Quick Note: Deterioration of Speed
A regrettable feature of rail transport is that often, the speed of a line deteriorates over time after it opens or finishes a major upgrade. This can come from deferred maintenance or from proper maintenance that includes stricter speed limits or more timetable padding; in either case, it’s because maintaining the original schedule is not seen as a priority, and thus over time service degrades. In some cases, this can also include a deterioration of frequency over time, usually due to inattention.
This is not excusable behavior. The networks where this feature exists, including the US, France, and Germany, are not better-run than the Shinkansen, where I have not seen any such deterioration of Shinkansen speed in many years of poking around timetables on Hyperdia, or the system in Switzerland. Switzerland’s timed transfers make it impossible for gradual deterioration of speed to accumulate – trains are scheduled to just make connections to other trains at major nodes, and so if they slow down too much then they can’t make the transfers and the entire network degrades.
I wish I could say degradation is a purely American phenomenon. It’s very common in the United States, certainly – on the subway in New York the deterioration made citywide news in 2017 (including one piece by me), on the trains between New York and New Haven the schedule is visibly slower now than it was in the late 2000s, on Amtrak the Northeast Corridor has degraded since the 2000s. Speed is not viewed as a priority in the US, and so there are always little excuses that add up, whether they’re flagging, the never ending State of Good Repair program on the New Haven Line under which at no point in the last 20-25 years have all four tracks been in service at the same time, or just inattention to reliability.
But no. France and Germany have had this as well. The TGV used to run between Paris and Marseille in 3:03 every two hours and in 3:06 every other hour; today I see a 3:04 itinerary every four hours and the rest start at 3:11. And here, the Berlin-Hamburg trains were timetabled at 1:30 in the mid-2000s, giving an average speed of 189 km/h, the highest in Germany even though the top speed is only 230 and not 300; the fastest itinerary I can find right now is 1:43, averaging only 165 km/h.
I stress that such deterioration does not have any benefits. It’s an illusory tradeoff. When New York chose to slow down the L trains’ braking rate as part of CBTC installation, this was not seen in reduced systemwide maintenance costs; speed just wasn’t a priority, so the brakes were derated. The 7 train, as I understand it, will instead speed up when CBTC comes online, a decision made under Andy Byford’s program to speed up service.
Nor has France saved anything out of the incremental slowdowns in TGV service. Operating costs are up, not down. The savings from slowdowns are on the illusory to microscopic spectrum, always trumped by increases in cost from other sources, for example the large increases in wages in the 2010s due to the cheminot strikes.
By far the greatest cost of speed is during construction. During operations, faster service means lower crew costs per km. This is where the Swiss maxim of running trains as fast as necessary comes from. This isn’t about derating trains’ acceleration – on the contrary, Switzerland procures high-performance trains. It’s about building the least amount of physical infrastructure required to maintain a desired timetable, and once the infrastructure is built, running that timetable.
Pedestrian Observations 
Isn’t deterioration of frequency usually cause byess than anticipated ridership?
Sometimes speed limits are introduce as a response to noise complain.
Addition of new intermediate stops could also slow down train journey.
Introduction of cheaper and slower vehicle might also cause some systems to be affected
Speak of which, isn’t China being one of the more prominent example of such deterioration? They dropped the operation speed across the entire network, both high speed and legacy, following the Wenzhou incident, and subsequently lowered construction standard of many lines to match the lowered max speed. Some lines’ speed have restored but many still haven’t now that roughly a decade have passed since then.
I was under the impression that China did Japanese style keep-midnight-to-6am-for-maintenance. I know Taiwan has that schedule. Korail seems to do a smaller version last trains on Seoul-Busan route arrive close to 1am and first are at 5am. But I don’t have any knowledge of speed/schedule deterioration in either.
Its interesting this deliberate trade off to give up nighttime trains for speed/reliability is associated with more profit conscious operators like the East Asians and the Swiss. A second order false-economy stemming from a long-term soft-budget constraint (i.e. governments and publics want nighttime service and underrate maintenance)?
SNCF has profit-oriented management and doesn’t run night trains on the LGVs either. This isn’t about reliability, it’s about treating 3:00 as a marketing gimmick that can be compromised, same way Finland explicitly calls its proposal for 1:15 Helsinki-Turku service “one hour train.”
