Random Map Generation

Asked by MP

Not sure if this should be a bug, blueprint or just question.

One of my favorite things in Serf City was to randomly generate a size 8 or 9 world, find my own starting place near a mountain that had a fair amount of coal, iron and gold (in that priority, I believe), as well as trees and a quarry position (or stone in the mountains). I'd often penalize myself by a low reproduction rate and/or low starting supplies, so I'd have to build a toolmaker soon. Then I'd play for a couple hours, let it run overnight and play again for a couple hours the next day (thank you for the "speed up" option in WL:P) and repeat.

In any case, the current map generation routines seem "inferior" to those in Serf City (Settlers I). In SC it was relatively easy to find a castle spot near a mountain or two that had everything that I could want. In Widelands, I often see huge mountains that produce nothing but gold. Or stone. A lack of coal is not as fatal as it was in SC, as here we get charcoal burners, but the conversion rate (6:1) is pretty horrendous if that's the only source of "early" coal, given other building needs that require trees. In the late game it can be more manageable if you set aside "tree farms" and don't build as many new structures. Not having iron is downright fatal in SC. Yes, you can take down interior military buildings in WL without losing land and allowing you to expand (which you can't do as effectively in SC), but expansion is slow and if you haven't found iron by the time you encounter another civilization, you're pretty much screwed.

In any case, in WL the map generation routines seem to favor only 1 or 2 ore types per mountain, no matter the size of the mountain. Further, a given ore is found in a contiguous swath only. Although this makes Geologists (and dummy mines) less needed than in SC, it makes expansion difficult. There will sometimes be one good castle spot on a 144*144 map, but then the other players are at a complete disadvantage, as there are rarely two good starting positions. In SC, the resources were confined to smaller patches and the different ore types were more interspersed.

I'd like to have the individual ores tend to "decay" faster away from the center point than the current routines call for. Further a given mountain should possibly be checked to see if they have a second ore before placing a second cluster of the first ore, although this may not be needed if the decay rate allows for a lot of ore changes per mountain. Some regions could be empty of course as well, but currently there seem to be a few large areas where there just isn't anything (which seems to occur on the south face of mountains more?). Further, some consideration should be given to the balance of the ores needed for development, although I'm sure this depends on tribe and playstyle a bit too. However, a starting point might be 6:4:2:3:3 for coal:iron:gold:stone:nothing. My guess is the current routines weight each of the ores equally.

Of course land resources (trees, stones, fish, game, etc) should also be considered during the starting position picking. I almost feel though that the castle positions should be selected before seeding the map with trees and stone so that each starting civilization has immediate access to these.

Also, if the map generation routines are improved, I'd like to see generate map put not only in the map editor, but also in the "new game" menu. Right now that would be inadvisable because of how unlikely you are to get a workable starting location this way.

Revision history for this message
Nasenbaer (nasenbaer) said :
#1

You are completely right - the map generator is not yet ready for the masses - almost all maps generated by the random map generator must be worked on before they are really playable.

We are actually planinf the following feature:
> Also, if the map generation routines are improved, I'd like to see generate map put not only in the map editor,
> but also in the "new game" menu. Right now that would be inadvisable because of how unlikely you are to get a
> workable starting location this way.

but as you already stated, it would not be wise to integrate it at the current state of the map generator.

You would definitely do us a favor :), if you open up bug reports for the different points (bad resource placement, better starting position selection, etc.) ... so feel welcome to do so.

Can you help with this problem?

Provide an answer of your own, or ask MP for more information if necessary.

To post a message you must log in.