Amtrak Lateness Statistics

[Green Maned Lion]

I've been working on a small project over the last two weeks to track Amtrak's OTP. I have seen several different people quoting Amtrak’s “percent of trains on time”, which vary greatly. The primary reason seems to be the definition of “late”. People who don’t like Amtrak and want grounds to criticize it or get rid of it seem to define “late” as any time more than one microsecond after scheduled time. Others seem to define it by being there within a certain amount of time after scheduled, while still others seem to define it as being more than a certain percentage of its schedule late.

I personally have allowed 30 minutes for a short distance, and 2 hours per day scheduled for acceptable on time performance on the trains I myself have ridden on in the past. I based this on the general on time performance of Silver Meteor, Silver Star, and Silver Palm during the times I rode those trains as a youth to Disney with my plane-phobic father. The worst lateness I experienced was on some Metropark NJ to Chicago train (Either the Cardinal or Broadway Limited- this was 91 or 92-, not sure which) at 16 hours. We almost missed my Uncles wedding. If routing is off, I was less than 10 years old at the time. My memory might be off.

In response to what I saw as biased determinations, I came up with my own system of observing this without bias. Basically, I enter the number of minutes late or early in a iWork Numbers spreadsheet, and calculate a bunch of averages and percentages. Whomever reads the data can decide for themselves whether the train was late or not. To call the 70+ hour Texas Eagle late for being 5 minutes behind schedule into Los Angeles (which I guess would be saying the Sunset Limited was late, but whatever) is, to my perspective, farcical, but if you want to do so, be my guest.

I have not selected every train on the Amtrak System, but a selection of them that I saw fit. I basically selected the trains I considered to be “Named Trains”, with three exceptions. The first is Reigonal number 143, which I refer to by its old Northeast Direct name assignment, Nutmeg State, and I put this one in because I am taking this particular train in march as part of a rail vacation. Its my spreadsheet, so I’ll put in what I want. The second is Empire Service number 233, which I keep track of because I took it back in November. And lastly, Cascades numbers 510 and 517, which are the Vancouver to Seattle, once daily trains, which I refer to as Canadian Cascades in the list.

Other than these I have chosen trains that are generally somewhat long distance, operate once a day in each direction at most, and have a unique name. The list of watched trains are: Sunset Limited, Southwest Chief, California Zephyr, Empire Builder, Coast Starlight, Crescent, City Of New Orleans, Texas Eagle, Capitol Limited, Lake Shore Limited, Cardinal, Silver Star, Silver Meteor, Palmetto, Carolinian, Piedmont, Pennsylvanian, Vermonter, Maple Leaf, Adirondack, Nutmeg State, Empire Service 233, Ethan Allen Express, Kansas City Mule, St. Louis Mule, Blue Water, Pere Marquette, Canadian Cascades and Heartland Flyer. This means that when I mention averages, they only include those trains, and when I mention minimums and maximums, they also only include those trains. Also, the Texas Eagle is into San Antonio every day, not into LAX on Sunset Limited connection days.

I’m willing to post my daily statistics each day to this thread, if people are interested in reading them. Several of my statistics are limited by the lack of information available. I have gotten quite a few so-called “service disruptions”, which from what I have read does not necessarily indicate the train not arriving, but most likely indicates significant delay.

I truly like to be thorough, so if members can provide me with even mildly accurate OTPs for service disrupted trains, I would be very grateful. If you are interested in daily updates, I will mention service disruptions for each of them. I record the information the day after, so I input information for Jan 1st today. Due to the complexity of doing so, plus the boring nature, I am not going to give you percentages of every train every day, nor give you the minutes for each train every day.

If you want that precise information, you can look it up on the various sites that host it, or if it is outdated, I will be happy to give it to you if you email me. My start date is December 20th, 2007. I can NOT convert this information into Microsoft Excel data. I don’t do Microsoft Office.

I appreciate any information as to why trains are significantly late or Service Disrupted, and I would enjoy discussion about related things in this thread. I’m trying to figure out why things happen, in addition to turning them into numbers.

For starters, the following trains are Service Disrupted:

12/28 Southwest Chief, E-bound

12/22 California Zephyr W-bound

12/24 California Zephyr W-bound

12/28 California Zephyr E-bound

12/28 Texas Eagle S-bound

12/21 Capitol Limited W-bound

12/22 Capitol Limited E-bound

12/22 Lake Shore Limited W-bound

12/23 Silver Star S-bound

12/31 Silver Star S-bound

12/28 Pennsylvanian S-bound

12/21 Blue Water W-bound

12/21 Pere Marquette E-bound

12/21 Pere Marquette W-bound

12/27 #517 Canadian Cascades S-bound

Times they actually arrived, if they did, would be much appreciated.