So how would SNCF go about making it right?
Profit-making, profit-orientated and soft-budget issues are not all the same or mutually exclusive. From your criticisms here and elsewhere SNCF seems to have a lot a behavior that looks like soft-budget problems. Not that pure profit-maximization is good idea for railways (US frieght).
Is it even viable to run night trains on LGVs even ignoring maintenance window needs? None of them go far enough to justify such a thing, and low speed, loco hauled night trains are already have bad economics. Buying expensive high speed rolling stock just for night use sounds like a money losing proposition that only makes sense as a political play, which is why afaik no one outside of China does it.
Even using modern EMUs for night trains is questionable, considering that Sunrise Izumo/Seto are more expensive than Shinkansen/airplane for all but the nobi-nobi sleeping slots which are (at least seemingly) denser than the lowest tier Euro sleeper accommodations.
Why not run the same trains 24×7? I doubt sleeper trains make economic sense, but there are a small number of people who work no-standard hours would would find an overnight train more useful than regular day trains. (the arrive at 7am train deserves special mention as it saves a night in the hotel while still letting you be there on time. Sleeping in a seats isn’t comfortable, but you only need a few riders to pay for the incremental costs of the trip. A bit of yield management pricing can get a few people to take the overnight trains.
Of course skipping maintenance on trains or track is not something to ever consider. Before considering any additional service the first question is when/how will you do the required work to keep the trains running.
The Sunrise Izumo/Seto trains are viable as they provide early morning arrivals before business opens (after a convenient Tokyo 10pm departure) at Takamatsu as well as Yonago et al- these areas have relatively few airline flights from Tokyo, and taking the (daytime) shinkansen would burn most of the day. The “up” Sunrise service picks up passengers at Kobe and Osaka just past midnight (0:13 and 0:34 respectively), making it the de facto “last train to Tokyo”, and deposits them in Tokyo at 7:08, ready for a full business day. For comparison the cheap night buses leave earlier and arrive later. This justifies the premium versus the other modes.
@Andrew in Ezo: That means those trains use the legacy network (and are probably the only direct trains between Osaka / Kobe and Tokyo over the legacy line).
@Max Wyss: Yes, that is correct. All the other legacy route night trains (other than the super luxury tour trains) were discontinued years ago (and the daytime trains *decades* back). As I wrote, the Sunrise is a de facto service for Kinki to Kanto rail passengers, as AFAIK it is not officially advertised as an overnight service between the two regions, and the “down” service is not available for the same segment.
“None of them go far enough to justify such a thing”
This wouldn’t be nearly as much a problem with a more integrated European/Eurasian rail system
@Henry Miller When China use their sleeper trains in daytime, according to my understanding, they use each bed as seat for three as passengers are asked to sit on the beds which make the ride less comfortable than sitting on a train with regular seats.
The reason why there are no night trains on those high speed network is because it is not possible to guarantee staff safety when they close off and work on one track while a high speed train passing close to them on the next track. I don’t think that have to do with speed deterioation.
And China do run night train on its Beijing-Shanghai corridor in some days of the weeks, and according to my understanding they do this by re-arranging maintenance timeslot of some tracks to be outside the period that those night train would use and have some other week days for longer maintenance time.
One possible solution is to run night trains 50% on high speed lines and 50% on legacy lines that switches to allow maintenance.
Legacy line between Beijing and Guangzhou is probably too slow for such maneuver to be competitive.
Better yet is reduced service. 1-2 TPH allows a lot of time for maintenance between trains. Sure it means they work in 20 minute chunks with enforced 10 minute breaks every half an hour, but those breaks can be planned around (possibly loading the machines doing the work). The trick is to figure out all the procedures and buy/spec equipment that can do the operation in 20 minutes and be safely moved off the tracks in a minute. This is a difficult engineering problem, but it is one that can be solved if effort/money is put into it. Then the foreman calls the halt, verifies the track is clear and 5 minutes latter the train goes by – or if there is a problem an emergency stop can be called and the train halted (other than regular tests this should be rare)
Nighttime demand will be less than daytime in almost all cases. If you are popular enough to justify more night service than the above you are popular enough to pay for another set of tracks during the day. (probably running a different city pair to avoid connections)
Deploying maintenance machines before repairs and check whether there are anything left on tracks take a lot of time and aren’t something that can be effectively fit in half an hour timeslot. Also, it doesn’t make sense to use the concept of “TPH” for overnight service
@Henry Miller: May I suggest an update course on track maintenance? There are some good videos on Youtube, some by infrastructure managers, such as Network Rail, some by manufacturers, such as Plasser & Theurer or Matisa.
