Print

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

[Devast]

The fact I can't make a giant 10x10 fortress with 500 dwarfs and get more than 0.001fps

[Thisfox]

I'm changing my answer.

It's massively frustrating to watch the dorfs dig out a huge rectangle of galena, and present me with just two mined galena ore. (Even more so when it's limonite). Meanwhile, every gemstone they dig through gets carried over to the gemstone stockpile to lie in glittering patterns on the floor. It's a small thing, but I really do wish that if they dug through one ore in their cavern wall, it would always return one ore for my smelter. Or more, for that matter. But not sodding less than one, dammit!

[Starver]

One ore (25% chance? 33%?) gives four bars, though, as do stone->block production. When one ore gave one bar (a few versions ago) the determinator for (non-gem?) digging-out was highly tied to miner skill (now only used to dictate speed of mining) so you could end up with net loss of processed material far greater than the current average.

Not saying it's better now (I remember deliberately training up on the commoner 'junk layer' stone layers whilst holding back on 'prime' extraction material as much as possible in the early days of an embark with less-than-Legendary miners) and the RNG probably can punish you badly (or reward you mightily... but inconveniently if it was just a clearance operation that you're not wanting and intend to just dump-clear the spoil).

So it's a balance, and a different balance. Perhaps a memoised-chance-adjustment[1] would have helped flatten the RNG-blues(/elation). But you could still find yourself with extreme malus/bonus results. Unless it was de-chanced and boringly 1:1(:1), without any variation at all.

(Also, previously suggested, potentially tagging mining as "mine with care", for slower but more certain extraction, and/or "mine to a pulp", for slower but more certain non-extraction. I remember these suggested before the Priority and Mark-Only thing went in, so it might well go in there rather than as a special-designation extension letter or Traffic/Zone-type designator as the original suggestion probably was.)


I feel your pain, but it should average out if only you stop tearing up four-leaf clovers while stomping on broken mirrors underneath a ladder that a black cat has just fallen from. 


[1] Perhaps something that remembered the prior results to arbitrary depth and weighted (by log N?) the chance that the distribution of the last N results (against the intended proportions) be used to directly dictate success/failure. So you don't get a regular and predictable "every fourth tile" but you do more quickly (and certainly) head towards the designed in ideal. *Warning*: Memory hog if you apply this tally to each individual subtype instead of/in parallel to a combined (shared) success-log component.

[Quietust]

One ore (25% chance? 33%?) gives four bars, though, as do stone->block production. When one ore gave one bar (a few versions ago) the determinator for (non-gem?) digging-out was highly tied to miner skill (now only used to dictate speed of mining) so you could end up with net loss of processed material far greater than the current average.

It's actually changed multiple times over the game's history - in the old 2D versions, everything (including raw adamantine) started at 25% yield and increased according to miner skill with different caps based on what was being mined (100% for metal ores, 62.5% for gems and coal, and 43.75% for stone), and in the later versions everything had a 100% cap (and raw adamantine always had a 100% yield regardless of skill level). In the current version, small clusters and individual tiles have a 100% drop rate and everything else is a flat 25%.

[1] Perhaps something that remembered the prior results to arbitrary depth and weighted (by log N?) the chance that the distribution of the last N results (against the intended proportions) be used to directly dictate success/failure. So you don't get a regular and predictable "every fourth tile" but you do more quickly (and certainly) head towards the designed in ideal. *Warning*: Memory hog if you apply this tally to each individual subtype instead of/in parallel to a combined (shared) success-log component.

The simplest method would be to just track that information within each "mineral inclusion" record, of which there is typically one per vein/cluster per 16x16x1 map region - they already keep track of the inclusion type (so that small clusters can give a 100% drop rate), so adding a pair of "successful mines"+"failed mines" variables would be easy and wouldn't increase memory usage in any significant way.

[Starver]

Spoiler (click to show/hide)

The simplest method would be to just track that information within each "mineral inclusion" record, of which there is typically one per vein/cluster per 16x16x1 map region - they already keep track of the inclusion type (so that small clusters can give a 100% drop rate), so adding a pair of "successful mines"+"failed mines" variables would be easy and wouldn't increase memory usage in any significant way.

I was thinking of something between the fully railroaded reversion-to-the-mean[1] and total memorylessness[2]. Ideally, it would have infinite full sequence memory, from which the compound sum of various length rolling-averages would influence the next pick. Not so much based upon either the full sequence or the immediate predecessors as a short-term FIFO stack, but blips would spark de-blips at various points, especially if the blip had not prevented a further re-blip recently (always should be possible) and thus now more de-blips are needed.

And to be implemented as everything(plus)subtype(plus)exacttype so you get exciting blips(/deblips) just to offset more-/less-global deblips(/blips) for an dditional piquant of de-dampened chaos.

This, far moreso than the simple count you give, would require much memory. At a pinch it could be rationalised to jot be a fully-literal sequence[4], but still would accumulate.

Also, I should say, this was not exactly serious as a suggestion. It was an overcomplicated solution in the fine tradition of overcomplicated dorfy things in general.

(I could have suggested a pre-prepared dithered procedurally-ratioed btmask of yay/nay results (or maybe skill thresholds needed), that only individually surprises but is strictly in range when you consider the whole. The randomness is introduced by player choice of where and how to dig them out. Does lend itself to DFHack interrogation (for 'cheaty' reasons) or save-scumming for the best approach (unless saves/reloads recalculate the unrealised determinations to a new pattern). Of course, I'm still leaving a component of annoyance, because I think it needs it for verisimilitude, but trying to knock the corners off the main annoyance of the current chances of getting a whole-lotta-nothing from much effort.)

