I had a look at what numbers we've got so far, and I think I'm fairly clear on how the armour scaling works now. Let's summarise it and see if anyone disagrees.
1. Roll for the base value of the stat.
For example, Gossamer INT bonus is in the range (1,10). Whole number? I suspect so - I'll explain below.
2. Roll for whatever boosts are given by the equipment suffix. I'm guessing this is done after rolling all the stats, not that it makes any difference.
For example, "the elemental damage bonus from Surth/Niflheim/Mjolnir/Freyr/Heimdall/Fenrir staff/armor [is] (10,15)". Add bonus(es) to the relevant stat(s). (Again, that's a guess, but it seems like the most logical way of doing it.) Note that this suggests that a specific suffix always gives the same range of stat boosts, regardless of what equipment type it applies to - but that's just a theory.
3. Multiply by the slot scaling factor. These appear to be 1.0/1.2/0.9/1.1/0.8 for head/body/hands/legs/feet. Round to the nearest whole number.
4. Apply the global "level 0" and level scaling factors for each stat to give the actual numbers for each piece of equipment.
Why do I think the base stat roll is a whole number? Well, let's take Gossamer INT bonus on robes. If we take the whole numbers from 1 to 10 and multiply by 1.2, we get
1.2 2.4 3.6 4.8 6.0 7.2 8.4 9.6 10.8 12.0
Rounding these gives
1 2 _ 4 5 6 7 8 _ 10 11 12
Note the numbers that don't appear: 3 and 9. And neither varst nor I had a robe with those base values (all the others were covered except for 12). Those correspond to "base" (i.e. level 0) values of 0.3 and 0.9 btw, in case anyone wants to look for a robe that disproves this theory.
Some examples from the data collected so far. I've removed the L300 numbers, looked at a number of pieces to determine the level 0 multiplier for each stat and derived the raw base values.
Elemental Damage Bonus (EDB)Old values from when the maximum was known to be 10.
Level 0 multiplier is 0.299.
Head: (2.99)
10Body: (3.59)
12Gloves: (2.69)
9Legs: (3.29)
11Feet: (2.39)
8This is a pretty good indicator that the multipliers are 1.0/1.2/0.9/1.1/0.8, though it doesn't prove those are the exact values.
ProficiencyLevel 0 multiplier is 0.305.
Head: (6.12)
20 Body: (7.34)
24Gloves: (5.51)
18Legs: (6.73)
22Feet: (4.90)
16Better resolution than EDB, and therefore a better confirmation of those multipliers The body multiplier,for example, has to be in the range 1.175 to 1.225, and we might as well assume they aren't weird numbers unless we have evidence to the contrary.
Phase Evade ("Preliminary.")
Level 0 multiplier is 0.25.
Head: (4.00)
16Body: (4.75)
19Gloves: (3.5)
14Legs: (4.50)
18Feet: (3.25)
13Are these uneven step sizes due to incomplete data? Not necessarily. Taking 16 as the base, and multiplying by the various factors, we get
12.8 14.4 16.0 17.6 19.2
Round those numbers, and we have the observed values.
Everything else seems to fit too, at least where we have sufficient data. The Fleet Silk numbers don't make much sense at the moment, but they are labelled "extremely tentative", so...