I thought I’d follow up on Martin’s post on Amtrak’s long-distance reliability numbers with some Cascades data that is highly relevant to the conversation.  Commenters expressed frustration with using mean performance as a single indicator of success, and many asked not only for Amtrak’s definition of ‘on-time’, but also to see median, mode, and ‘full-distribution’ data for Amtrak trains.  The problems with ‘mean performance’ have already been kicking around the transit blogosphere lately, and I see little reason to simply reiterate something others have already said well.  But I always find it useful to add data and visuals to a conversation.

By Amtrak’s on-time performance definitions, trains can be 10-30 minutes late (depending on route distance) and be considered ‘on-time’.  With such a wide scope, the concepts of ‘on-time’ and ‘reliability’ begin to drift apart in meaning.  Reliability should mean a consistent travel experience, both qualitatively (comfort and service) and quantitatively (speed and on-time performance).

Check out the chart above.  Using data from the invaluable Amtrak Train Status Archives, I charted the performance of two morning train segments (#510 Seattle to Vancouver BC, and #513 Vancouver BC to Seattle only) for January 1-June 30 2010.

More after the jump…

The chart depicts actual deviance from scheduled arrival time in minutes.  Each blue-green data pair indicate one day. Blue lines denote the arrival of Train 510 into Vancouver, while green lines denote 513’s arrival into Seattle. Gaps indicate days in January and March in which service was shut down due to mudslides.  A negative bar indicates an early arrival.  I notice two things right away:

(1.  EXTREME VARIANCE, which further shows the misleading nature of mean data.  Trains arrived anywhere from 25 minutes early to 4 hours late.  Train 510 frequently runs 15 minutes early (3h 45m), while 513 (largely due to the extra customs check) rarely runs less than the scheduled 4h 25m.  Worst, however, is the volatility of journey times from one day to the next.  On a chart like this, one should ideally see a straight line.

(2.  PERFORMANCE COUPLING.  Notice how delays on #510 frequently cause reciprocal delays on #513, and vice versa.  I chose these two trains for a reason; they run on single-track and must pass each other in a siding just south of Mt. Vernon.  If one is late, the other train will usually be late as well, and the effects ripple throughout the system.  Very rarely on the chart is there a day in which one train arrives early and the other very late.  On single-track rails things are so finely tuned that minor mistakes can translate into major operating issues, and margin for error is always small.

Anyway, the point of this post is not to undertake a major analysis, but rather to (1.  provide a visual that strikingly shows how average performance numbers offer little in the way of relevant customer information, and (2. to let our readers know that the data is indeed available if one is willing to do the stats themselves.

(Note: This post revises and substantially updates a post from my earlier blog.  The original will be integrated into the STB archives soon.)

31 Replies to “Parsing Reliability Numbers Further”

  1. Fascinating. Also, performance can vary from stop to stop- I was on the 506 train last weekend, and it ran about 15 minutes late until it got to Seattle, where it was only five minutes late due to how the schedule was padded. So for any passenger along the way, waiting for it or getting off at intermediate stops, it was late, but for those of us at the end it was pretty much on time.

    Also, this shows that we need to double track the line north of Everett. :)

    1. Absent super-efficient dispatching from BNSF in Ft Worth (!), and given the lack of a Shengen-type agreement between the US and Canada to speed things up, double tracking is much needed Marysville to Sylvana, Stanwood to Mt Vernon, Burlington to Bow, and North of Ferndale to the International Frontier. Double tracking the bridges over some of the rivers (Snoho, Stilly, Skagit, Nooksack) would be lovely, but very expensive. North of Blaine will need tons of work, but that is a topic for another post. If we wish for 3-6 trains daily, we’ll need this level of double tracking at a minimum. Paula?

      1. When it comes to the laundry list of improvements needed — and as usual your list is spot-on — I like to stick to the principle ‘fix what’s wrong before you improve what’s right’. In this corridor, glitzy improvements are worth much less than bringing the worst sections up to current Class IV standards. Simply fixing the worst two(only two!) miles of track (the bridges just north of Everett, and the last remaining mile of jointed rail between Pacific Central and Commercial/Broadway) would shave 20-25 minutes off the trip immediately.

