Notes:

I'm actually working on a method to make this even better!

Since I figured how to deal with a skew, I pretty much figured out how to rotate a wave around either it's mean or median from a scale of 1 to 100% with a desired average of .5 (or at least drawing closer to it).

What this will allow is for an older fort, who has higher trained values than lower trained [attribute] values be counted appropriately. In other words, the distance in values in themselves [especially on the higher end] will be better accounted for and sticking true to the original shape of the data, whilst maintaining a ~.5 mean and a range from ~0 and ~100%, preferably split around the median. It's possible to do two transforms though. First a transform around the mean, then a transform around the median of those results. I have a few pictures, but I'm still working on it

Splinterz has given me the okay to submit this

Quote from Splinterz on some of his changes

"what you'll notice is that the new roles can be compared against each other very easily at a glance. you can see immediately who would be your best pick for which jobs based on the roles.

this also ties into the optimizer, as now the priorities you set will be upheld, as long as you have a decent spread among the priorities (explained in the GUI). so if you want to ensure certain jobs are chosen before others, you can.

the other feature visible in the new role view are the shaded cells which indicate if at least one labor associated with the role is currently enabled, and yes, you can toggle them on and off.

[...] the equipment [...] will show exactly how much ammo is missing"

Alpha testing of 15 different versions now, but now the formula is working as we want, just cramming new features in at this point.

The roles are directly comparable, and all roles averaged = 50% mean (barring any mods where dwarf's can't gain any skills, such as priest castes in mw. If that is the case, the dwarf is listed as 0% for roles due to skill limitations).

Please report any issues on the forum.

before roles:
http://i.imgur.com/gsUCLdt.png

after roles:
http://imgur.com/d9YTv6H&Uoxe4Q5#1

Here's a gridview of just skills:

http://imgur.com/qwU2ynS

Notice all those high skills? That's due to cultivating them using the labor optimizer and the new role calculation methods (valuing skill with the most weight is what's important). Also notice how happy all the dwarfs are? That's due to preference matching

Think that's impressive?

Here's how my military looks role comparison wise.

http://imgur.com/CHOhFZc

An indepth explanation of how we derive %'s for each category (aka attributes vs skills vs traits) can be found here:
http://www.bay12forums.com/smf/index.php?topic=122968.msg5410773#msg5410773

but more or less, we use a rank function to derive a % by dividing by the count of elements within a category. We resolve rank ties by returning the center rank position of ties vs first element of ties (standard rank behavior) or last position (standard ecdf behavior).

In Progress:
We're working on two new changes that will more evenly spread out skills as well as preference %'s. As is, there is a large jump from 50% to the set of %'s used for skills. I did a simple min/max transform around the upper values (aka >Median) to 50% to 100% respectively, and then factored UP the <=median values to achieve a [max values] less than 50%, but also an overall output mean of ~50% (wasn't easy). - implemented

updated:
fixed issue with skills being poorly accounted for due to their exponential curve.

This will allow more relevant skill comparisons, as it will reduce them to a scale of 50% to 100% for relevant skills (aka > median) vs ~95% to 100%, which will be more in line with how attributes and traits are scaled.

To use with MW, copy of the memory layout as well as binary to a masterwork installed version.

I also recommend replacing the default .ini for mw roles with the one found in this thread
http://www.bay12forums.com/smf/index.php?topic=132010.msg5374573#msg5374573

Check out the homepage for a description on how to setup autolabor to manage hauling.