@Max Wyss I will accept your rebuke.
However I’m going to stand by my statement, engineering machines that can do the job in half an hour will take 10 years, but it is an investment worth looking at if you want to run night service. Though I’m not sure it is worth the investment, which in turn means night trains might not be worth it because you never skip maintenance.
Note that there are other options. Building 3 tracks so that you can take one out of service is possible, though obviously expensive (you need track spacing large enough for safety). Or Monday maintenance where everyone knows there is no service so they plan for it. There are other possibilities.
@Henry Miller: It sounds nice, but in order to achieve that, you will have to reinvent physics.
Tamping does simply take its time, because the ballast has to move and find its new place. Ballast replacement, including cleaning and adding new ballast where needed does take its time. Rail grinding does that too, and you can not grind too fast, because you may overheat the railhead and change the internal structure of the steel.
Yes, there is development, but it is evolutionary.
You may only be able to tramp ballast for 10 meters in that time. Or grind a similar length of rail. I’m expecting that because of all the setup/tear down labor costs will be 4x higher than shutting down the whole line for the night. Is that worth it is an economic question, not a possibility question.
If speeds aren’t improved, it’s difficult for rail to compete with regional airlines and nagging car culture.
Are there sometimes speed vs capacity tradeoffs that lead to these lower speeds? I’m thinking about the NWK-NYP stretch of the Northeast Corridor, and how the high-density signalling that allows them to cram 24-26 trains each way during peak hour also reduced running speed from 90 mph to 60 mph.
Also have they ever fixed/can they ever fix the L train braking rates in the CBTC system there? As CBTC expands throughout the NYC Subway using standardized technologies and programming, the conservative, bespoke system there is increasingly becoming an oddball.
I’d think that by the time they’re done, the systems on the L would be up for modernization. At that time it can then be brought up to spec, same for the 7.
Not really. Signal systems that allow you to space trains more than one block apart, like LZB, are such old tech that they’re starting to get replaced with ETCS – and the signal system used for the Northeast Corridor, ACSES, is a variant of ETCS. Not that any of this is even required at the speeds in question; multi-block braking is only required here at speeds above 160 km/h.
There still is. Even if you assume some kind of moving block-type signalling installation that allows you to closely approach maximum theoretical capacity, doubling your speed still means means that while on the one hand the train itself occupies a given section of track for only half as much time, on the other hand the required safety distance between trains quadruples (because of braking distances scaling with v²) and therefore requires twice as much time to pass before the next train can follow.
At low speeds of course the former effect is dominant, but eventually you reach a point where the gains from reducing the track occupancy time of the train itself are outweighed by the increasing braking distance required between trains. This means that given a certain train length, there exists an optimal speed that provides the highest possible amount of trains per hour.
How do you measure speed? If it is brake off to brake off on a single segment, then not really. However if you factor in stop time to get an average speed from source to destination, then the best case trains are stopped for only a minute at each station, while the worst case need 15 minutes to stopped to get people on/off the train. Running a timed transfer system increases time as well because you need to allow people time in the station to get from one train to their next after they are off the trains (I’m not sure if the 15 minutes for worst case captures that or not)
The fastest train unloading means you have 4 large doors on each car with level boarding to the platforms. If you have less small/doors that leaves more space for seats, but it takes longer to get people on/off the train as more people are trying to get through the door. If you have bi-level trains you can have even more seats, but now you need to wait for people to go up/down the stairs. (Note that most bi-level trains are locomotive hauled, which means that in practice they can hold about the same number of people over the entire train – the locomotive holds zero – but they still take a lot longer in the station)
Station transfer time depends on station design, which in turn means if you screw up it is expensive to fix. I don’t really know much about this, but it is important to acknowledge for anyone planning a time transfer system. A well designed station can take a lot of time off of your transfers. Anyplace people need to transfer between trains (or any other system) needs to be designed to make those transfers fast.