      2. Looking at the WS DOT’s list of current projects and the ones that the stimulus money is going to get spent on, the only Everett area project is to re-align some curves around the city. So the changes mentioned will have to wait until later.

      3. Well, they’re also rearranging stuff around the Everett station (something related to plans for double tracking), I believe. But no, no bridge replacements in there.

      4. WSDOT was selected to receive $590 million in high speed intercity passenger rail Recovery Act funding. The list of proposed projects is here:

        WSDOT is waiting on final approval from FRA to get started on projects between Seattle and Portland that range from track upgrades to new siding to safety upgrades. Reliability is the end game.

  2. Fantastic!

    Since you’ve clearly done the hard work of parsing the data, any chance you could create a histogram binned by deviation from scheduled arrival? Maybe bins of 5-10 minutes, ranging from -30 minutes to +60 minutes, or something like that.

    I’m such a nerd, but I love this kind of data.

    1. “extreme variance” is a strong term when you don’t actually say what the variance is! Could you actually compute the variance, give a histogram or provide the raw data? I fully agree that the previous presentations of this data were inadequate, but a plot showing deviation from the mean would be very educational … it looks like the distribution is about gaussian, so variance should be a meaningful statistic…

  3. I’d love to see year over year data – it looks like there’s been an improvement from winter, but winter could just have worse weather.

    1. Easily doable, it’d just be alot of work without access to the data in a better form. Just doing these 6 months took me 8 hours of personal time…they should really publish aggregated raw data to the public.

      1. Yes, when I submitted an FOIA request they replied that I must pay $38 per ‘research hour’ and 25-cents for each printed page. Naturally I declined. =)

      2. Sounds like a real bargain compared to my little request.
        I asked BNSF what the height clearance was on the Bellingham sub-division, thinking US Railcar bi-level DMU’s could make the trip.
        They asked for a $1000 upfront payment before processing the request.
        I also declined the generous offer.

      3. Jesus. And the fact is, BNSF has to have the clearance information on the Bellingham sub ready and available.

        Try approaching them as a freight customer: call their “dimensional load” department and tell ’em you have a railcar with dimensions such-and-such and can they give you a quote to move it from point A to point B, or is it impossible…. :-)

  4. The 510 departs Seattle at 7:40am and arrives Vancouver at 11:40am


    A four hour trip.

    According to Google Maps, this is 2 hr and 46 minute ride by car (see link below).

    Being 30 minutes late, then, seems quite horrific, since one way to look at it is that it already penalizes you by 1 hour and 15 minutes.


    1. That map shows Vancouver, WA, not Vancouver, BC. Here is the Google Map directions from Pacific Central Station in Vancouver BC to King Street Station in Seattle:


      It lists the driving time as 2 hours and 30 minutes, or 3 hours with traffic. On a weekend, wait times at the border crossing can add a heck of a lot more than 30 minutes to your drive time…I don’t think I’ve crossed the border on a weekday in years, so I’m not sure how bad it is on weekdays.

      Not that the delays on the Cascades are OK (they make me crazy too), but the convenience of avoiding the volatility of the car border crossing in Blaine is a big selling point for that route. Also, the route itself is GORGEOUS, especially right after the border when you go through that stretch along the beach where there are always a lot of bald eagles.

      1. With a clear border, driving is much faster without question, and I’ve driven a rental car to BC a few times this year in addition to my many train trips. In my experience the train is only time-competitive on Friday and Sunday, the days in which you’re most likely to wait 1-2 hours at the border. When the construction on the southbound Peace Arch crossing is finished, driving should again nearly always be faster. But speed isn’t really my reason for taking the train. Getting through my latest book, the ability to have a beer while seeing the Chuckanut Bay, or sleeping…ah, that’s travel. =)

      2. I recently had a convention in Bellingham so I decided to take my family to Vancouver and just drive down each day. Out of 8 trips my average time between Vancouver and Bellingham was 2 hrs. According to Google Seattle to Bellingham is 1 hr 40 minutes which makes it a nearly identical trip to the train time wise (on average). I live north of Seattle and my *fastest time from Lynnwood to Granville is 3 hrs 15 minutes. Amtrak Cascades gets me from Edmonds to Vancouver BC in 3 hrs and 33 minutes and it takes about another 10 to take the False Creek Ferry to Granville. So in my experience at the very best driving saves 30 minutes. At the other end it’s much worse. One day I took 110 minutes to cross the border driving south for a Vancouver to Bellingham trip of 3 hrs!