The Actual Stats

The two earliest trains by percentage AND time, are both New Orleans trains, but ironically are differently bound:

The City of New Orleans, New Orleans-bound, runs an average of 22.84 minutes early, 1.95%, its best time being 58 minutes early on Christmas, and worst time being an hour and 5 minutes late.

The Crescent, but when New York bound, has an equal percentage early, 1.95%. Obviously, being a longer-schedule, it is considerably earlier in minutes, 35.53. Since I started timing this, it has been more than an hour early twice, on Christmas and on December 30th. Its most delayed time was 30 minutes late- the only time it was late since I started measuring.

The latest train in percentage is the St. Louis Mule, which runs an average of an 1h 38m late on its 5h 40m run- 28.94%. It consistently runs between 45 minutes and an 2h 30m late, with its worst performance being on December 21st, 3h 15m late on its 5h 40m schedule- 57.35%.

The latest train in percentage for a day is on Amtrak’s Michigan service, specifically the Pere Marquette. On December 23rd, it ran 4 hours and 15 minutes late on its 3 hour, 55 minute schedule- 108.51% late!

On the flip side of the coin, the best percentage for a day was achieved by the North-bound Heartland Flyer, running 23 minutes early on its 4h 14m schedule, 9.08%. The earliest time in minutes was the south bound Texas Eagle on Christmas, an incredible hour and 10 minutes early- 3.5%.

The latest train was the December 30th Sunset Limited, which ran 9 hours 58 minutes late on its 46h 15m schedule, 21.54%.

A not-very accurate daily percentage average for these trains is 3.24%- its not accurate because it is difficult to measure these since some trains do not run every day. The average number of minutes late for the those trains is 39.23 over those days. The total number of minutes late or early for these trains over that time period is 24,067.

For each of the days between December 20th and January 1st, the group watched ran:

12/20 37.86

12/21 45.48

12/22 39.23

12/23 73.04

12/24 32.62

12/25 -7.76

12/26 19.57

12/27 44.52

12/28 41.15

12/29 33.45

12/30 63.42

12/31 38.34

1/1 19.24

Minutes late.

Any suggestions, comments, perspectives- would be welcome.

 

[MrFSS]

Have you seen this LINK?

 

[PRR 60]

The Amtrak criteria is a 10 minute tolerance (10 is on-time, 11 is late) for short-distance trains and a 30 minutes tolerance for long distance trains. For this purpose Amtrak defines long distance as trains with end-point distances of greater than 300 miles. The only exception to that are the Boston - Washington Acela Express trips which are held to the short-distance 10 minute tolerance. Note that Amtrak simply ignores trains that are cancelled or do not complete the trip.

The tolerance used by the Department of Transportation for airline reporting is 14 minutes (flights 15 or more minutes late are counted as late). For airline reporting, flights that do not operate or are diverted count as "late" in the statistics.

 

[Green Maned Lion]

I saw it just today. It would be nice to add a backlog of statistics, I guess, but I wouldn't be able to get anymore consistently, which would add to the inaccuracy of my averages.

 

[meatpuff]

Wow, if your stats showing record early-ness on Christmas Day don't show the neagtive effect of rail congestion on Amtrak, I don't know what does. You listed three trains that were about an hour early on Dec 25, but I suspect there were many more all across the system.

Christmas Day being of course a day when the freight railroads take a break almost nationwide, but Amtrak keeps running as usual. This is a good test case that really shows either how freight railroads don't really give Amtrak trains preference or how strained the rail infrastructure really is under normal conditions, depending on your perspective.

And of course those trains being an hour early is just the time made up (and time padding not used) between the penultimate station and the end of the line. They may have been 5-10 minutes early on average into each station along the line and had to sit until the scheduled departure.

-meatpuff

 

[Green Maned Lion]

I agree. You're right; the trains ran very early on Christmas, but not that early. Only 3 trains were an hour early. But the earliness was substantial.

 

[Trogdor]

The only exception to that are the Boston - Washington Acela Express trips which are held to the short-distance 10 minute tolerance.

Are those trips counted as one train or two? I think Amtrak might count the WAS-NYP and NYP-BOS segments as two different trips for on-time reporting purposes.

Several years ago, Amtrak counted the Sunset Limited as two different trains (Sunset East and Sunset West, splitting at New Orleans) for OTP tracking. I don't recall when they stopped doing that, but it was at least a couple of years before Katrina.

 

[PRR 60]

The only exception to that are the Boston - Washington Acela Express trips which are held to the short-distance 10 minute tolerance.

