Print

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

[Shandra]

Sutremaine... mh, sounds reasonable that they are equal in their interpretation (and that appearance ranges finally are of gaussian type, whereas the attribute ranges are not). Damned, that would mean the percentiles for position 2/3/6/7 beeing reals and not decimals - well, just somewhat unusual. So the percentiles used for the graph would be 0/16.66../33.33../50/66.66../83.33../100. Next interesting thing would be to know in wich percentile-ranges the descriptors for the attribute value will fall. 8 named ranges, plus the non-descriped range around the median seem to me that they are either 11.11...% each, or should we assume a somewhat higher % range for the non-described/normal value?

Spoiler: Example Graph (click to show/hide)

And the trouble with the descriptions is with the cases where the Quantiles are of equal value, always wondered wich one will be used in that case - ok, sounds like some testing with min/max is needed here.
Edit: ok, tried with all zeros and 5000 - and I can't recognize any pattern. So it's either randomly picked, or there are some other parameters for the descripions I miss so far.

[Sutremaine]

Edit: ok, tried with all zeros and 5000 - and I can't recognize any pattern. So it's either randomly picked, or there are some other parameters for the descripions I miss so far.

Is there anything in the error log? The string dump contains these entries (pasted from wiki):

Attribute range not in increasing order, repaired
Attribute range less than
, repaired
Attribute range greater than

Seems odd that these would exist given that there are vanilla creatures with max stats across the board, though.

I've done a little bit of messing around with attributes and modifiers, but just enough to know that a lot of them are linked and some of them are relative values and not absolutes.

[Shandra]

Nope, nothing in the gamelog.txt, except for the creation of the dummy region and no errorlog.txt created. And for that string_dumbs I would guess they mean a) case 0:200:100:300,etc., b) less then - and c) more then 7 parameters provided

For the descriptions I assumed from the description within creatures_standart.txt that they are based on the range-distribution for that particular race, but how they are distributed along the quantiles I have no idea.

Spoiler: note in creatures_*.txt (click to show/hide)

   Attributes for dwarves are still described in terms of the median value below, but the actual game effects are altered according to the raw numbers.

I've updated the linked image in my previous post to include an assumed average and modified the whole diagram to be more readable in general. The assumed default values are (pure choice of mine in regards to the neighboring graphs and not based on any facts: 250/700/900/1000/1100/1100/1900)

[knaveofstaves]

topline: AVG is [200:750:900:1000:1100:1300:2000]. The 750 is bizarre to me as well, but I've got 8960 data points backing it up.

These seven numbers are the fenceposts for six ranges. A creature is assigned to a range, then with equal probability to any value in the range. So racial median and racial mean are different.

at ++, racial median is 1500, racial mean is 1516.66...
at +, racial median is 1250, racial mean is 1266.66...
at avg, racial median is 1000, racial mean is 1025
at -, racial median is 900, racial mean is 862.5 (!!)

In other words the patterns of equal steps are a bit of a red herring, not systematically used but used often, such as in how attributes are verbally evaluated under Thoughts and Preferences. Reference spoilered to ameliorate digression:

Spoiler (click to show/hide)

++ (Focus, Spatial):
0-542: -4
543-792: -3
793-1042: -2
1043-1292: -1
1293-1791: 0
1792-2041: +1
2042-2291: +2
2292-2541: +3
2542+: +4

+ (Str, Tough, Analyt, Creat, Pat, Mem):
0-250: -4
251-500: -3
501-750: -2
751-1000: -1
1001-1499: 0
1500-1749: +1
1750-1999: +2
2000-2249: +3
2250+: +4

AVG (All others except Agi):
0: -4
1-250: -3
251-500: -2
501-750: -1
751-1249: 0
1250-1499: +1
1500-1749: +2
1750-1999: +3
2000+: +4

- (Agi)
NULL: -4
0-150: -3
151-400: -2
401-650: -1
651-1149: 0
1150-1399: +1
1400-1649: +2
1650-1899: +3
1900+: +4

Open question: Why did Toady break pattern and offset his verbal evals of ++ by 42? Douglas Adams callout?

I'm not a modder as such -- I just want to understand how default dorfs work -- so this project is half-finished. This is what I did if anybody wants to see what other settings do: Dwarf Fortress instantiated by the More Embark Dwarves modified Dwarf Fortress 224 app in version 0.31.21. Runesmith 0.1.9 used while DF is on the Prepare for the journey carefully screen, and dwarf data exported to CSV and manipulated in OpenOffice Calc 3.3. For the above I did this four times for a sample size of 896. I estimated with the =percentile() function, got an idea how -,+,++ work, and applied that idea to the avg data sets.

Regrettably Runesmith exports traits verbally, so hard to work with and having no clear value for dwarves in the 40-60 range, but based on this I'll say half of dwarves gen with (e.g.) Immoderation of over 55, and half under 55. (No actual idea whether the fenceposts themselves are in one side, both, neither...)

And now the plug! I'm doing all this because I'm writing a little JavaScript/HTML script that will, hopefully, one day, understand all attributes and traits, and tell you what job a dwarf should do with them. It's just a proof of concept right now, but it does understand a handful of traits (among them Cooperation, Self-Discipline, and Straightforwardness). I call it Dwarven Guidance Counselor, and any input is welcome.