Print

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

[Uthimienure]

Is there a way in DT to sort by or display the level of cave adaptation?

[Clément]

By creating a custom grid view (either from scratch or from a copy of an existing one), add a column "Trait" → "Cave Adaptation".

[Uthimienure]

By creating a custom grid view (either from scratch or from a copy of an existing one), add a column "Trait" → "Cave Adaptation".

Thanks!  It worked 

[thistleknot]

I seem to vaguely recall discovering a problem with the way nurse is calculated

It doesn't make sense it's all in red

https://imgur.com/gallery/oyhn5V9

yeah, this post

http://www.bay12forums.com/smf/index.php?topic=122968.msg7306290#msg7306290

dwarf.cpp

2962 to 3126

I'm trying to figure out how to open a ticket in github.

The issue seems to be there is no check if the role's aspect has no members. If so.  The aspect should be excluded from the averaging (weighted sum).

Edit
It's definitely the skills.  I looked through the logic and the logic is coded in there.  I'm in the works of debugging it.  If I remove the skills section altogether, the roles look "normal" for nurse.  But if I do by either forcing 50 or not.  It's either all red or all black.  I'm thinking the aspect isn't reporting empty for some reason.

[Clément]

I don't have any issue with nurse. Are you sure you don't have a broken override?

[thistleknot]

Hmm. Idk

Is your lawmaker good too?

[Clément]

Do you mean "Lawdwarf (Weak)"? I don't see anything wrong with this one either.

[thistleknot]

Sorry early. Ye lawdwarf. I'm using a build w mwmod. So maybe thats it. I'll check against vanilla

[thistleknot]

While I'm checking it out (I did see skill at 0 confirming my concern, but still haven't fired up vanilla and/or downloaded new .ini files).

The way the weighted sum calculates right now is if an "aspect" doesn't have any elements in the vector, it returns 50% which doesn't harm or improve the overall score.

This was chosen quickly because we thought it made sense, but now I see how it penalizes the score

Say I have a role with all aspects defined.  Then that role potentially could have scores ranging from 0 to 100%

Now say I have a role that is attributes only and the rest are 50% (7)

This means that if I have 100% in Attributes, that 100% will be brought down to 57% ((6*50%+100%)/7)

Instead, it makes more sense to let the role be defined by the score of the attributes itself without averaging in the "null" elements.  Instead of .50f being stored for m_role->[aspect].empty(), a null should be stored and the end average should be done over non null values.

This is what I have for vanilla using a fresh build.  I'm not sure what an override is.  I reset all settings and the nurse is still getting 0

Spoiler (click to show/hide)

Code: [Select]

2020-Aug-23 05:54:43.458 VERBOSE core   + "Nurse" - "Mebzuth Zal?slogem" [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3023] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -attributes: 58.2721 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3131] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -skills: 0 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3132] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -facets: 50 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3133] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -beliefs: 50 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3134] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -goals: 50 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3135] (calc_role_rating)
2020-Aug-23 05:54:43.458 VERBOSE core       -needs: 50 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3136] (calc_role_rating)
2020-Aug-23 05:54:43.459 VERBOSE core       -preferences: 50 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3137] (calc_role_rating)
2020-Aug-23 05:54:43.459 VERBOSE core       -total: 26.034 [C:\Users\user\Documents\GitHub\Dwarf-Therapist\src\dwarf.cpp:3138] (calc_role_rating)

[Clément]

If you are comparing dwarves for a given role, it does not matter. If you are comparing roles for a given dwarf, yes it is important. But I am not sure what is the best solution. If you want to have each role use the best of the 0-100% range, I think it would be best to apply role ratings normalization per-role, instead of globally. Then applying the empty aspects ratings or not is simply a linear transform and should not affect the result of the normalization. This brings the question: which normalization method? I currently have three: Stratified MAD (+ factors), ECDF, double linear transform (around median).

[thistleknot]

Stratified MAD (+ factors) should be capable of replacing everything. 

As to per role, I compare across roles.  I see what your point is if your doing it job by job, but that was something I tried to overcome (I was doing that in spreadsheets way back) but the point of the labor optimizer was to start from a single list of %'s and start from the highest and work down (across roles)

[thistleknot]

I don't get it... for one my breakpoint didn't stop, but two it looks normal now...

https://imgur.com/a/xNF7zTv

[thistleknot]

I reran some comparisons. I *think* it has something to do with the version of mwmod I'm running. I thought I checked vanilla earlier and saw the same behaviour. It could still be an edge case bug, but rechecking the exact same build against mw and vanilla shows no issue with vanilla (at least in terms of the original issue, low role scores for nurse and lawdwarf).

[thistleknot]

I did a really cheap hack to work around my aspect concerns.

Here is a screenshot showcasing the difference.

Spoiler (click to show/hide)

Code: [Select]

for (int i = 0; i < AspectCount; ++i)
        if(rating_aspect[i] == 50.0f)
        {
            //exclude
            global_aspect_weight_total = global_aspect_weight_total - global_aspect_weight[i];

        }
        else{
        rating_total += rating_aspect[i] * global_aspect_weight[i];
        }
    rating_total /= global_aspect_weight_total;


https://imgur.com/gallery/iJOtZIQ

that seems to solve my problems.

I still get lower than normal nurse scores but they are much better, in the 40% range.  Looking at the log (using --trace) I see they have higher scores than what is drawn, but I believe that's because the final scores are ran through another normalization method at the end and these scores are on the lower end of other scores.

*In my log it no longer shows skills reporting at 0% but 50% and I exclude that from the weighted sum with the above code [reasoning: knowing that's hard coded for missing aspects]

Suggested workaround until a better solution:
https://github.com/cvuchener/Dwarf-Therapist/compare/rolecalc...thistleknot:rolecalc

[thistleknot]

Clement

I forget how skill rust was calculated into role ratings if at all.  Do you know?  Is it counted at all?  Is there a penalty applied (like potential) for skills that are experiencing rust?