If you run at least 6tph on all the lines you can just ignore the transfer time – the worst case someone gets to the platform just as their trains closes the door and needs to wait 10 minutes for the next train – 10 minutes is the maximum I’d consider acceptable in this case, though the less the better of course. In turn this means at 6tph you can run faster average speeds as you don’t need to plan around station transfers. Anything less than 6tph and missing a connection needs to be rare as the wait time for the next train is too long for someone who doesn’t have something else to do while waiting.
Note that the 6tph mark to not do timed transfers is itself a trade off. If you do assigned seats you need to either account for transfer time so that people can get on the train their seat is for. Assigned seats are great when trying to get the maximum number of people on the train (particularly if standing isn’t allowed or is a different price), otherwise you need to do something to ensure the train isn’t over capacity. Up to somewhere between 12 and 20 TPH you can just add more trains when you are in danger of getting trains overfull (there is a couple years lead time to order new trains though so it needs good planning), but eventually you reach the point where the track infrastructure cannot allow more trains and so you need to assign seats to get the maximum capacity out of each train and this puts you pack to slower average speeds because of transfer times (meaning your now very busy stations need to have a place for people who could catch the train before theirs to wait)
I’ve only touched the surface of how complex it can be to run the fastest possible system in the real world. This is something that all operators should strive for though. Time is valuable to people, so the faster you can get people from point A to point B the better.
Bi-levels: mostly loco-hauled (for “commuter” rail) is pretty much an USAn practice. Taking the NJTransit cars as an example, they are very inefficient, because they have only very narrow doors (looks like two single wide doors). Efficient bi-levels have two triple-wide doors on the middle level.
Again if you talk about 10 minute intervals, assigning seats is pretty dumb. At these intervals, it means get on the next train, find a seat, and that’s it.
There is a fundamental law in motion science, which is valid for rail operation as much as for stackers in glass bottle production. To get a fast system, speed up and shorten the slow sections. For rail operation, this means speed up station throats, for example.
That was the point, at 5 to 10 minute intervals you run more trains if there is danger of someone not finding an empty seat. The ecconomics of rail is different from airplanes because you can do this.
At more than 10 minute intervals you are not the spur of the moment trip category as people won’t risk missing their train, so you can assign them a seat. (If someone does miss their train there is time to change seats to the next ) at less than 5 minutes you are running out of space for more trains and so assigned seats allows you to fill the train full as people know where they need to go to find a seat.
People will absolutely choose to get the next train on a whim at 15, 20, even 30 minute intervals.
> Assigned seats are great when trying to get the maximum number of people on the train (particularly if standing isn’t allowed or is a different price)
Why would you ever make standing not allowed? In addition, even if you want to make seats a different price than standing, a solution like Suica Green Car Seats could be implemented, where you just find an open seat, sit down, and pay for it with your transit card or phone.
> so you need to assign seats to get the maximum capacity out of each train
The maximum capacity of a train involves removing seats altogether. To load balance between different cars, you could have signs telling people the fullness of cars in the next trains (I don’t think this happens, but the technology is there, considering apps with per car crowdedness, and dynamic digital signage with per car information, both already exist).
In London they tell you how full each car is on some trains.
Yes, but it’s based on car weight, which is a function of passenger weight. Since passengers vary greatly in weight, this would not be exact enough to tell you if one or two seats are open. You’d still have to walk down the entire train until you found or didn’t find a seat.
This is just to balance loading between carriages though right? So you don’t need to know if *every* seat in a carriage is taken – just if most are (and people will always start to stand before the last seat is full).
I don’t think there is a *need* to balance loading between carriages – the carriages run fine even if all are crowded. Rather, if you even out the number of passengers per carriage, it’s more comfortable for the passengers, more of whom get to sit, fewer of whom are crammed together standing.
The existence of timed transfers makes it hard to have deterioration of speed. However, the creation of those timed transfers can be done by large infrastructure upgrades that speed up the travel time, or by keeping the same infrastructure and slowing down trains in some places to make it fit.
