News:

Want to praise Simutrans?
Your feedback is important for us ;D.

Discussion from server game about extended features

Started by Vladki, February 23, 2019, 02:43:28 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Vladki

Discussing playability with Elemental on the Stephenson-Siemens game. We noticed that several players started a game, but went either quickly bankrupt, or lost interest even if that managed to make a profitable line. The map is not big (1024x1024), and quite hilly. What we have identified as potential problems.

1. curves and hills: New players do not realize that they should avoid sharp corners and steep hills.
2. classes: To have enough passengers, you have to reassign Low to Very Low. Otherwise you don't have enough passengers to be profitable. Higher classes are just a nice bonus (maybe more important on bigger maps).
3. goods: in 19th century the demand is so low that it is uneconomical. Maintenance of loading bay at farm costs more that you can earn transporting those few bags of grain (wool, whatever).
4. farms: are everywhere, and they are generated at the minimum size, so that you cannot remove a single field from them.
5. signalling: new players just don't know how to use it, and get their trains stuck. (Or they hit some bug in signalling, not knowing if it is a bug or not).

jamespetts

Thank you for your feedback. It is not easy to know how to deal with all of these points. Items 1 and 5 would seem to suggest adding tutorial videos, although there are already some for signalling, and further videos will have to wait for more important things to be done.  Any bugs in the signalling should of course be reported with a reproduction case.

As to item 2, is it still the case that there are very few passengers travelling with the "low" or a higher class? There had been a bug that caused all returning passengers to be "very low", but this has been fixed. The proportions are otherwise set by the individual buildings. I believe that these are accurate proportions, but is there some reason to believe that these are not quite right? Making any change in balance always requires a very large amount of work as every single item needs to be changed in a consistent way, so it is important to have a good empirical basis for being quite sure that some other proportions are definitely more accurate before making any changes to this. Does anyone have any empirical data on this?

As to the third point, this has been discussed extensively, and it is difficult to know what can be done about this pending the planned major rewrite of the town growth code, which will have to wait until the completion of the work on schedule and related features, which itself has been suspended pending the resolution of the critical bug relating to the loss of synchronisation, which I can only productively work on when I have many, many consecutive hours spare, which is very rare.

As to item 4, I had changed the code a significant amount of time ago to prevent farms from being generated at a size below which it is not possible to remove any fields: one can now remove fields from a farm down to half its minimum generation size. Is this mechanism not working? If so, I will need a dedicated bug report thread with a full reproduction case to look into this.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

We'll try default classes on the server game, to see what happens. However these are the current numbers of passwngers on my trains (VH/H/M/L/VL):
0/17/6/7/67
3/13/25/9/74
8/11/9/15/74
5/17/4/9/105
4/66/35/81/49
4/20/12/55/110
0/38/19/57/90
5/8/18/12/51
As you can see it is more than half of passengers that is of Very low class

The game is in march 1897

jamespetts

Do you believe that these figures are outside the actual proportions of people of the respective levels of wealth in 1897?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

Quote from: jamespetts on February 23, 2019, 03:49:04 PM
Do you believe that these figures are outside the actual proportions of people of the respective levels of wealth in 1897?
I have no idea. But if it is, then either the transport companies should be profitable even if they do not offer any service for very low class, or there should be some options for these people to travel.
Do you think that more then half of the population at that time was so poor that they could not afford traveling even in the third class?  How about: 
QuoteThe Railway Regulation Act 1844 was an Act of the Parliament of the United Kingdom providing a minimum standard for rail passenger travel. It provided compulsory services at a price affordable to poorer people to enable them to travel to find work. It is one of the Railway Regulation Acts 1840 to 1893.[1]
And nowadays there are usually only 1st and 2nd class carriages. If they are coded as medium and high, the situation would be even worse. The designation should be by price/wealth, not comfort...

The question is what ratio is reasonable - if I have original class - higher revenue per passenger, but much less passengers. The current balancing seem to be in favor to get more passangers even if they pay slightly less.

jamespetts

It may be better to have one topic for each of the issues, as it is extremely hard to follow a topic with so many completely separate issues as this. I suggest that this topic continue with  passenger classes and that fields be split into a separate topic, which I will do presently.