Are those trips counted as one train or two? I think Amtrak might count the WAS-NYP and NYP-BOS segments as two different trips for on-time reporting purposes.

Several years ago, Amtrak counted the Sunset Limited as two different trains (Sunset East and Sunset West, splitting at New Orleans) for OTP tracking. I don't recall when they stopped doing that, but it was at least a couple of years before Katrina.

Amtrak counts them as one train. I agree with you. They should be counted as two since, for all practical purposes, they function as two trains. At least Amtrak changed the reporting criteria to the 10 minute toloerance. Up until about two years ago, the Boston - Washington trains were only held to the 30 minute LD criteria.

 

[PRR 60]

I don't normally post these stats here, but since this topic brought up the subject, here are the latest on-time results for Acela in the NEC. I keep this data myself and do count the BOS-NYP and NYP-WAS city pairs as seperate trains. I also apply the DOT airline citieria so I can make an valid comparison with the airline on time results when those are released (usually about 6 weeks after the end of the month).

Acela On-time for Rate for Dec-07

Boston to New York 80.4%

Average trip time (hr:min): 3:36 : average trip (min:sec) 05:31 late

Trips scheduled 194 ; trips 15 or more minutes late 32 ; trips cancelled 6

New York to Boston 81.4%

Average trip time (hr:min): 3:44 : average trip (min:sec) 08:12 late

Trips scheduled 194 ; trips 15 or more minutes late 32 ; trips cancelled 4

New York to Washington 84.4%

Average trip time (hr:min): 2:51 : average trip (min:sec) 05:05 late

Trips scheduled 347 ; trips 15 or more minutes late 42 ; trips cancelled 12

Washington to New York 91.4%

Average trip time (hr:min): 2:49 : average trip (min:sec) 04:21 late

Trips scheduled 347 ; trips 15 or more minutes late 25 ; trips cancelled 5

All trips 85.4%

Average trip (min:sec) 05:29 late

Trips scheduled 1082 ; trips 15 or more minutes late 131 ; trips cancelled 27

The fine print: The NYC/WAS and NYC/BOS city pairs are each counted separately. Through trains between Washington and Boston are counted as two trains. On-time is arrival early, on-schedule, or up to 14 minutes behind schedule (DOT Airline Reporting Criteria). Cancelled trains include all trains that either did not run or did not complete the trip. Data is from the Amtrak.com train status feature.

 

[Green Maned Lion]

I keep thinking of adding a sheet to the spreadsheet and doing NEC trains. I already sort of have it formatted. I'm trying to figure out a way to organize the morass more easily. I was thinking about doing it by cross referencing the numbers with the old names assigned to them, like I did for the Nutmeg State/Regional 143, and also dividing them by number sections. I was surprised that only five Metroliner numbers were still being used (110, 111, 121, 125, 127). I personally hate Amtrak's decision to retire all the names and simply call them Reigonal. I think, for example, "South bound Charter Oak" sounds better than "Reigonal 86". But whatever.

Also, I keep telling my girlfriend that one advantage of Amtrak over the various airlines is that they may run (substantially) late at times, but they always get you there. For some reason the various airlines seem to think it is somehow acceptable to cancel a flight, hand out a refund, pat you on the head, and say "you're on your own, we are going to say we are really sorry, but we really don't care." Whereas Amtrak usually makes a nice-sized effort to get its displaced passengers where they are supposed to be, perhaps not on time, but there nonetheless. I usually feel that, assuming I don't miss the train, when I buy a train ticket I have bought a certificate giving me travel from point A to point B, preferably by the trains I specified, but if not, via some other way. On the other hand, with a plane ticket, I feel I have bought myself an option to travel on a given flight for a given price, if that flight is operating. (Probably preaching to the choir, end of rant.)

In anycase, does anyone know if 2.5% cancelation (27/1082) compares favourably to the airlines?

As for yesterday's stats:

First the generals. Yesterday was a relatively late day, running an average of 59.84 minutes late for the trains I watch, versus 39.56 late for the whole time I've gathered once the new numbers are factored in. Several trains posted new record times. Additionally, the Pere Marquette had a service disruption for its West bound train.

The latest train yesterday, and the latest I have so far observed, was the east-bound Cardinal, which pulled into New York Pennsylvania Station 9 hours, 59 minutes late- 37.1% of its 26h 56m schedule, the largest percentage late for an overnight I have so far recorded.

The earliest train was the Southwest Chief, pulling into Los Angles Union Passenger Terminal 47 minutes early, 1.8% of its schedule. This is also the earliest it has been since I have kept track of it.