The latter is what happened in some parts of the Netherlands, and that’s why people complain that trains were faster 50 years ago. The pre-takt trains that ran a few times per day without timed transfers indeed were a few minutes faster than the half-hourly trains we have today that have timed transfers at all major stops. But I bet that the average person has a much quicker trip at a more convenient time today.
This also makes it hard to improve speeds, because you have to save at least 15 minutes on a 45 or 60 minute segment, which means relatively expensive projects, either in terms of money or in terms of political cost from skipping stations. Most upgrades are being done to increase capacity, and the additional speed benefits that flyovers, better switches and improved station throats create, end up mostly in improved reliability through increased padding and dwelling.
In Japan, incremental improvements to the Shinkansen from the 1960s to the 1990s allowed the Hikari to maintain the 3-hour schedule while making more stops. Then the Nozomi came, making the old Hikari stops (at this point 2 more stops than the old Hikari, in fact) in about 2:30. So it’s fine to bank such reliability upgrades and then use them to push for a speedup past the next discrete trip time. I wish Germany did that, cutting Berlin-Erfurt to an hour with HSR, then completing Erfurt-Nuremberg to make that 0:45.
I wonder how many upgrades are already banked, considering the iteration speed of European HSR is slower than Japanese HSR. The N700 series running on Tokaido Shinkansen were all upgraded to N700A spec in about 2 years of the N700A being introduced, and the remaining 700 series were upgraded with better acceleration to help keep up around then as well. The ICE 1 from 1991 is still in service today, even though DB started their switch to EMU with the ICE 3 over two decades ago, in 1999.
While introducing new technology can improve a line, having to run alongside older technology means it’s not possible to take advantage of the full potential. Could the retirement of ICE 1 trains improve German HSR? Or does the nature of ICE having to share track with low speed trains much of the time mean that ICE 1 trains weren’t really the limiting factor anyways?
There are no significant differences between the N700 and N700A, which was how the small-a upgrade was so easily done. 700 series trains have 275kW motors on each bogie while N700 trains have 300kW motors.
In fact, just last year JR Tokai banned 700 series from running on the Tokaido in order to do the 12 Nozomi/16tph schedule, so we can surmise they weren’t upgraded. They can iterate fast because of the low lifespan of the trains – they’re built rather cheaply and today there are already reports of N700s going to the knackers.
What might now be harder to overcome is the alignment, and that’s why they’re building the Chuo maglev. You could argue that they could throw more dakka at it and have even faster acceleration rates, but that might be pushing the limits of power electronics and would only benefit stopping services.
JR East, though…
The first N700 retirements began almost a year ago, in tandem with the introduction of the N700 “S” series. And I wouldn’t call the differences between the n700 and N700a insignificant- the “a” version has 10% reduction in braking distance (20% reduction compared to the 700 series), and the ATC has been fitted with a system where after a signal check, even when coasting or on a grade, a constant speed will be maintained, which allows a faster timetable recovery during service disruptions. Both are essential upgrades on an operation like the Tokaido Shinkansen, where everything is operating on precise margins (and subject to the very real risks of natural disasters like earthquakes or typhoons).
The slowdown is almost always due to increased schedule padding, which is increased to ward off complaints about unreliability. According to Mikko Mukula (2008, sadly only in Finnish) the main reason why Japan doesn’t suffer from this slowdown is a different philosophy regarding reliability.
Whereas European systems (excluding the Swiss, to some degree) tend to view padding as unavoidable (because delays are unavoidable), the Japanese approach to scheduling seeks to eliminate delays due to track infrastructure (using reliable technology, fallback systems), rolling stock (keeping it simple, relatively short rolling stock lifespans) and staff (avoiding complex rostering and rolling stock circulation patterns; staff and rolling stock usually sticks to a single line). This focus on eliminating sources of delay allows Japanese operators to operate with very little schedule padding. It’s essentially Lean principles applied to railway operations.
The vertically integrated nature of Japanese and Swiss railways might make this style of lean operations easier to implement, but that doesn’t necessarily mean it’s impossible to achieve in other kinds of regimes. The challenge is in aligning the financial incentives of each party so that you don’t introduce perverse incentives, e.g. making it much cheaper to cancel a train than run late.
