Drain II/Drain III Testing

So, I went out and did some testing on Drain II and Drain III. I tested on Plateau Hares in Abyssea La-Thiene due to them having 0 MDT, with 0 Drain/Aspir Potency, 0 Dark Affinity, and on days other than Darksday/Lightsday. The results are sorted from highest to lowest with the bracket for easier reading.

Drain II, 460 Dark Magic Skill
616
611
605
604
603
601
592
587
565
563
563
557
551
548
543
541
540
532
523
516
507
502
501
491
481
460
443
435
434
430

Drain II, 520 Dark Magic Skill
684
679
678
677
673
672
668
663
663
657
652
645
616
606
596
594
593
583
570
560
540
532
524
511
505
501
500
498
493
491

Drain III, 460 Dark Magic Skill
843
835
828
819
813
805
799
796
778
778
767
762
751
728
726
723
719
707
701
693
690
684
678
677
669
657
656
651
647
646

Drain III, 520 Dark Magic Skill
935
928
925
924
918
908
885
882
881
875
874
868
868
850
842
820
816
811
793
775
766
757
755
754
753
752
738
734
732
708

From these results I can make the following conclusions:

At Dark Magic Skill levels relevant to knowing Drain II and Drain III, the base potencies are (plus or minus a couple points due to small test samples):

Drain II: Dark Magic Skill + 165
Drain III: Dark Magic Skill * 3/2 + 105

It is likely that they have different formulas at much lower Dark Magic Skill values (like with Drain's formula), but as I have no way of testing it (and it's not useful information since Dark Knights should always have high enough Dark Magic Skill for it to not matter), they are not relevant to this test.

Drain II has a random variance between the base value and ~70% of the base value (might be 2/3s, more testing samples would be needed).

Drain III has a random variance between the base value and 75% of the base value.

Bg-wiki lists the hares as lvl 75-90 which I find hard to believe but I wasn't around for Abyssea... Do you have a way of knowing you were in an area with all the same lvl or could lvl differences explain some of the variance

Only asking because maybe some attribute "dif" factor is involved idk

Mob levels have no effect on MDT, and the MEVA differences aren't nearly high enough to affect an ilvl 119 player's land rate.

Drain spells just have a very large natural variance; Drain I's variance is even larger than these two's, going as low as 50% of the base value.

Just as an addendum, I wanted to see if INT affected magic accuracy for Dark Magic. Under the assumption that INT likely affects all of them the same, I used Bio II for testing. I tested on level 123 Quarreling Hippogryphs, and made sure my magic accuracy and magic accuracy skill stats were exactly the same for each test, with values adjusted so that the baseline test had low, but not floored hit rates.

100 INT (247 samples)
1/8 Resist: 75 (30.364%)
1/4 Resist: 37 (14.980%)
1/2 Resist: 52 (21.053%)
No Resist: 83 (33.603%)

155 INT (277 samples)
1/8 Resist: 80 (28.881%)
1/4 Resist: 44 (15.884%)
1/2 Resist: 78 (28.158%)
No Resist: 75 (27.076%)

221 INT (246 samples)
1/8 Resist: 58 (23.577%)
1/4 Resist: 40 (16.260%)
1/2 Resist: 52 (21.138%)
No Resist: 96 (39.024%)

It does look like there is a minor difference between the resist rates in the three samples. However, the sample sizes are on the low side, and one would expect a huge 121 INT swing to have a FAR larger effect on resist rates if INT affected Dark Magic magic accuracy, based on dINT values for black magic and dMND values for white magic.

My tentative conclusion is that INT doesn't affect Dark Magic magic accuracy, but I may have to do another test at some point with much higher INT values, to make sure there isn't some kind of dINT flooring going on.

Thanks for working this out.

INT Testing, Part 2

266 INT (337 samples)
1/8 Resist: 24 (7.122%)
1/4 Resist: 40 (11.869%)
1/2 Resist: 77 (22.849%)
No Resist: 196 (58.160%)

321 INT (315 samples)
1/8 Resist: 4 (1.270%)
1/4 Resist: 16 (5.079%)
1/2 Resist: 57 (18.095%)
No Resist: 238 (75.556%)

So, it seems that INT does affect the magic accuracy of Dark Magic (or at least Bio, but I'm willing to assume it follows for the others as well). It seems that the previous samples were encountering a dINT floor due to my INT being too low.