Bugzors: changing the skill weighting down to 0.00 in the options, and then changing it back to 1.00(or any number actually) has zero effect. DT requires a restart for it to start working again. You can set it down to 0.01 and changing it still works.
And something must be wrong with how DT values a skill, 2 examples:
dorf1 with just 500xp in gem setting is listed as being 97.79% fit for the job, without any skill weighting he is 45.23% fit.
dorf2 with just 1500xp in gem setting is listed as being 100% fit for the job, without any skill weighting he is 43.89% fit.
If you took out skill weights, then it wouldn't matter who had higher lower skills, it would then base the role on attributes/traits.
I wouldn't be surprised to see that having just set 1 gem(from 0xp) would change dorf3's fitness for the role a fair amount, i'll try and test that today or tomorrow.
I would hope to expect a skill level of 7-10 to start to out weight attributes in the calculations.
you can't use these assumptions. As it's based on what the average xp value is for all dwarf's (i.e. the mean), depending on what your min value is (most likely 0), and your max (whatever your max dwarf is at, say 1500), then deviations from the mean would start to throw %'s off, say your mean was 750.
Edit: Quick test i did with dorf3, odd results.
At 0.01 skill weight 1 gem set took him from 49.36% to 49.71%
At 1.00 skill weight 1 gem set took him from 43.80% to 43.84% O.o
This makes zero sense, a far far far lower skill weight = a far higher fitness% increase? O.o
I don't get this last part. I get the weights, but what do you mean by 1 gem? 1 gem experience? If it's 1 gem experience, then your deviation from mean barely changed, (ex, mean is 750, 1 exp vs 0 exp to 750 is nothing). These numbers themselves would produce z-scores, and then these z-scores would be multiplied by either .01 or 1.
Example: A 1 gem (i.e. a 1 exp?... vs a mean somewhere around 750), would yield a deviation from mean of -749, converted to some negative z-score. Which would be multiplied by the weight. The .01 weight would minimize the z-score to .01 it's value, it also means the stdev of that dataset would also be reduced to .01 before being combined with the other standard deviations. In the end, the .01 weight would mean that the deviation from mean would contribute very little to the overall %, meaning attributes and skills would have a higher weight. The 1 weight would retain the same zscore, and standard deviation, contributing more to the overall %. Which means skills would contribute more to the overall %.
It's hard to see without using a spreadsheet using the formula vs your results, and known mean's, sdev's (which could be done as v9 output's means and sdev's I believe when you hover the mouse over specific instances of attributes, skills, traits).
What I can do is use pastebin to paste the formula that Splinterz used as his basis if you want to follow along the calculations, but I'll have to do that later tonight...
Here you guys go!
http://pastebin.com/v71VrBfMUpdate:
I am not able to spot check against a spreadsheet as the stdev' and mean's are no longer listed in (v9), they were in v8... oh well. I was hoping to post a link to a mediafire spreadsheet showing the calculations...