That’s an interesting perspective, but won’t the much-vaunted through running services add to the uncertainty? Then again, maybe that’s why through running services are always first in line to get canned at signs of trouble. When the troubles hit, run every train as if it were your own, sort out who should get what back later.
Out in the sticks, perhaps it manifests itself as the many short-hop shuttles down ostensibly continuous main lines, some even run by third-sectors – if you wanted to go far, you should be taking the shinkansen. Any Seishun 18 user, who wouldn’t take the shinkansen anyway, should be familiar with that.
In any case, I’m not convinced myself that padding practically doesn’t exist in Japan – especially after Amagasaki, where JR West’s fetish for punctuality and corresponding reduction in schedule slack was blamed for the driver taking a curve too fast and sending his train into an apartment building.
Japan has chosen a different timetable-infrastructure mix. Timed connections Central European style are not done, but not really needed in the major cities (e.g. ones with subways). Shinkansen are kept separate from conventional network and even each other to maximise frequency. It screws non-megacity mainline cities though which could definitely benefit from s-bahn style service (Sendai, Hiroshima etc). I’d love a detailed comparison with Swiss railways since while Japanese operators have to juggle subway through running and for JR a national freight system, Switzerland has an entire continent to through run with no dedicated HSR lines to relieve mainlines..
I’d say the segmented conventional mainline services are more about about punishing you for not going Shinkansen then padding by stealth. Ishikawa-Toyama-Niigata operators on the Hokuriku mainline and Aomori-Iwate on the Tohoku mainline do operate some through-running services but mostly to create tidy city pairing services. Given the existence of profitable daytime highway buses between the relevant cities that padding better be big or JR West and JR East are squeezing pricing power to stop a cheaper and more frequent rapid services between Niigata and Toyama/Kanazawa.
This is all getting stuck in post queue, but when I get around to doing the Japanese Way of Building Rapid Transit post, I’m going to talk about line identity; tl;dr the Japanese way has independently-operated subway lines, which is also a thing in France (and Boston) but not here or in New York, and unlike France also has this on the Shinkansen.
Subway through running are usually done by other private rail operators instead of JR because of inability for private rails to build lines into inside city center in the old times. In many cases they should have been one single line.
That’s why they invented the technique with the Asakusa line, but places which didn’t have the Yamanote style barrier have done it Fukuoka, Kyoto etc. And there are JR through-running subways, I was on the platform opposite the Joban line just this afternoon. And we have specially made for through-running commuter lines like the Saitama Kosoku and the Toyo rapid. And then there is Seoul which didn’t have extensive pre-car legacy private rail to connect up, but has found it quite useful. And the reality is that subways are usually playing catch-up with urban development so souping up existing mainlines is a good idea.
I’m looking forward to Alon’s comparing through-running to S-bahn a bit more. I have wondered if through-running is better suited to megacities and s-bahn to sub-megacities.
@borners
I think mainline-subway through running and S-Bahn are effectively the same thing. One has two operators cooperating, and one has one operator just doing their thing, but I don’t think JR East vs Subway+Private, or Seoul Subway vs Seoul Subway+Korail S-Bahn style services are substantially different. They have different development histories, but the end product looks pretty similar to the user.
The bigger difference is whether metro lines should be thought of as trams isolated from traffic (most European metros) or mainline heavy rail that (also) serve city center trips (S-Bahn, Tokyo).
Sticking staffing to a single route is best practice for buses – https://www.freewheeling.info/blog/how-to-get-more-people-onto-buses
The Swiss timetables do have padding. But it is more granulated than usual. It also fits well in the “not as fast as possible, just as fast as necessary” dogma.
In the Zürich S-Bahn network, there are stations (usually kind of nodes), about 15 minutes apart, which are called Zeitvergleichsstation. Their departure times are set fixed, and well achievable. The departure times of the intermediate stations are set that they correspond to the earliest possible time. In reality, the train may have 1 or 2 minutes delay at an intermediate station, but will leave the Zeitvergleichsstation on time. This leads to a very short padding (a minute or two, at most), and when the train enters the core, it is very much on time. This is opposed to other places, where the padding occurs in the last segment before the terminal station, or the main station, leading to ridiculously long times in the last segment.
