak's API was woefully optimistic, clearly not taking the lower speed into account.

But it's a lot of maths, the feed only updates every 4-5 minutes (meaning we could be that amount of time out of date before we start), and though fun I don't have the time to figure it all out.

Amtrak program does not take into consideration of  lower operating speed when a fright unit is being used.  Hurts real bad on # 3-4 where it has sections of 90 mph operation and the freight unit is only goot for 70 mph or less.
Nor does it take into consideration with major delays where crew turns have to get their hours of rest.  This has hurt the #3-4 very bad the last few days, waiting for hours at some stations for a rested crew.

Seems the running time could be adjusted when a freight unit being used.

Years back when I was operating group tours, I would ask what the delay was.  If freight  unitI knew we would be iin no hurry to the station.  In the days of many more staffed stations, one could get a better ETA from the local agents.  Years back when I had the phone number for DRG, dispatcher said, the add unitl is just a "little guy from the yards".

Remember