Feature discussion: journey time tolerance

[jamespetts]

I have had a thought for a new feature for Simutrans-Experimental that would be quite simple to add, but would potentially make a significant improvement to the economic realism of the game. I should be interested in people's views before I start to code it.

The idea is that passengers (and also possibly mail) would have a variable level of tolerance for journey times. If the actual (estimated) journey time of the shortest route from their origin to their destination exceeded that time, the passengers would not travel. The tolerable journey time for each packet of passengers would be set randomly, so, as improving technology gradually reduces journey times, new markets open up to players to transport passengers further than they would previously have been willing to go. With lower journey times, a higher proportion of town passengers would be transported, thus making the growth of towns dependant in part on the speed, as well as the capacity, of the network. This system would also make it easier to simulate very early transport networks more realistically: large stagecoach networks would attract relatively few passengers because of the long journeys involved. Those fewer numbers of passengers would be commensurate with the stagecoaches' small passenger capacity, thus preventing extreme overcrowding of early, slow, low-capacity networks.

A few things I have not resolved yet, and on which I should like feedback. Firstly, ought this apply in addition to, or instead of, the existing Simutrans-Experimental feature of having different proportions of passengers making local, mid-range and long-distance trips? If it should be in addition to this system, how ought the two interact? Would having a different range of journey time tolerances for short, medium and long-range distances work? If so, how should the numbers be calibrated? Secondly, ought this be just for passengers, or ought it also be for mail (on the basis that some mail is only worth sending if it gets to its recipient within a certain time)? Thirdly, is there any room for this working for freight, or does the system of linking particular factories make it impracticable? Fourthly, should the ranges be hard-coded, or set in simuconf.tab? Fifthly, ought a proportion of passengers have an unlimited tolerance of journey times, or should there always be a maximum? Sixthly, how should passengers who can find a route but only one that would exceed their tolerance level be handled in terms of statistics? Should they be recorded under "no route", or should there be a new category called "journey too long" (or something similar - any suggestions?) to deal with this particular eventuality? Finally, can anyone think of any gameplay/economic implications of this idea (whether good, bad or indifferent) of which I have not thought (and listed in the previous paragraph)?

I should very much appreciate any thoughts on this topic.

[The Hood]

Historically speaking, there have always been the most trips within 1 hour of the origin (no matter how far).  So if a new road is built to relieve congestion, people benefit for a while from shorter journeys, then they realise that they can travel further in the same time they used to travel, so travel further and create new congestion...

So it might be interesting to model that - as soon as more/faster connections are available then more people actually want to travel (rather than simply are able to travel, as present)

[jamespetts]

The Hood,

thank you very much for your input :-) On a slightly more detailed level, what are your thoughts on how this might interact with the existing system of different propensities to travel short, medium and long distances?

[The Hood]

Err... my comments were more general support and adding my own ideas.

One possible way is to have it so that instead of short, medium, and long distances, you could do it for time inestead, somehow.  So e.g. 50% of trips generated would want to go somewhere they could get within 1 hour (or whatever).  The only thing is I'm not sure how this would fit with only having limited choice of destinations.  In sim-experimental though, this could be related to car driving and the car speed for a given year - so there are so many trips generated for a certain number of tiles which increases with the private car speed.  Then if people can't get to where they want (either at all or within their time allocation), by the public transport network, the trips are still "there" in the game and could be transferred if a new train line was built, but otherwise they don't appear at stops.  The other problem though is early in the game, private "car" speed would be very low (horse and carriage!), and building a new railway attaining the dizzying speeds of 20mph should generate a lot of additional demand too because of the new opportunities...

[Bernd Gabriel]

Let's ask the question: why do people, mail and goods travel?

Goods: industries and wholesales need goods which suppliers offer.

Clever salesmen prefer the most cost-saving goods. In simutrans this means: from the nearest supplier. They pay the carrier for the distance between start and end. And this is the challenge playing simutrans: can you transport the goods and make profit or are the goods sent through the detour hell? For simutrans there is no distance limit, except the one (or am I dreaming it?) for industry chaining.

BTW: There might be a warning, that a convoy cannot make profit due to too long detour (or/and show the max. return according to schedule, capacity and shortest way. Show also the monthly maintenance of used infrastructure).

BTW: Sometimes I'd like to view the calculated route (longer routes in the overview). Should I try to program that?

Mail: industries mail to (possible) suppliers and consumers, people mail to people which they cannot travel to.

In simutrans business mail should have no distance limit. Private mail quantity might depend on radius from home. E.g. same quantity for each radius or slightly falling with increasing radius. This results in a high density of sent mail in your local area and sparse in the remote area.

BTW: Is it an idea for a new industry chain: postal shopping? Producing mail from several (boxed) goods. Hm, there is no 'or' relation between the resources, isn't it?

People: they travel to employment, to relatives and friends, jaunts/vacation trips.

They only travel, if it is not too cost-intensive or time-consuming. Some people have more time and/or money than others. Rich people have more money than time and will take the expensive aircrafts, while the middle class will take the train or bus and the unemployeed poor man has a lot of time, but no money and stays at home.

The 3-zone-model now appears to be an intuitive approach to this time-and-cost-tolerance-model, where distance is an origin for time and cost.

The time-tolerance-model is one part of this time-and-cost-tolerance-model, but together with The Hood's interesting finding about travel times the time-tolerance is a fixed value for all times. Of course, this does not mean that noone ever goes on a trip that's scheduled to last 1 hour and 5 minutes. But there is a considerable loss of tolerance beyond this value.

The cost-tolerance makes the difference between people. Lately I added an aircraft line to relieve a 2 hop train route. Noone ever entered a train again and the airports were overcrowded at once. In real life, some people would have changed to the aircrafts, but not all. In a balanced simutrans without inflation the cost-tolerance could be a fixed value for all times. A random deviation from this value (maybe influencable by (residental) buildings) models the various liquidities.

[jamespetts]

Bernd,

thank you very much for your input :-) We're not really in a position to add any major new features that would require changes in lots of places in the code at present (such as price sensitivity or  different handling of private and business mail, although at least the former would be good to include one day). It may well be worthwhile setting the minimum tolerance level to 60 minutes by default. Any thoughts about the other issues raised?

[Bernd Gabriel]

Sorry, like The Hood I was primarily writing some food for thought.

Let's get concrete:

to 1) The old model will be replaced completely by the time-tolerance as there must be a considerable loss of tolerance beyond the time-tolerance-threshold.
to 2) For passengers only. Mail is intended to reach the remotest places.
to 3) No, each suppliers goods should be able to reach the consumers. If there are any constraints, the chain builder should consider them.
to 4) simuconf.tab is a good place for the time-tolerance value.
to 5) The loss of tolerance beyond the time-tolerance-threshold should be randomized. So there will always be some long-haul travelers.
to 6) There should be a statistic for 'route too time-consuming' ('long' is mistakable for distance), because it is important for the carrier, which could try react with building a shorter route. Hm, the carrier does not know the abandoned destination(s)...