In relation to classes, as I believe has been discussed before, there was a time when many people could not afford to travel by rail, even third class. There were special excursion trains as late as the 1930s (and I believe even 1950s or 1960s) which had reduced fares to allow people who could not otherwise travel to do so. Workmens' trains also had fares below the level of third class fares. There were fourth class carriages in the 1840s and 1850s in some places.

The change of name from "third" to "second" class was purely cosmetic - the price did not increase to match the old second class carriages. This only happened once the old "second" class had been abolished entirely, which only occurred when the ocean liner trains, which had classes to match the classes on the liners, were terminated. This enabled the railways to complete the process that began in 1875 when the Midland Railway abolished second class and increased the comfort of its third class carriages to match what was formerly second class.

You write about what is reasonable - but what is reasonable in a game that aims to simulate reality is that which most successfully simulates reality. The mere fact that, if one allows passengers at the lowest level of income to travel, there will be more passengers does not by itself mean that there is anything wrong with the simultion. Detailed comparison with reality is necessary to calculate with some precision whether or not this is correct. If a profit cannot be made without setting the classes to "very low", there are a large number of things that will require exhaustive investigation before it can be determined whether or not (and if so to what extent) buildings have the wrong proportion of passengers of "very low" class. If there is such an issue, it may well be that the buildings in tows are not regenerating to buildings that produce higher class passengers, or too many buildings that produce lower class passengers are being built, which is a town growth algorithm issue.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

Sorry for mixing it all together.

Regarding the classes, we have a few observations:

- especially in the beginning of the game, higher classes make so little traffic, that you go bankrupt due to maintenance and running costs.
- most (if not all) vehicles are profitable even when set to very low class, but most important is that the numbers of passengers gets much higher (more than double).
- from the previous example I have summed up the numbers of passengers according to wealth: 25/190/128/245/620. That is sligthly more than 50 % being the very low class.
- once your network is big enough, you may survive switching to default classes, but will see a big drop in profit.
- we did an experiment on server with quite a big network (many players interconnected, no cargo, some mail), and all players experience drop in transported passengers to approximately 50 %
- the drop in profit is critical, some smaller companies (with monthly profit <10000 cr.) are not profitable any more. It seems that railways are not affected that badly, but omnibuses and boats are losing money. Maybe because trains have high overcrowding ratio, so now they are just not so overcrowded. And thus the higher revenue makes for the reduced number of passengers. But buses and boats have little overcrowding and lose money. Trams were affected too (although they are very low by default), due to network effects. You just may rise the prices, as all the passengers will transfer to some other vehicle, which is low or higher class. (Perhaps a really big tram network might support local traffic for very low class).

So, the main question is:
- is it really intended that 50% of passengers are "very low", only 2% "very high" and the rest is almost evenly split between high/med/low ?

If it is really the intended distribution of wealth, then the vehicle and infrastructure costs and maintenance might not be balanced properly against the possible revenue.

If not, then the passenger generation, town growth or class levels in buildings might need to be adjusted.

Nowadays everybody who want's to travel (at least to commute), can probably afford it (excluding homeless and unemployed people).  Companies sometimes lower the ticket prices to attract more passengers, but usually only to take them over from competing company. But we don't know how it was 100 years ago. How big were the profits of transport companies (not only railways, but also stagecoaches, omnibuses, boats, etc.). Would they get more passengers if they lowered the prices? Would the number of new passengers be high enough to offset the lower price? At least in simutrans the answer is yes. Why then it did not happen in real world? Cheaper travel for everyone?

jamespetts

#7
Thank you for the observations. Looking into this is likely to require some very extensive work: I have spent a few minutes trying unsuccessfully to find data on wealth distribution in the 19th century. I believe that the existing figures for the buildings are based on at least approximate data, but I do not have the detailed data to hand now.