[1] Is totA/totNotA>ratio? Then NotA. Is totA/totNotA<ratio? Then A. Else flipacoin?[3]
[2] An infinite sequence of any single result is (albeit infinitesimally) possible.
[3] If not, but >= or <= are used it's a strict repeating sequence. Done differently it could be some higher-order version of the Thue-Morse sequence, maybe.
[4] Sequence= [last], then [last,prior=previous.last], then [last,/,lastpair=previous.last+previous.prior], repeat to insert newer lastpair with priorpair=previous.lastpair, then the next shuffle involving the pairs bumps them into a combined lastquartet. (Names here just for 'readability'. It'd be just stacks that only lengthen as order LogN to the total individual values individually stored to it. And in a format (sum of each like-'sized' total) compatible with my intended sum-of-running-totals 'scheme'. 

Cheers for the overview. I know it had changed in parts since 38whatever, 40d and onwards, but I thought it might take too long for me to peruse all the changelogs, announcements and/or wiki-namespaces. But you, as always, actually have the encyclopedic lowdown. Or certainly close enough to brook no arguments from me.

[A_Curious_Cat]

Not sure if this is what you are looking for, but I embarked in a good biome one time and set up a gathering zone under a bunch of feather trees.  According to the wiki, it should have taken a lot longer for cave adaptation to kick in, but here were my dwarfs puking their guts out as the collected the feather tree eggs!

[rico6822]

Invading a dark fortress alone as overpowered adventurer only to get lagged because beak dogs and trolls seem to never sleep. I literally had to mod the game to remove beak dogs and trolls from dark fortresses and power up goblins to balance the challenge.

[Fikilili]

Glad you asked!

1. Lag, granted, it's not the game's fault. I'm not a programming genius (because I don't know anything about it), but I know for sure it's the game's fault 100.
2. Anything that spits webs. I know its supposed to be annoying, but it isn't really funny to see my militia dying of starvation because of a giant jumping spider that is spitting webs in their general direction.
3. The big wait. I don't feel too affected by this, because I'm waiting for things that have little to no hope of coming out; like a Season 2 of Dark Crystal
4. The early game. It's always a chore to me, because you have to make sure every dwarf is well fed, dressed, quenched and housed somewhere in your new fort. Perhaps I'm just impatient to get to my little projects?
5. Elves, but that's a given.
6. My militia inexplicably beating the fuck out of anyone they meet while going to attend their duty. I dunno what happened, but I lost a whole fort because of this. Seriously! They were supposed to kill a bunch of fire imps!
7. The very lackluster physical descriptions. I don't really wish to know how long is my dwarf's schlong, but I think we could use more descriptions regarding their appearance.
8. How the game tells its stories in Legends mode. It's very much the opposite of ergonomy. I don't expect Shakespearian writing, but something a bit more... Erm... Well, you know...

[PlumpHelmetMan]

On the topic of physical descriptors, it bugs me how often a description is generated that (for me at least) is really hard to visualize, like a dwarf with a head that's both extremely short and extremely narrow.

[Starver]

On the topic of physical descriptors, it bugs me how often a description is generated that (for me at least) is really hard to visualize, like a dwarf with a head that's both extremely short and extremely narrow.

Spoiler: This? (click to show/hide)

Spoiler: Alternately, assuming standard 'depth' (click to show/hide)

(It's the "three-legged quadruped" stuff that is hard to reconcile.)

[PlumpHelmetMan]

I mean maybe my visual imagination just isn't the best, but yeah "three-legged quadrupeds" are pretty maddening (though I thought they were fixed).

[Starver]

Possibly. I'd never even had them (or noticed them, not being so asiduous as to examine every detail like that) so I haven't particularly noticed anything of their later absence/non-generation either. But that (or four-legged biped, or somesuch) is what others reported.

(TBH, I'd have assumed a four-with-three to be an amputee or developmentary-misformed in the limb department example of what-should-have-had-four turning up with just the three. If it's a two-with-four then you're into conjoinedness territory, usually. Though making such RL assumptions might be just a little over analytical to support an 'observation'.)

[PlumpHelmetMan]

Also whenever I try to imagine the "great swinging lobes" descriptor for ears, it's not that I can't visualize it but it always ends up looking extremely cartoony in my head. Again, probably just the limitations of my own imagination more than anything else.

[Thisfox]

I mean maybe my visual imagination just isn't the best, but yeah "three-legged quadrupeds" are pretty maddening (though I thought they were fixed).

I had a cat who was a three-legged quadruped. She had an argument with a car and lost, but she was fine after the surgery. Got around on three paws very well. She was great at takeoffs, terrible at landings (It was a forepaw). She was not allowed outside again, of course. But yes, I know that's not what you mean.

[Magmacube_tr]

I mean maybe my visual imagination just isn't the best, but yeah "three-legged quadrupeds" are pretty maddening (though I thought they were fixed).

Think about it like this.

The creature has three limbs initially, but then the third limb splits into two halfway. In a pair of hyper-thropied digits kind of way.



''Urist McForgottenBeast is a Forgotten Beast. A three-legged quadruped made of steel. It is moving deliberately. Beware its venomous bite!