In addtition, several other trains posted records for my list. The Coast Starlight into LAX was the earliest it has ever been, 17 minutes early. That train shall not be referred to as the Coast Starlate today.

The Northbound Silver Star set one- for lateness. It ran 3 h 10 minutes late on its 30h 53m schedule- 10.2%. Admittedly, this was not much later than its performance on 12/30, where it posted 3h 6m late.

The Northbound Carolinian also set a record for lateness, at 2 h 1m late- 15.16%. This, however, was considerably more than its next-latest performance of an 1h 10m. The Southbound Piedmont also set a lateness record of an 1h 6m which is more than double its previous lateness record, 32m. Percentage-wise, the Piedmont ran 34.92% late, perhaps the latest train percentage-wise.

The usually-early NYP-bound Pennsylvanian also recorded its latest-so-far at a still-good-for-a-9h-30m-train of 23m- a mere 4% late.

The Toronto-bound Maple Leaf did its poorest at 1h 45m late- 14.03%. Ironically, the other New York to Canada train, the Montreal-bound Adirondack, coming right on the heels of a 5h 16m debacle, also set a record- its best time so far, a mere 2 minutes late.

The Rutland-bound Ethan Allen Express also set a record-lateness time- 1h 55m late. That would come to 35.38% late.

The St. Louis Mule, for the first time since I've been watching, broke the double-digit barrier- yesterday it ran only 9 minutes late. On the other hand, the Ft. Worth bound Heartland Flyer didn't fly yesterday, posting a lateness of 1h 18m, its worst performance so far- 30.7% of its schedule.

 

[PRR 60]

Typically, the NE Shuttle airline cancellation rate is higher than Acela. It varies depending largely on weather. In October, the last month with airline data, the rate varied from about 4% to 8% depending on route and carrier.

I do not know what you mean by airline cancellations leaving you on your own. If a flight is cancelled you do not simply get a refund and then get sent home. You will be accommodated on another flight just as Amtrak will accommodate you on another train. I have had it happen, and I have always been treated well. I have not found either Amtrak or airlines to be consistently better in rebooking cancellations. When you dump a bunch of displaced passengers into a crowded system on short notice, it can create problems. Cancelled trains or planes are a hassle for both the passenger and the carriers regardless of mode.

The reason for eliminating traditional train names is simple. They are an anachronism and mean nothing in 2007. Except for rail fans, all today's corridor travellers care about is when does the train leave and when does it get in. The service name - Acela, Regional, etc. - tells them what to expect on board. In the old days, the names did have a meaning. The Afternoon Congressional was a top-class train that could not be confused with the Senator. But a seven car Amfleet train named Congressional would be no different than a seven car Amfleet train named Senator, so I fail to see the point of retaining names just for the sake of it. Names for the LD's it makes some sense, for corridor services, it does not.

 

[Green Maned Lion]

My stats for today are done, but I've noticed there doesn't seem to be all that much interest in this thread. It does take some time out of my day. I don't mind doing it, but I don't want to do it just to post my findings on the net if nobody is interested in looking at them. Naturally, there are other reasons why I seem to get no replys, such as people finding it interesting but not having anything to say. In that case I am still happy to keep it up. But I don't want to do it just to see my own text on a web page. So if you guys want me to continue, just drop a mention of such.

 

[Anthony]

The Amtrak criteria is a 10 minute tolerance (10 is on-time, 11 is late) for short-distance trains and a 30 minutes tolerance for long distance trains. For this purpose Amtrak defines long distance as trains with end-point distances of greater than 300 miles. The only exception to that are the Boston - Washington Acela Express trips which are held to the short-distance 10 minute tolerance. Note that Amtrak simply ignores trains that are cancelled or do not complete the trip.

OTP tolerance (for all trains besides Acela and the defunct Metroliner, which as you said, is a flat 10 minutes) varies based on the total mileage of the trip. Short-distance trains don't necessarily always have a 10-minute tolerance... there are 5, 10, 15, 20, 25, and 30 minute tolerances based on the total trip mileage, which may differ from that on the timetable if a train is annulled short of the scheduled endpoint.

 

[MrEd]

thanks for working on the stats project.

 

[haolerider]

My stats for today are done, but I've noticed there doesn't seem to be all that much interest in this thread. It does take some time out of my day. I don't mind doing it, but I don't want to do it just to post my findings on the net if nobody is interested in looking at them. Naturally, there are other reasons why I seem to get no replys, such as people finding it interesting but not having anything to say. In that case I am still happy to keep it up. But I don't want to do it just to see my own text on a web page. So if you guys want me to continue, just drop a mention of such.

