Comparing percentages like that is pretty cheesy when you stated all the information you need to calculate a paper-napkin average damage per attack round (not individual swing).
But first, your implied attack values seem totally unacceptable. Where does "1.8% DoT" come from? The typical naif (you can hammer me on this point if you want because I have no choice but to guess at your "thinking process") would take the ratio of, say, 555 and 545 attack and observe a percent change of ~1.8%. Seems like you were assuming the STR+4 Kheten had no additional attack. But if you did (being charitable here), your "1.8%" could imply attack in the sub-400s (for G. Colibri...) or a high-defense mob where melee damage is sh*t.
503 vs. 510 on lv82 Greater Colibri. Probably a little low, but at least reasonable for a baseline (the assumption being 500 attack using a GAxe with no STR+ or attack+). I used cRatio in place of pDIF because even BG doesn't have the two-handed pDIF formulas figured out yet
. If you link that thread again, you're an idiot: what it proves is that (1) the parity of your damage is twice as likely to match the parity of the mob's level vs. not matching (that is, you're twice as likely to get an odd number for your damage against mobs of odd level as you are to get an even number for your damage, and vice-versa against mobs of even level); (2) they agree that nameless is flat-out wrong; and (3) they don't have enough different cRatios to give any sort of pDIF formula that's remotely
I mean, ****, the best I managed to pull off from their Lesser Colibri testing is that there's an inflection point for pDIF minimum at or near cRatio=1.6, and the multipliers are absolutely ******* backwards (the maximum increases at half the rate, roughly, that the minimum does). I don't even trust the formulas I've got that are a half-assed rough draft for the top end of the range to be reliable because of that
Anyway... let's use a high average pDIF like 1.825 for Perdu Voulge and 1.800 for Kheten. (Suppose these account for some non-zero, sub 20% critical hit rate).
Nice assumption of 700+ attack.
DA doesn't have to cover an alleged difference of "2.6% DoT". Subtracting percentages is a complete joke. It just has to be enough to let Erlking's Kheten surpass Perdu Voulge when the whole accuracy +5 still matters, and it does for my contrived values, with pDIF being the only questionable factors.
It's not subtracting percentages (though with figures in the single-digit percentage range and my habit of normally giving two significant figures, there's no noticeable difference between adding/subtracting and multiplying/dividing the majority of the time).
The math as I did it (the 38/37 is .95/.925, multiplying both the numerator and denominator by 40):
38/37 * (510/327 - 0.35)/(503/327 - 0.35) * 105/107 * 1.2/(1.2 + DA) <= 1
Solving for DA, we get 3.118% or higher.
Bumping the base attack to 600 (accounting for food and switching from double March to March/Min, I suppose):
38/37 * (610/327 - 0.35)/(603/327 - 0.35) * 105/107 * 1.2/(1.2 + DA) <= 1
And now DA needs to be 2.672% or higher.
But all this is besides the point when the only conclusion I gave was that, once again, DA+2% doesn't cut it
, which your figures agreed with.