Print

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

[Antsan]

However, I would request that knowledge of geography wouldn't be likely to be forgotten within a couple years just because nobody has walked there recently. That's fairly silly

I was in Nordhausen last week and there's a tower there to see (now below ground because garbage) that was built in the 16th century (I believe) and known up into the second world war. When it was found again via construction work there was noone, who remebered it, until someone said "Hey, wasn't there something about a tower?".
A whole city forgot that thing after, what, not even a century.. Forgetting geographical details over a few years seems really likely.

[Dirst]

I agree with GavJ, though I'm not sure what you mean by X% accuracy.

What I meant about pre-embark information is based entirely on prospect's behavior.  It gives different numbers pre- and post-embark (and the former are marked as estimates).  Since it's a cheat tool, I can only assume the lack of detail is because fine details aren't set in stone until you embark.

When reporting generic layer information, it should come with summaries based on accurate math using blind projections (that is, based on the raws, not any specific knowledge of that particular site).

[GavJ]

Antsan, World War 2 was 70 years ago... not "a few." Also dwarves live twice as long as humans, so something in human history forgotten in 70 years would generally take 150.

I agree with GavJ, though I'm not sure what you mean by X% accuracy.

Actually, reading more closely, I agree that "% accuracy" doesn't mean anything. 3 significant figures at 80% accuracy is basically the same thing as saying "1 significant figure." If you want a more rigorous/controllable measure of variance and error and so forth, ditch the whole significant figures thing and use a proper standard deviation amount:

1) Take the actual amounts of everything (hidden), and apply whatever level of standard deviation.
2) Now figure out the ratios
3) Print the ratios at just always XX%.
4) Also print the standard deviation you used on the screen for that area. Something like:
"5% of actual amounts = 1 standard deviation" for the highest detail level around your major cities
15% small towns
25% your political border as per the normal legends map
50% some distance beyond that
(and places sufficiently traveled get bumped up then slowly decay over many decades)

[Antsan]

I know how long ago it was (although it was only 50 years at most as it was found 20 years ago), the point is that it was right in the middle of their city. They buried it under garbage and rubble and forgot about it.
Something that is more than a week of travel away won't last very long and only because there is something written that doesn't mean anybody reads it. Information deterioration can be quite high.

[SixOfSpades]

I would request that knowledge of geography wouldn't be likely to be forgotten within a couple years just because nobody has walked there recently. That's fairly silly. . . . Like... maybe 2-3 trips might build a level, whereas losing one might take 100 years or something.

Maybe not a couple of years, or 100 years, but more like 8-10 years. So let's say that "light-moderate" traffic (1 or 2 visits per year) can raise a road's surroundings by 1 detail level, "heavy" traffic kicks it up by 2 levels, and no recorded traffic at all (from your civ) causes those areas to drop by 1 level every 5 years. After all, there's landslides, wildfires, floods, I assume Toady will work in earthquakes & volcanic eruptions at some point, theoretically there may even be hurricanes. And now that the trees are huge, a single tree falling down could (in the real world, at least) dam a stream and create a small lake. And then there's logging, mining, terrace farming, construction, destruction, defensive earthworks, and apparently people burying entire towers under garbage--I'll just call that "landfills". Plus the wake of devastation assumedly wrought by megabeasts. Now, granted, this is pretty much everything that can happen to the landscape, so the list seems a little extreme . . . but you've got to remember, the detail only drops by 1 level after 5 consecutive years of no recorded traffic. It takes a lot to stop/divert traffic for that long, and a lot can happen in 5 years.

That's not to say it should magically keep up to date with new buildings or mining, though.

Yeah. Until/unless Toady actually implements things like forest fires and new towns popping up, the game should only act like these things actually happen. So if you go for 10 years without going near a particular river, nobody will have actually built a bridge over the river . . . but that area should still be shown at a lower level of detail, because somebody might have built that bridge. Or that dam. Or mined out all the aluminum. Etc.