I think your stats are interesting, but since Amtrak already keeps these records and that is what is used throughout the system, all your work seems to be redundant. Amtrak's figures are what are used in discussions with the host railroads and they have agreed to use the same parameters.

 

[billinwilliamston]

Have you seen this LINK?

Thanks for the link.

In one way it's actually kind of scary. We were thinking about sending our daughter to Chicago on a route which is supposed to arrive at 10:38 p.m. (bus from East Lansing connecting to the #355 Wolverine) . 10:38 p.m. was pushing our comfort level with her traveling alone.

I just looked at the last three weeks for this train. With an average delay of 91 minutes (average arrival 12:08 a.m.), a median delay of 84.5 minutes (since averages can sometime be misleading) and a max delay of 288 minutes (arrival 3:26 am!!! ), I'm not so sure about putting her on that route. I think my wife will definitely be against it.

Anyone had much experience with arriving at Union Station after midnight?

Thanks.

 

[caravanman]

Well, yes, I did arrive at 3am one morning, SW Chief a tad late. It seems quite a safe environment to me, I guess it depends what your daughter is doing next? Will she be met? Is she planning to take a cab?

I at least got a free night in a (grotty) hotel !

You know your daughters limitations, just be aware of the possibility of late running..

At the end of the day, (say, 3am!), it is the lack of being able to plan arrivals with any degree of certainty that is the bugbear, no matter who is to blame.

Just for interest, my latest train ever was about 4 years ago, the Sunset limited from LA to Orlando, about 18 hours late!

Ed. B)

 

[Green Maned Lion]

But recently, the California Zephyr ran 31 hours late.

 

[billinwilliamston]

But recently, the California Zephyr ran 31 hours late.

I guess caravanman should feel lucky then, eh?

 

[Green Maned Lion]

Dude, being late is fun!

 

[opaque]

Depends on what plans you have in the place you're going to and how long you intend to be there.

A few hours is fine but if you're talking 10 hours and you're only there for 15 it's no good.

Suppsoe it depends on what time you're expecting to arrive, if you're arriving after say 8pm does it matter as much if you arrive at 4am rather than ecpecting to arrive at 8am and arriving at 4pm?

I also guess it's not much fun spending an extra 10 hours sat on your bum in coach. You could travel the Atlantic on a plane in less time.

 

[Green Maned Lion]

I'm a pragmatic person, as well as a insatiable cynic. I give myself 12 hours or so minimum before I have something important to do. Most people should do this too.

You have something important to do? Cool, leave several days before it. Relax. Stop worrying about the time. Its not worth it.

 

[Kramerica]

I'm a pragmatic person, as well as a insatiable cynic. I give myself 12 hours or so minimum before I have something important to do. Most people should do this too.
You have something important to do? Cool, leave several days before it. Relax. Stop worrying about the time. Its not worth it.

In an ideal world, that would be great. But most people don't have "several days" to arrive early. From what I gather in your other posts, you don't have a job. So it is relatively easy to not worry about time. Most people have to be at work and have a limited amount of time off. And that time off is used for a multitude of things, including leisure, family time, home maintenance, etc. And most people want to make the most of it. And for business travellers, most companies won't send you out days in advance of a big meeting. That's just not going to happen in today's business environment.

That's not to say that people shouldn't plan things with a bit more time padding, and relax a little bit more. But several days? C'mon.

 

[Green Maned Lion]

To me, the most important thing about time off is to relax and not worry. If I have to worry about anything at all, I feel like I am not on vacation.

The world as a whole would work a LOT better if everybody, companies included, would slow the heck down. Everybody wants everything, they want them yesterday, and there is something very wrong if its there some time after yesterday.

A lot of companies are doing poorly because of this attitude. Restructuring Chrysler to adapt to a changing world, for instance, was definitely possible. A year before Daimler sold them, they made a profit. They didn't even sell it because they themselves wanted to. All of a sudden it was unprofitable, and investors wanted them to sell it. Stock prices, not intelligent business planning, made them sell it. Fixing problems take time. General Motors could have been saved 15 years ago. It would have taken time, someone who was looking for a long term solution. Now its not possible, for one good reason: they don't have the money to hold out until they can be profitable again.

The whole world needs to slow down, because everybody is rushing and making stupid decisions in the process.

 

[haolerider]

I'm a pragmatic person, as well as a insatiable cynic. I give myself 12 hours or so minimum before I have something important to do. Most people should do this too.
You have something important to do? Cool, leave several days before it. Relax. Stop worrying about the time. Its not worth it.

Nice work if you can get it, but as noted above, the bulk of America does not have that option. This is why Amtrak is simply not a business decision out side of the NEC and other short haul corridors.