        Moral of the story is the train will come very close to matching driving on the best day. If they tweaked a few things the train would always beat driving. From Everett the train has to turn around and that takes time. I’ve had the US border officials hold up the train for 30 minutes coming back into the states even though we all went through passport checks before boarding and the train didn’t stop anywhere. Stuff like that.

        I will say however the experience of taking the train is 5x better than driving. If you’re staying downtown Vancouver your car is a liability and will cost you extra money. The trip up on the train is very nice and a car could never compete. The ONLY reason I see for not taking the train is it doesn’t run often enough.

      3. It will be interesting to see IF the second train into Vancouver BC will become permanent. That’s where a midday train would be a great option. At first I thought the second train was just going to siphon riders from the morning train, but that isn’t the case, plus about a third of the riders are from the Portland area. New Riders!

        I’m sure the decision won’t come until the last possible moment (the end of September).

        Oy Vey!

      4. I do not know what is wrong with the Border Control people these days, but they’re becoming a major impediment to travel for no good reason. Happening in NY too — they’re even delaying *domestic* trains.

      5. @ Nathanael – it is called Security Theatre – the DHS/TSA goons have themselves become a clear and present danger to the US.

      6. Great for me (I’m divorced and live alone) and maybe you.

        But what about the nuclear family.

        Once you put four people in a car, versus 4 train tickets, the cost skyrockets.

        If Amtrak wants to compete, the need to sell a Family 4 for 1 ticket.

      7. I agree with you, for a nuclear family that owns a car, Amtrak doesn’t compete on the marginal cost of a trip. Maybe not a 4-for-1, but “Kids-under-18 free with 2 paying adults” sounds more reasonable.

      8. Are cars suddenly free to drive? People think it only costs gas money to drive their cars. If you’re travelling 200 miles ore more it’s actually cheaper to rent a car from Enterprise on the weekend deal than it is to drive your OWN car! Do the math, it’s true. If you calculate the standard cost per mile on a car and gas you save about $30 for two adults and two kids riding the train and you’ll end up paying that for parking once you get to Vancouver. If it’s cheaper to drive it’s only by pennies even with a family. I have the two adults and two kids in my family and I’ve driven and taken the train both several times and I’ve calculated it. With AAA you save another $30 and on occasion there’s steeper discounts. I never get a discount on my cost per mile when I drive…

      9. Actually Grant, what I see happening is the “nothing but roads” people now compare costs by ignoring the personal capital costs (sunk costs – depreciation, insurance) of a roads based system and will perform the comparison based only on the personal maintenance (marginal) costs (gas, regular maintenance items – oil changes, tune ups, and for those of you with interference engines, the requisite Timing Belt change at 65k-100k (depending on model, pay attention Honda owners!)).

        What this brings up a brand new facet of the argument, and I’d be happy to take them on.

        That is, if the ‘buy in’ to mobility via the roads requires me to purchase, insure, and maintain a vehicle, at $.54/mile per person, if each family member wants that same mobility,


        I had to vote for Sound Move, and ST2.

        Don’t tell me I vote by driving when it’s my only option, and how do you justify forcing me to pay $300 year in gas tax with no say on where or how it’s spent?

      10. By the way Grant, the family of 4 argument is actually legit.

        Essentially if you take that $.54/mile and divide it by the number of riders in the car, at around 3-4 people, versus the price of the Amtrak tickes, when they are at the lowest fare, is about the same.

        If you’re trying to book at the last minute and the train is full, it costt more for that family to ride the train than the AAA cost of taking them in the car.

  5. Yeah, whenever I drive to VAC I use an Enterprise car. When I need the flexibility, it’s great and not too expensive. When I have the extra hour, I happily take the train. Both costs pale in comparison to the sunk costs of car ownership.

Comments are closed.