Hi!
Looking around, this seems to be the best place to share some observations of mine concerning world gen.
First of all, the factors I am considering are humans, elves, goblins, dwarves, megabeasts, length of time, event number. Basically, I assume that people want all races with civs, at least one megabeast on the main continent, a long length of time, and an exciting history. And it seems that it is relatively difficult to get all of these things at the same time.
I am currently in the process of generating worlds with a very promising approach, but I don't have any concrete results yet (the first world I accepted had no goblins and 1 dragon, probably on the main continent, in 963 when probably the other long time megabeast got killed). For people having similar notions, I want to share this approach in general terms as it seems the most promising thus far to me.
The world size I am using is MEDIUM (I had considered LARGE, but since the depot would not accept such saves, I figured it wasn't worth it).
Of course I put in a percentage for megabeasts dead (90% for the above cave, now I am trying 80%) and have the check start in year 2 (after all, depending on your luck, things may collapse much sooner than 200).
The variance levels are somewhere between 400 and 800 (espially height near 800), although reducing temperature variance has no negative impact.
I multiply the number of good and evil squares and regions etc. by 4 each.
Biome requirements are 300 - 600 squares for wetlands (0:0 regions), 20 or some such squares of desert (0:0 regions), approx. 5200 squares of forest with maybe 5:5 regions, about 2800 mountain squares (3:3 regions), 100 - 1000 squares ocean (0:0 or 1:1 regions). All other biome minimums are nullified, especially grassland and hills!
Maximum number for rivers and erosion (but erosion for cliffs turned off) is rather a matter of preference, as is having about 100 volcanoes and several dozen mountain peaks.
I suggest border ocean 1 or 0, as this setup is designed to have few oceans.
Maximum number of regions should be maxed out (otherwise you could get last time rejects).
Cave size minimum is a matter of preference, I think, although I suggest 200-400. Cave size max should probably be maxed out. Cave number must be 800+800 (in other words, maxed out).
100 starting civilizations and the population cap for civilizations should be nullified!
The final set of minimums (drainage and so on) should all be nullified as they are unnecessary rejects, except for high savagery 3500,low savagery 1000 and mid savagery 3000.
These values are all vague suggestions.
Anyhow, my reasoning is as follows:
Dwarves and goblins settle only in mountains and elves and humans only in non-mountains (although humans and dwarves will take over other civs' settlements).
With decent height variation, dwarves and goblins have good chances of developing regions of control - one mountain chain may be dwarven with destroyed/conquered dark fortresses while one may have destroyed mountain halls next to flourishing goblin fortresses. Because of them not connected, there seems to be little likelihood of one conquering the other (provided you reach that situation in the first place, of course
).
Elves are also not a big problem.
First of all, they seem to be most vulnerable to non-mountain caves with ruined settlements showing up in the first 40 years without any other civ around once you have more than 200 non-mountain caves.
Secondly, they only settle in forests and do not conquer other civs' settlements (they only destroy them). They seem to be much more likely to be held by natural boundaries than humans.
The biggest problem seem to be humans as they tend to spread rather quickly and have good chances of conquering complete elven civs (as they take over forest retreats, they are more likely to reach even the remote elven settlements while the elves stay close to their native forests).
Therefore, I figured that increasing the number of forest squares may help encourage the elves (who are usually the first to get whiped out) while weakening the humans. Thus far, it seems that you do not need to worry about human civs; I have yet to see a world with more than 10 civs without a human civ - just like cockroaches, you simply can't get rid of them.
Now, with 100 civs, it seems probable that you get quite some activity in the history (=lots of interesting events) as things get mighty crowded on the map.
And with the elves as buffers for human expansion, you are likely to have a few remote swamps and the like where megabeasts can retreat to while feasting on the comparatively weak elves.
The results have been promising thus far (although you do need some luck to get the right distribution, of course), so I think this information may be helpful for others.
Deathworks