There are a number of possible causes of the profit effects seen, which do not seem entirely realistic, as the profits ought to be maximum when the default classes are used, at least in most cases. The possibilities, so far as I can work out at present, are that one or any combination of the below things might be wholly or partly responsible:

(1) the buildings have too high a ratio of "very low" passengers;
(2) too high a ratio of buildings with a high proportion of "very low" passengers are built;
(3) the fares for "very low" passengers are too high*;
(4) the infrastructure costs are too high; and/or
(5) the vehicle maintenance costs are too high.

* These are the only fares that were not directly based on data about 19th century railway ticket prices, as data on 4th class tickets/workmens' fares are harder to come by, and the relative prices are often not consistent.

The current balance in the game was an interim balance set up by Dr. Supergood last year; Dr. Supergood routinely reduces the class of vehicles to "very low", so I infer that the costs were calibrated to this level. Reducing the fares of "very low" passengers and also reducing maintenance costs might well mitigate this possibly anomalous profit effect, giving higher relative profits if higher classes be carried.

This is a complex balancing issue that is likely to require much detailed work over the long term - I will have to concentrate on the bugs and critical features before I can work on this fully. However, any detailed real data based calibration experiments or just real world data obtainment would be most useful in the meantime.

Edit: What would be especially helpful is if people could test the in-game data against the real world data to see which of the five hypotheses above are most likely to be correct and in each case in what way and to what extent.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Jando

I believe that the ratios of the different passenger classes are balanced quite nicely in the current game and that the ratio is quite realistic for the 19th century as well.

However - a player will always do what benefits his gaming experience (and esp. in a competitive server game what benefits his competitive rank). If a player gets better profit when he/she throws reality out of the window and offers transportation to people that could in reality not afford the ticket price then that is what a player will do. I'm not critical of this behaviour, I have done the same in my current game and offer very low class transport. I'm aware that this is not realistic for the time period but playability always wins over historical accuracy.

In my opinion there's no reason to combat this behaviour, players will always find a way to maximize their gains. Even when you, James, rebalance the wealth classes and costs players will still do whatever is needed to increase their profits.

My current personal excuse is that I subsidize my freight routes with my passenger routes. My company would have gone broke since long if not for the magical money tree called passenger service. :)

Vladki

Maximizing profit is ok, but vehicles should be profitable at the default class assignment.

It is especially big problem in the beginning, when you have small network and not frequent service. Thus many potential travelers will not travel because the journey will take too long, or there is no route at all.

DrSuperGood

I think the main problem is that time tolerances are too small early on. Rich people, the sort who could afford to travel, would not think anything of spending many weeks to do so since that was the only way one could travel. Instead they will choose not to travel because it is too slow because they expect something like a 1920 car or train speed in 1750.

On the Bridgewater Brunel server I only started to notice practical (enough to fill vehicles) very high traffic in the early 1900, and especially when aircraft were introduced (when they worked...). Before then Very High was an extremely rare sight, as if the gentry did not bother to travel at all.

That said coaches do work with low and medium passengers. One must just have a long enough line that is regularly serviced. Boats on the other hand do not. Until steam it is near impossible to fill anything above 1-2 holds of high  and low passengers in a boat with 7 holds. I suspect boats are meant to move wares, but that suffers from other issues currently (not worth it, bad cargo granularity).

Vladki

I'm analysing the graphs from the server game experiment, when we let it running for a few game months on default class settings, and then returned back to setting low to very low.
It is very interesting, that overall the profits dropped - for some companies even started to lose money. But if looking at individual lines, I can find cases when the increased revenue was able to offset the drop in passenger number, so the profit was almost the same, and some that were significantly affected. What is even more interesting is that they used the same vehicles (knifeboard horse omnibus + 4x clydesdale horses). Ships have been affected much more, trains were not affected too much.

Spenk009

I would like to extend a little caution about the topic. I have found that if I load an older savegame, companies usually get in trouble with ridership. If I switch to the public player and increase the town's population manually by clicking repeatedly, initially the town population stays similar but the job numbers increase drastically as the town rebuilds itself. When the amount of jobs reaches around 20-25%, the population increases (somewhat) in relation to the amount added by the tool. If this is done, passenger numbers match that of games generated later. I'm afraid I don't know whether this is directly related to DrSuperGood's balancing.


