extremely large cities

[ojii]

Running e684a47722e4f5b40b0216094a6b427fd299519d with pak128.Britain-ex 0.8.3 (on OSX) I get very odd behaviour when creating a new map.


The settings are all the default, except for "Use timeline year" which I turned on to 1899.


Creating a map in that environment takes a very very long time, particularly the "creating cities part". This time varies a lot, sometimes it's just a few minutes, sometimes it's seemingly stuck and after 10 minutes there's still no map. Once the map finally loads, the reason for it becomes apparent, it creates absolutely massive cities way beyond the "Median Citizens per City" number of 1600.


Screenshot of a very "fast" load (~1minute) but still with a, in my opinion, grossly disproportioned city:


http://uploads.ojii.ch/simutrans/1.png


I'd assume this isn't expected behaviour, so is this a config mistake on my end (maybe turn "big cities" to 0?), or a bug?




EDIT:


Started a new game with "big cities = 0", all other options as above (Median Citizens = 1600), I still get 8 cities over 10'000 citizens (seems a bit far away from the median), with the top one at 40'000 citizens. City creation seems to go a bit crazy on my system..

[Junna]

I can confirm that - even on the start-up default little island that is created - the program loads very slowly during the "creating cities" part occasionally. Sometimes it loads very briskly, but at other times makes no progress for a 2-3 minutes. Usually this seems to be when the cities are larger. Neroden's been working on the city code recently though, so it is in a state of flux, I guess.

[jamespetts]

The generation of cities in Simutrans-Experimental is designed to follow Zipf's law, which does, in fact, mandate a very great disparity between the largest and most other cities.

[MCollett]

The generation of cities in Simutrans-Experimental is designed to follow Zipf's law, which does, in fact, mandate a very great disparity between the largest and most other cities.

More quantitatively, if you have N cities with a median population of p, the largest one should have a population of around N*p/2, and the next-largest around half that size.  So if you have 100 cities with a median of 1600, the largest should be about 80000, and the next largest about 40000.

Best wishes,
Matthew