Sometimes, a monopoly eliminates the need for speed:
France passed a law that bans air travel that can be done by train in less than 2.5 hours, even if the airline can provide lower fares. Germany is considering a similar law.
Great article. I can’t speak to the situation Europe but in Japan, where I reside, I think the competition with airline services a key factor in driving rail performance. Domestic air travel is quite convenient with good access from airports to city centers. Further, the lax security protocols when boarding domestic flights (no ID required), mean that domestic flights are not a trial of patience and endurance. Poor flying experiences in the US allow for poor rail performance with the only alternatives being driving or personal jets for the uber wealthy.
Always take “record runs” with a pinch of salt, mind – during the recent attempt to beat the 1980s London-Glasgow record with a Pendolino – the modern train had to obey the speed limits, unlike the APT!
(3 hours 53 minutes – failed to beat the record by 21 seconds)
Yeah, but I’m looking at actual timetables. TGVs routinely did Paris-Marseille in 3:03-06 every hour in the late 2000s, and thy were on time when running on LGVs; today they’re somewhat slower. I’m not looking at record runs or at speeds that are only sustained once a day.
“an average speed of 189 km/h, the highest in Germany”
If this is true, this is disappointing number for the German HSR network. I know that ICE is slow for long trips (>500 km), but I don’t realize that they are also not very fast for short direct trips like Berlin-Hamburg (or even Cologne-Frankfurt, where the average speed for the trip is no faster than 190 kph). Isn’t it faster to drive a Mercedes on the autobahn?
Only at 3 am. The Autobahns are congested as fuck.
But yeah, German HSR average speeds are pretty meh. The routes are always compromised by something, and Cologne-Frankfurt has seen deterioration too, I think (it should be doable Hbf-Hbf in 0:57 for a nice takt, but they struggle to do that Hbf-Flughafen, and I think the fastest Hbf-Hbf now is 1:10?).
Hbf-Hbf was never done under an hour, which is a bummer. They started with 62 minutes, if I am right and now it slowly creeped to 70, as you say.
On highway speeds: Every German claims to know somebody who did the 750 km of Hamburg – München under 4 hours, but this is bs. In practice, average speeds of 130 km/h are the max and this already requires very aggressive accelerations (and decelerations) with absurd fuel consumption.
While you can go 200kmh on the autobahn, in my experience most people drive at about the same speed as US freeways where there is a speed limit. If anything the average speed in the US us faster as those speed limits become a goal to exceed. Of course the outliers on the autobahn going much faster than average are far more common, but most drivers are much slower.
Either way human drivers are not able to safely drive a car that fast without some luck.
Even “normal” US highway speeds diminish your gas mileage a fair bit from the attainable maximum, so crazy autobahn speeds would hurt even more. I suspect many Germans don’t want to take the hit to their wallet since gas is so much more expensive.
That’s a very interesting point, Henry. Here’s an article discussing the data (in German): https://www.zeit.de/mobilitaet/2019-02/autobahnen-geschwindigkeit-tempo-schnelligkeit-raser-verkehr
The upshot is that on autobahn stretches with no speed limit cars indeed drive only 122 km/h on average, that’s 76 mph.
Isn’t increased timetable padding supposed increase on-time performance?
If slight increase in travel time increase helps greatly increase on-time performance then the trade-off is probably be worth it.
But if on-time performance remains the same or decreases, then the trade-off isn’t worth it.
It is a bit of a cheating. On time performance is often used as a measurement for overall performance. The on time performance is usually measured at either the main node, or the terminal station of a run. Depending on the place, “on time” means less than 3, 5, 10 or more minutes late. So, the padding artificially makes believe that the trains are on time, even if they did perform lousy during their run.
Lowering the nominal journey time to make a train consistently slow isn’t helping travellers.
A bit of a trade off. I sometimes have a time I need to arrive. On time performance is thus critical – if I miss the my meeting I may as well not have made the trip. Thus a reputation of on time at the expense of padding the schedule isn’t all bad – I need to plan when I arrive at time. However this is a compromise, and as such a little is good, but a lot is bad. Trains should be on-time, but also as fast as possible. The Swiss model doesn’t run trains in 2 hours that could do the trip in 61minutes, they fix the rails so they can run it in 55 minutes.