Empire Builder Train, photo by Mike Bjork

Contributor Emeritus Brian Bundridge relays these numbers:


Northbound Into Seattle (Train #14)
June On Time Arrival = 73% (Average Arrival 19 Minutes Early)
July On Time Arrival = 74% (Average Arrival 8 Minutes Early)
August On Time Arrival= 74% (Average Arrival 8 Minutes Early)

Southbound into Los Angeles (Train #11)
June On Time Arrival  = 93% (Average Arrival 19 Minutes Early)
July On Time Arrival  = 87% (Average Arrival 3 Minutes Early)
August On Time Arrival  = 84% (Average Arrival 3 Minutes Late)


Westbound into Seattle (Train #8)
June On Time Arrival = 63% (Average Arrival 11 Minutes Late)
July On Time Arrival = 71% (Average Arrival 2 Minutes Late)
August On Time Arrival = 48% (Average Arrival 25 Minutes Late)

Eastbound into Chicago (Train #7)
June On Time Arrival = 13% (Average Arrival 54 Minutes Late)
July On Time Arrival = 42% (Average Arrival 69 Minutes Late)
August On Time Arrival = 23% (Average Arrival 83 Minutes Late)

25 Replies to “Amtrak Reliability Numbers”

  1. That’s interesting. What is it about going Eastbound that makes the trains later? I would assume they use the same track eastbound and westbound. Wind? :)

    1. Near Chicago, the Empire Builder shares tracks with Metra, the area’s commuter rail system and arrives close to the PM rush hour. If it gets there a little bit late, it might get stuck behind a commuter train that makes a lot of station stops. Thus it will just get later and later.

      1. Exactly what happened to me. I took over two hours to get from Glenview to Union Station. Miserable ending to an otherwise pleasant trip.

      2. Also of note, there are no passing tracks on the two-track segment from Milwaukee to Chicago Union — and IIRC some sections are even single-track? With Metra running both directions in the PM rush hour, there is simply no opportunity for the Empire Builder to pass a commuter train.

        My question is why the northbound Coast Starlight is so much less reliable than the southbound. The southbound has to integrate into Metrolink traffic in LA in the PM rush, but apparently there are enough tracks to keep it on schedule mostly.

        Is the northbound hitting interference with Cascades, freight, or Sounder, or what?

  2. If you’re computing averages, could you compute the standard deviation as well? If anything, that’s a better measure of “reliability” since a train that is 3 hours early half the time and 3 hours late the other half is, on average, exactly on time, but that’s not what I would call reliable.

    1. I suspect it’s more of a Poisson distribution, rather than a Gaussian, because trains almost certainly won’t be more than an hour early, but plenty will be an hour or more late. So, perhaps the median or mode would actually be a more useful number. Of course, just seeing the distribution itself would be ideal…

      1. Amtrak schedules have a lot of recovery time in them, and they put a disproportionate amount of that at the run into the terminal station. So the westbound Empire Builder could be ontime in Everett, but very early into Seattle. Eastbound, Seattle-Everett is scheduled to take 59 minutes; westbound, Everett-Seattle is scheduled at 107 minutes.

        It works similarly for the Coast Starlight between Tacoma and Seattle.

      2. Although with the Westbound Builder, if the C/S leaves on time, they just bring it in on track #3.

        But you’re right, there are a number of places that there is ‘catch-up’ time built in.

        Or as they like to say “Airline Time”.

      3. The Southwest Chief, westbound, is routinely arriving 2-4 hours early into Los Angeles these days (and usually an hour early into Albuquerque too). They haven’t tightened the schedule up, I guess so they can retain “on time” status even if something fairly serious happens on the route.

      4. Note that the Chief waits at Albuquerque (it can’t leave *ahead* of time). The slowdowns in Kansas and Colorado will probably remove the hour-early arrival in ABQ. Early arrival in LA will probably remain routine, however.

    2. Also remember, the average doesn’t say anything about a particular run. My round trip on the Empire Builder last year got in 30-60 minutes early both ways. The stats show that you have at least a 50% chance of being on time, except the unfortunate eastbound Empire. But even there, it’s only an hour or 1 1/2 hours. What’s an hour in a 48-hour journey? When I fly, I assume I may arrive a day late and make contingency plans (by either going a day early or considering whether the first day is expendable). The train’s the same way.

      1. Interesting- you can also pull similar data for each Cascades train. Overall, Cascades is 69% on time. The best train is # 504 with 93% on time and the worst is # 513 at 25%.

      2. And let’s remember that on a daily train, the most trains you can have in a month are a meager 62. Each run is worth 1.6%! Not much to go by. And a thrice-weekly train such as the Sunset Limited only makes 25-30 runs per calendar month, each run is worth 3.3-4%. I want year-over-year modal data…I think the question, “What are you most likely to experience?” is probably the most relevant.

  3. The main reasons for the Empire Builders tardiness would be track work on the BNSF in North Dakota, Montana and various parts of Washington State.

  4. About 2 decades ago I took the Empire Builder out of Seattle.

    I think it was about 6 hours late to Chicago.

    So, looks like they’ve improved things.

    1. They have.

      Biggest change was driven by a change in the law which allows Amtrak to actually penalize the freight railroads seriously when they cause unnecessary delay. The “passenger train priority” was previously written into law but unenforceable.

      Also, the entire attitude has changed at several of the freight railroads (though not UP); BNSF actually likes dispatching the passenger trains on time.

      Finally, the schedules were lengthened. :-P

      It means most of the remaining delays are due to actual reasonable causes of delay, such as bottlenecks, track work, or engine failures. Some money is finally going into fixing some of the bottlenecks.

  5. There’s another on-line tracker that provides a graph of the last month’s performance: http://amtrakdelays.onlineschedulingsoftware.com/

    If you type in SEA (Seattle), 14 (Coast Starlight) and get 4 week’s history, you see that the Starlight arrived early 21 times and late 5 times. The earliest arrival was 55 minutes ahead of schedule and the latest arrival was 151 minutes late.

Comments are closed.