Print

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

[Bumber]

How about observer skill? It's used for detecting incoming attacks.

I thought it was used to detect things which were sneaking.

That too, but since the combat rewrite that added manual blocking it has a new purpose. It gives info about attack types, and when and where they're going to land (in adventure mode.) I suspect it might affect parry rate for your dwarves.

[Admiral Obvious]

How about observer skill? It's used for detecting incoming attacks.

I thought it was used to detect things which were sneaking.

That too, but since the combat rewrite that added manual blocking it has a new purpose. It gives info about attack types, and when and where they're going to land (in adventure mode.) I suspect it might affect parry rate for your dwarves.

I might be able to give it a wing some time in the next two days on the most recent build. Sounds interesting at least.

[kontako]

Could you upload your graphs to the df wiki? (If you're confident in what they represent)
I'm sure whoever is reading about military on the wiki is looking for exactly this!

Links to pages:

Spoiler (click to show/hide)

http://dwarffortresswiki.org/index.php/DF2014:Combat_skill
http://dwarffortresswiki.org/index.php/DF2014:Armor

[MiguelT]

Ptw.

[Magistrum]

Well, that is some interesting stuff.

Giving miscellaneous object user skill to shield users might tip the scales a bit, seconding the dual wielding tests too.

[Cathar]

Mind explaining in a bit more detail how you tested it? Did you pit one dwarf wearing a set of arms and armour against subsequent goblins until the dwarf died, while having another dwarf in a different set of arms and armour through the same goblins? How well armoured were the goblins, stats etc?

Sure thing :

A Lethal Dose 50 is considered to be the most reliable test when it comes to the dangerosity of chemical substances, however it is frown upon in the scientific community due to ethic concern, which we need not to worry about in DF.

The set up is the following : you take a population of rats or mice (100 let's say) and increase the dose of the tested chemical until it exactly kills 50% of them. Then you note the dose you have to come to to achieve that result, and it gives you the LD50 value of that chemical.

So I conducted a ...reversed ? LD50 test with a standardized goblin (competent in all combat related skills and clad in copper).

I first took 10 dwarves armored in copper, single wielding+shield and conducted tests with increasing numbers of goblins until I could reliably kill 50% of the dwarves, then I repeated the process with dual wielding.

This test has many advantages compared to simple pitting, the two obvious ones are 1°) it tests the dwarves under similar conditions and 2°) the testing conditions are closer to actual field use of the tested gears

The test result gave a slight edge to dual wielders, but it was a very close call. It is not as overpowered as people seem to think - it is however viable as a strategy.

TL;DR : "How many standard goblin do you need to kill 50% of a population of X ?" is the question my test aims to answer, basically

[Fleeting Frames]

50% of population rather than 50% of groups is bit more troublesome because of snowball effect.

But besides that, you shouldn't test dwarves against group of enemies because of martial trances (use two groups of goblins instead). To illustrate: Suppose you set up 10 dwarves against 10 dwarves, and 1 from group A gets early martial trance - then kills 3 from group B during a trance. Now there's 3 dwarves who can make 3 more fights 2 on 1, likely cinching them. If the remaining 4 fights go equally, you get that group A wins with only 2 fatalities, leading them to look 5 times superior to naive eye even though they're actually equal.


You're absolutely correct about my testing counting only for specific duel scenarios, though. While gobbos do come in with shields where possible in fort mode, the case is usually some dwarves versus many lower-skilled gobbos - which means flanking and the like can come into play. And of course, regular rocs are lower skill as well.


I think a possible "downfall" of massive tests might be legendary dwarves being pretty much immune to losing limbs until they tire themselves out, but that's only a "downfall" in the sense of ability to distinguish the groups rather than finding anything.

[Cathar]

The snowball martial transe effect was indeed a critic that was mentionned on my initial tests, however I don't think it does invalidate the test, as martial transe are part of the gameplay, and tests conducted could determine a reliable LD50 despite it.
However if you assume martial transe snowball is a factor that should eliminated from testing (why not, after all) you can replace dwarves with goblins alltogether, as it is the material, not the wearer that is being tested. I will do precisely that for subsequent tests, I am very confident the results should match more or less the same.

Another solution could consist of a battery of tests and use aggregate data to eliminate oulying results

That's the good thing with those tests, they are easily replicable. I will do them again after work and share the results

[Fleeting Frames]

Indeed, changing all dwarves to goblins is just single line in dfhack after all if you saved the setup. (Though I guess they'd keep dwarven attributes.)