Print

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

[Slitherrr]

Ah, thanks, I had completely missed it. It does work perfectly, cheers to that.

[Raven]

it doesn't work

it says "I don't know how to talk to this version of DF.
checksum:0x4f7592fe"

I tried creating an .ini copy of 34.05 (the last one I have) but it doesn't work either...

whut?

[slink]

9 posts up is the complete file.  Enjoy.   

[thistleknot]

much thanks to Thuvian, we brainstormed a way to not have to use frequency distributions, now we can get exact % by taking (instance-mean)/range = a perfect of ~-50%/~+50%, which will also work for skewed distributions!  the distribution is the exact distribution of the sample!  no guessing with frequency distributions!

top is mean, bottom is median

Spoiler (click to show/hide)

I was afraid originally that it was going to give a flat distribution curve, but after a lot of thinking, I realized it wouldn't and then graphed it.  It's also a perfect histogram.  This is an example of agility.  and I did abs(cell)-.5 to get the results, then flipped it.  I'll have to figure the exact details out

[Kon]

Thanks, slink!

[dreiche2]

much thanks to Thuvian, we brainstormed a way to not have to use frequency distributions, now we can get exact % by taking (instance-mean)/range = a perfect of ~-50%/~+50%, which will also work for skewed distributions!  the distribution is the exact distribution of the sample!  no guessing with frequency distributions!

Sorry, that was a bit difficult to comprehend. Is "range" relative to the population or the absolute range of the aspects? What's "a perfect of ~-50%/~+50%" supposed to mean?

[Raven]

Thanks, slink!

yeah thanks a lot

it seems that I tried to paste another code and the right one was the one in your post

I suggest placing that stuff in the open post until the next version

[slink]

Er, well, if you mean the opening post of this thread, I can't.  It belongs to DwarfEngineer and if he was available you wouldn't need my post.   

[miauw62]

Thanks slink!
(oh god, this is getting horribly repetive)
Anywayz, have a nice vacation (or whatever) dwarfengineer, and thanks slink so we can play df without bashing our heads against the walls.

[i2amroy]

The OS X values are up, sweet! This way I can actually get some play time in rather then wasting all of my weekend on more graphically (but less awesomely) advanced games.

[maseck]

So, that lua scripting I was talking about, I actually got it to do something useful. Given a file with the following code, only dwarves who are the most skilled in a skill are shown.

Spoiler (click to show/hide)

Code: [Select]

local included_dorfs = {}
local dorfs = dwarf.get_dwarf_array()
local n = 1
for labor in pairs(labor_to_skill) do
local skill = labor_to_skill[labor]
local best_dorf = -1
local best_val = -1
for index, dorf in ipairs(dorfs) do
local val = dwarf.get_skill_exp(dorf, skill)
if val > best_val then
best_dorf = dorf
best_val = val
end
end
if best_dorf ~= -1 then
included_dorfs[n] = best_dorf
n = n + 1
end
end
dwarf.ui_only_show(included_dorfs)

I was discouraged a little bit when I noticed that Qt Creator has a bit of a script engine built in. It makes me feel unsure if any of my changes will be considered for implementation.

[shovelmonkey]

Thank you, slink, now works like a charm. My weekend is saved!

[Edv1819]

Thank you slink for the values

[Blaises]

I think I found the layout for Linux.  It's my first time finding the offsets for DF so let me know if it works or not:

[info]
checksum=0x61b4fa72
version_name=v0.34.07
complete=true

[addresses]
translation_vector=0x9653ee0
language_vector=0x09653ec8
creature_vector=0x09612b8c
dwarf_race_index=0x095fa500
squad_vector=0x9650a94
current_year=0x095f8000

[thistleknot]

much thanks to Thuvian, we brainstormed a way to not have to use frequency distributions, now we can get exact % by taking (instance-mean)/range = a perfect of ~-50%/~+50%, which will also work for skewed distributions!  the distribution is the exact distribution of the sample!  no guessing with frequency distributions!

Sorry, that was a bit difficult to comprehend. Is "range" relative to the population or the absolute range of the aspects? What's "a perfect of ~-50%/~+50%" supposed to mean?

It could go either way, I'm going to calculate it from the population (which I know some of you guys hate).  For two reasons.

One is only traits are static

skills and attributes change, attributes are trained by skills, and vice versa, and player strategy can result in different ranges for different players.

This new approach should account for skew bias (without the need for fancy distribution frequencies which are not as accurate anyways).

And instead of ~-50%/~+50%, I'm gonna calculate a 0-50% and a 50-100%, 0 representing population minimum, and 50% representing median, and 100% representing population max.

Lastly,
Skills will now be based on a skill (non 0) median.

The same concept can be applied using averages.  However, I feel median's are more appropriate for "skewed" data.

Perfect meant that if you added the - % to the + % (if they were absolute values), they would equal 100%.  but... I decided that's not good, even though a max may be 53% above average, that 3% will bias that scale towards combining it with other attributes... So I scaled them down from median to max and median to min by dividing by median-instance by (median-min) and (max-median)

Update:

What I could do Dreiche2,

Is use the median from the wiki (or mean), and then take the min/max I have already for the various stat's (as taken from a sample of 3300 dwarves) and increase/decrease them by 10% for error, and that should be good.  There you go, everyone's happy.  Min's/max's/median's/mean's are all static.  I would still let skills be determined dynamically, but attributes/traits could use a static/means setup.  I would need to collect some data on traits though.

I was considering modifying the formula to say that anything between the median/mean was 50% but I'm not really all for that, but it would bridge the gap between median/mean.

Currently the min's max's I have are:

Code: [Select]

          -                +       ++
Min's   146   198    444     695
Max's  1595   2108   2250  2537