Quote from: Vladki on February 26, 2019, 05:09:58 PMIt is especially big problem in the beginning, when you have small network and not frequent service. Thus many potential travelers will not travel because the journey will take too long, or there is no route at all.
I have found that passenger numbers feel quite reasonable. In the early stages of a railway company I need to reduce fares in order to attract more passengers that need a more local service. I believe a fourth class passenger is more likely to undertake a commute into a big city than a random visiting trip 4 hours away, whereas a first class passenger will wish to make use of the opposite (as they have their own more suitable means locally). Later on, with a broad range of destinations, focusing on enabling the possibility of higher class travel we can do away with cheap (and by then overcrowded) travel.
Quote from: Vladki on February 26, 2019, 05:09:58 PMMaximizing profit is ok, but vehicles should be profitable at the default class assignment.
To me, this is exactly what the intention of the classes system is. This update my warning focuses on feels spot on, where you need to adapt to what you want to achieve. Transport for everyone with potentially extreme passenger numbers or pricey services tailored between services.

jamespetts

Thank you all for your feedback on this topic. A few notes on some of the issues raised, in no particular order.

First of all, the design intention is that there should not be a tension between doing what is realistic and what maximises profits: in reality, transport company owners wanted to maximise profits, so, if the game is balanced properly, doing what will maximise profits in game will tend to produce realistic behaviour. That is how Simutrans-Extended is intended to work.

Secondly, I note with interest Vladki's experiment - but I am slightly unclear on which way round that the results went as the post is unclear: was revenue mostly higher or mostly lower with "very low"?

Thirdly, as to variances in journey time tolerance with wealth and era, empirical data show that this does not occur: people tend to have an overall daily travel time budget which remains remarkably constant. Increased journey speeds therefore lead to people travelling further and/or more often. In reality, people now travel further and more often than they did 100 years ago, and people 100 years ago travelled further and more often than people 200 years ago.

Fourthly, as to "very high" passenger numbers, do not forget that a high proportion of people in this class have access to private cars (or private horse carriages), so the door to door public transport service will need to be faster than the equivalent by private road transport to attract these people (who are far fewer in number than any of the other classes).
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

The revenues (profits) were much higher when the low class vehicles were reassigned to very low. In general many lines have turned unprofitable with default class assignment. Only a few, specifically trains had similar profits with both default and very low, but it was imho due to significant overcrowding with very low class. So they were quite full in default setting as well.

jamespetts

Thank you for that: that is useful. It may well be that some very extensive analysis of the numbers will be necessary in due course, but that is likely to take a gargantuan amount of work and will only be viable once vehicle cost balancing is complete, which requires a number of significant additional features.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Jando

Personally I don't mind the revenue issue that much. However, when a player does not transport very low class passengers they can still overcrowd small stops because they show up in the stop for the horse omnibus even though the omnibus does not service very low class passengers.

Thus a player sees an overcrowded stop and his horse carriages leave that stop empty (or almost so). For that reason alone I have switched my horse carriages to transport all classes.

jamespetts

#17
Quote from: Jando on March 11, 2019, 11:25:40 PM
Personally I don't mind the revenue issue that much. However, when a player does not transport very low class passengers they can still overcrowd small stops because they show up in the stop for the horse omnibus even though the omnibus does not service very low class passengers.

Thus a player sees an overcrowded stop and his horse carriages leave that stop empty (or almost so). For that reason alone I have switched my horse carriages to transport all classes.

Did we ever manage to get a reliable reproduction case for this? I must confess that I have been so focussed on the loss of synchronisation bug recently that I have not always kept up to date with other reports. I did not spot a thread for this issue in my recent audit of bug reports, suggesting that it never had its own thread and possibly that it never had a reliable reproduction case. If you (or anyone else) has a reliable reproduction case for this, please post it in its own dedicated bug report thread.

Edit: Now that I think about it, I recall that there was a bug report with a good reproduction case of this sort earlier this year, and I fixed the bug underlying it (which was caused by returning passengers erroneously reverting to the very low class rather than retaining their originally assigned class). Can you confirm that the behaviour that you describe has now been fixed? If not, I should be grateful for a reliable reproduction case so that I can investigate further.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.