Shadow (warrior) skill slvl

[Crystalion]

I've noted oddities with Shadow skill slvls (e.g. observed #shots synergy on a trap after increasing my test chars skill in the synergy). Others have posted questions about shadow slvls and still others have given various answers.

Code reading would be helpful, but would still need to be verified by testing, when possible.

Testing is a pain, in this case.

But I had an idea for a test earlier that seemed pretty straightforward.

v1.10 lan pvp test, modded skills.txt, monstats.txt

Concept: have a hostile sorc deliver fire, cold or lightning damage of known quantity vs. a fixed HPs shadow warrior that is using some unknown level of Fade as her only source of resist and observe how many damage units it takes to kill the shadow (whose regen has been disabled).

I probably should have modded just ice blast (convenience of freezing shadow in place) but I instead made fire bolt, lightning and ice bolt into a constant 10 damage (lightning required eliminating a synergy also, and I noticed surprising damage min/max/level parameters, other than emin/emax, which I also zeroed, but need to remember to play with at some point, as I don't grok their purpose).

I turned off the Warrior regen, fixed her HPs to always be 100, and changed Fade to have linear 0% at slvl 1 plus 10% resist per slvl.

The test sorc' and 'sin were -act 5 clvl 33s, testing in blood moor, just outside town. The Assassin helpfully goes hostile and turns it off, so the shadow can be attacked (but doesn't attack). This trick can also be used to force a new shadow to turn on the Fade aura (on right mouse for the 'sin).

With and without Fade the sorc' also attacked the 'sin 6 times, with each element, checking the damage delivered, to verify that I had correctly gotten them all to a constant 10. After this tests were conducted using this constant damage to verify that I had made the Warrior have a constant 100 HPs, regardless of slvl cast at.

I also verified that the 10% per slvl on Fade "worked" by pvp vs. the 'sin.

So here's the raw test data, in format fade/sw/#hits (fade slvl, shadow warrior slvl, #hits taken to kill shadow):

Code:

none, 1 and 5, 10 (verification of 100 HPs constant before and during the other testing...)
 1,  1,  10  (0%)
 5,  1,  12  (shadow almost dead on 11th, according to LCS(TM), so probably 10%)
 5,  5,  13  (20%)
 6,  6,  17  (40%)
13, 6,  34  (70%)
13, 7,  34
14, 7,  50  (80%)
14, 8,  50
15, 8,  50
16, 8, 100 (90%)
16, 9, mu  (100% i.e. 'sin immune.. btw, with a poisoned dagger I confirmed immune to poison also)

Someone else should probably verify this or at least my "fit" for the data, but here's what it looks like to me...
(bear in mind that I can goof in my experimental procedures, or even transcription, etc.)

slvl 1 should be 0% resist, slvl 10 should be immunity. It looks like 5/1 might have been slvl 2 (almost dead on 11th hit) and 16/9 is slvl 10 with the other points in between. Clearly boosting either Fade (*my* skill points in the skill the warrior is copying) or boosting the clvl for casting the warrior has an effect, if you do it enough.

I'm assuming that either the 1..8 slvl warrior doesn't ever get resists on their equipment, or I just didn't get any in these tests. Note that my clvl 33 'sin was "naked" in the sense of only having the initial katar/shield from character creation (iirc someone said the warrior copies your base armor types on some slots).

The implied slvl of Fade, from my % comments above...

Code:

 1,  1,   1
 5,  1,   2
 5,  5,   3
 6,  6,   5
13, 6,   8
13, 7,   8
14, 7,   9
14, 8,   9
15, 8,   9
16, 8, 10
16, 9, 11

For this limited dataset I see bumps in slvl output from Fade input only on even transitions, but I see both odd and even transitions for SW. If I guess, therefore, that my Fade is div 2 and my SW is div 3, and then truncate, etc., so a spot check is...

int(16/2) + int(9/3) = 11
int(16/2) + int(8/3) = 10
int(15/2) + int(8/3) = 9
int(13/2) + int(7/3) = 8
int(6/2) + int(6/3) = 5
int(5/2) + int(5/3) = 3
int(5/2) + int(1/3) = 2
int(1/2) + int(1/3) = 0 (and there is probably a minimum of 1 check)

Seems likely to be a fit!

Whether this is correct, and if so, applies to all shadow warrior skills, synergies, and shadow master skills, is rather open to question. Still, it is a reasonable theory, and can be used to make predictions and do more testing (or code reading, I suppose).

If it is true, then your level in a skill is more beneficial than your level in warrior (and, presumably, master). But since the points in shadow help everything the shadow does, I don't find this surprising (i.e. that's the way I'd expect a reasonable formula to bias). Aside from a couple of other quirks/differences between warriors and masters, which might be more important, this does suggest that being able to bias a shadow toward using a skill you know well is valuable.

I'm very suspicious, actually, that the Master's ai is a bit of a copycat (or, at least, biased by the slvl of her skills in selection). Understanding how this works, *if* it is true, could be very helpful.

I certainly don't understand the (obsolete?) specific skills and levels mentioned in monstats.txt for shadows.

Lots more to figure out here (and I'm not the man for the job).

edit: misc. touch ups and this parting thought...
if shadow (warrior at least) gets synergy levels from this formula, a huge investment in getting your slvl for shadow warrior up (presumably when cast, fyi, though could be dynamic) might result in insane synergy to fire blast (you should also max fireblast, of course). In such a build--if my guess is correct--the value of teleport would be enormous (teleport on your right mouse, fire blast on your left mouse, teleport whenever you want to reset the shadow warrior, attempting to get her back into spamming fire blast). The fireblasts from the shadow, due to the huge number of fire blast synergies, would vastly exceed the damage from your own (master might work here, it depends on the ai bias toward skills with high slvl theory etc.)

edit2: still plugging away at testing aspects of this (can't report yet, too many things to figure out)... the formula:
shadow (warrior) skill level for their casting of a skill = int(player_current_skill_level/2) + int(shadow_slvl/3)
seems to hold up so far. Of special note is that it is level, not blvl (i.e. the bonuses to slvl are included). Also the shadow slvl is from when they were cast and the player's skill level for the skill the shadow is casting is current (i.e. "sampled" at the time the shadow casts). My big perplexity, still, is in figuring out how the synergies do/do not fit in. I have conflicting indications so far, so I guess I need some sleep and *more testing* ;D