I agree with GavJ, though I'm not sure what you mean by X% accuracy.

Don't feel bad, I don't know what I mean by it either. I was just shooting for a rough estimate of how "trustworthy" the embark window should be for each of the 3 depths, at each level of detail. As for the significant figures, I was trying to mirror how the Bookkeeper job has varying degrees of precision, although really that was just for flavor. Once I/we know more about how the game actually generates site composition from the raws, and understand more about pre-/post-embark differences, then I/we will be able to get into more exact nuts & bolts.

When reporting generic layer information, it should come with summaries based on accurate math using blind projections (that is, based on the raws, not any specific knowledge of that particular site).

Well, in areas of Full and High detail, your civ actually quite likely would have specific knowledge of that particular site, at least on the surface. They might even know exactly how many of which type of ores/gems are sticking out of the ground, it's just that no Miner has chosen to come collect them yet. But continuing this line of thought leads to awkward "half-formed" sites, where just browsing over it in the Embark window causes the game to finalize the Surface details, but still leaves the Shallow and Deep sections unfinished until you actually show up there.

If you want a more rigorous/controllable measure of variance and error and so forth, ditch the whole significant figures thing and use a proper standard deviation amount:
1) Take the actual amounts of everything (hidden), and apply whatever level of standard deviation.

Fine, but we'll need numbers we can crunch before sites are finalized, I'm hoping Dirst can provide some insight on that. As long as the game generates "probable composition", whose accuracy we can artificially degrade with distance from the Mountainhome and below the surface, I think we're all good. I'd also like it if the margin of error included the possibility of failing to detect features and/or generating false positives, at least at the low end of the site-familiarity scale, but that's not an absolute necessity.

[GavJ]

Fine, but we'll need numbers we can crunch before sites are finalized, I'm hoping Dirst can provide some insight on that.

The system is deterministic. You can generate the actual map temporarily if needed for calculations and then trash it and then go back and generate it later and get the same thing.

As far as I'm concerned, if you desire super accurate deepest level info when available, you can sit there for 5 seconds while it loads.

For example, if a location has level 1 info available, it still shows up at level 3 or whatever can be done easily without actually genning the tiles, but then still SAYS "level 1 available" blah blah, and there's some key you can hit to switch to that, which will take a little while, if you think the rougher scan looks worthwhile.

If you decide not to embark there later, it can just chuck the info in the trash from RAM, and still takes up no more disk space than before.

[2074red2074]

Jeez, if it got to the point where I didn't know what to expect, I would stop playing. There is a reason that people look at shallow and deep metals and aquifers and such. You can't sacrifice fun (not "fun," just fun.) for realism.

[GavJ]

Jeez, if it got to the point where I didn't know what to expect, I would stop playing. There is a reason that people look at shallow and deep metals and aquifers and such. You can't sacrifice fun (not "fun," just fun.) for realism.

I'm not sure what you're responding to. There are multiple reasons you shouldn't be worried if the system being discussed here were implemented:
1) Everybody has agreed/independently mentioned that a "classic mode" should be included.
2) In a small-medium map at 200 years history, probably half the map would be at or above current levels of information.
3) When there is less information, it TELLS you to expect greater variability. You would never be surprised about being surprised.
4) The suggested system includes a significant amount of land with HIGHER than current amounts of information, so you could, if you so chose, actually know even more about what is going on in your embark than you currently do.

[Dirst]

As GavJ mentioned, a set of pre-embark geology data deterministically gives you the same post-embark geology data every time (you can try this duplicating your saves and embarking in the same spot) although I'm not sure about creatures.  It probably uses the random number generator, but from a specific seed so that the same "random" numbers appear each time.  This could take longer than 5 seconds, though.  It might be useful to generate it in the background while the player is browsing the basic info.

I should clarify that when I said "generic layer information" I meant a level of detail that doesn't incorporate actual deposits, just what could be expected if all you knew was "sedimentary layer".  Would be a bit more useful if you knew "chalk layer".