Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Maolagin]

Not sure if Dwarf Mode is the best forum for this, but seeing as "World gen recipes" lives here happily...

The basic question is: if you regenerate a world from an exported params file (thus, exact same settings and seeds), how closely should the resulting world match the original?

In 0.40 just now I wanted to get a snapshot of a world-gen fort that I reclaimed before my dwarves had started messing it up. Since I didn't think to make a pre-embark backup, I tried re-genning from the exported params, thinking I could just go back and find the site. World gen wasn't even done and I could see it wasn't going to work -- sites were popping up in completely different places, to the extent that the world was colored differently from the different distribution of good and evil. On closer inspection, I see that the geography and region names all match, but I got totally different civs and I think different beasts/gods/demons. No relation to the timeline from the previous iteration of this world.

Just to check that I wasn't expecting misremembering how things worked, I genned a 0.34 world, exported params, and re-genned. Comparing the two I seem to have the same sites in the same locations. The historical maps are, by eye, identical. Checking events, major events seem identical as well, right down to the same sieges with the same combatants and outcomes taking place at the end of world gen. However, there are some interesting discrepancies. In both worlds a necromancer writes a book in 124, but the books have different names. And in one world, the Age of Dragon and Titan ends in 124 when a marsh titan is struck down by a dwarf. In the other, the same marsh titan defeats an elf that year.

And take that last battle, the destruction of dark fortress named Blackhates. It happens in 119 in one world, but in the other world the exact same battle takes place in 122. Firing up world gen a third time, it seems that human army was a little quicker on the draw, and a dwarf army a little slower: the battle that destroyed Blackhates the first two times is only a pillaging, and the dwarves destroy it in 121. Again, the relevant battles share names.

So back to my question at the top -- how repeatable do you expect world gen to be? Those of you who were around for various earlier versions, how does this compare to your experience? It seems to me that in 0.34 re-genning apparently gave you Bizarro Realm, where everything was almost the same as you remembered, whereas in 0.40 you get the same planet but history reboots completely.

[RealFear]

Ain't tried anything like using the same seeds, but It seems that in df2014, history seems to tend to snowball alot more harshly (probably because of the fact that all invasions are successful at this point), so if you do see more changes in history, it's going to be favoring one side or the other more.

[Melting Sky]

I think since the history generation is far more complicated than it used to be that a small change in initial conditions and battle outcomes can lead to radical changes later down the line.

Edit: I was ninja'd.

[Maolagin]

Ain't tried anything like using the same seeds, but It seems that in df2014, history seems to tend to snowball alot more harshly (probably because of the fact that all invasions are successful at this point), so if you do see more changes in history, it's going to be favoring one side or the other more.

I can totally see where history would be more chaotic -- as in sensitive to initial conditions -- in DF2014. But that's not what I'm seeing here. The game is actually placing totally different initial civilizations in totally different starting locations when it first starts running history.

Also, I think Toady only said that invasions always succeed in post-world-gen world activity. I've definitely seen invasions in legends where the defenders won.

[Melting Sky]

The starting civilizations at year 0 should be the same if all the seeds are, or at least I believe this is how it was intended to be. I could be wrong. I will try generating a couple of worlds using the same seed. Unfortunately all the worlds I have created thus far were in the initial release of .40 and I now have the newer version of it so I will have to start from scratch.

In the past you would get a battle that occasionally had a different outcome while running two worlds with the same seeds but it really wouldn't effect the map lay out, just some of the individual legends. In the current version the change of an outcome of a single battle could radically change the face of the world.

In the example you provided from the previous version where a marsh titan struck down a dwarf instead of an elf in one world an not the other. If that elf or dwarf was destined to later lead a revolution and conquer half the planet it would radically change the sorts of sites and civilizations you would find between the worlds that had otherwise started out the same. But what you seem to be saying isn't that the world's started out with the same starting entities and historical figures and later diverged but rather that it is starting out at year 0 with a completely different set of entities etc?

[krenshala]

I think the key here is the phrase "if all the seeds are [the same]".  OP, when you regenerated the world, did you verify that all the seed values were the same between the two?

[Maolagin]

I think the key here is the phrase "if all the seeds are [the same]".  OP, when you regenerated the world, did you verify that all the seed values were the same between the two?

Yeah, that's exactly what I did -- I exported the gen parameters and used that file exactly to generate a new world. In 0.34, that gave me back the exact same starting conditions, and nearly identical history. In 0.40, doing the same thing gives me the same geography and regions, but the initial state for history is totally different.

[Melting Sky]

I just generated two worlds using the same seeds and parameters and it all seemed to work as expected. All the starting entities and historical figures appear to be the same. The general history and site locations of the world also appear the same although I didn't look into that too deeply since I already wasted about half an hour trying to clone a nice world I made in a earlier version and kept making damn typos in the seeds.

My advice is to check and make sure that your world generation seeds and parameters are exactly the same. Try making two new worlds by putting in the same short and simple seed wherever it asks for one and leave all the parameters on default so there is little or no chance for making errors copying things. I used "bob" on all of mine.

Edit: I should add I did not let history run for long at all. I can't think of what else might be causing your problems. Are you generating this copied world on the same computer and OS as the original and the same version of DF?

Are there a load of repeated rejections when you generate this world?

Have you tried copying other worlds and if so do you have the same issues?

[wierd]

Also, make sure that your raws are identical.

Even just changing the order of otherwise identical materials in the raws can have profound effects on the resulting world. (Whole biomes being generated differently, and the like. The previous version was found to NOT alphabetically sort the mineral layer data entries loaded from the raw files, and instead just loaded them into the array in the order they appear in the raw file, meaning that if the seed generator says "This layer will be made from material listed in location 5 of the array", then that whole layer can be totally different, just by reordering the materials in the the raws. Identical raws, including order, are essential to consistent worldgens!)

[Maolagin]

I just generated two worlds using the same seeds and parameters and it all seemed to work as expected. All the starting entities and historical figures appear to be the same. The general history and site locations of the world also appear the same although I didn't look into that too deeply since I already wasted about half an hour trying to clone a nice world I made in a earlier version and kept making damn typos in the seeds.

My advice is to check and make sure that your world generation seeds and parameters are exactly the same. Try making two new worlds by putting in the same short and simple seed wherever it asks for one and leave all the parameters on default so there is little or no chance for making errors copying things. I used "bob" on all of mine.

Edit: I should add I did not let history run for long at all. I can't think of what else might be causing your problems. Are you generating this copied world on the same computer and OS as the original and the same version of DF?

Are there a load of repeated rejections when you generate this world?

Have you tried copying other worlds and if so do you have the same issues?

Well, I think it might have been a transient issue with 40.01, because I think I'm seeing the expected behavior in 40.03.

But yes, I did about the most simple test you could possibly do for this. I saved a world gen parameter set, complete with seeds, so that it shows up in the menu on the advanced world gen screen. I think I had it set to run for 50 years, but my diagnostic check has been comparing historical maps for year 1 and year 10. Then I just created a few worlds in a row off of that parameter set without touching anything else.

When I do that in 40.03, the maps are identical, just like in 34.11. When I did the same thing in 40.01, they didn't match for whatever reason. Think we can chalk it up to buggy early release weirdness. Thanks for the suggestions though!