evo crit: evidence for 0.6-additive base cc calculation

[Identity]

blubb.

[Identity]

#12, using foci from #1 (#5), but with 6% base crit.

expectations:
additive ~ 74% crit
higher hand ~ 50.32% crit
0.6-additive ~ 50.32-51.504% crit

sample size was 300, i got 145 crits (48.33%)


this still doesn't resolve the odd results from #1, but at least is in line with the expectations this time.

forgot to include this sample into the main posting, sry (different sheat, so many on desk by now )

[danker]

Big THX for the tests! ...even i don't like the results. xD

Should be clear enough now, that the 0,6 additive is the real one.

[Bryan]

Amazing test material,many thanks for carrying out the tests and of course for sharing 

Now what I believe the issue will be is,how this affects the standard endgame foci set.For instance,does it not seem that the offhand foci's base crit is more of an issue now?And how about ccm allocation between hands?The norm till now was to just fill one hand with it all.It would seem that the best approach would be to equally distribute ccm between hands,and through the addition calculation possibly even save a ccm relic in doing so.
Unless I misunderstand your theory.

[Harrowing]

The whole reason for doing a big sample size is it gets rid of all the skewed results.  You think that the 1,000 sample size is the outlier, I can tell statistically all your others don't agree with the one snapshot you did which has an error margin of 2.2%  Every other test you can have a -/+ swing % of 9% that means each result could never be conclusive enough.  Do each test with at least 500 sample sizes (3.7% margin of error)  That would be conclusive enough to determine actual results.

Secondly some of your gear settings result in a higher than expected crit and lower than expected result. Look at the statistical swing on these sample sizes.  You see that each is within the margin error size for a statistical sample.  I am not saying redo, just simply add more samples.

[Identity]

The whole reason for doing a big sample size is it gets rid of all the skewed results.  You think that the 1,000 sample size is the outlier, I can tell statistically all your others don't agree with the one snapshot you did

what do you mean by snapshot?

i'm well aware that sample size is important in such tests. that's why the major tests do have a sample size of 1k, rather than 100. where i only had 100 shots, the tests were declared as control groups for that reason.

on a sidenote, i did each test on Blabi in a pre-test which are not included in the results i posted. they were also in line with them.

two more theories i could imagine is, ccm is additive and hands behave separately (as stated by other testers), which i find highly unlikely but could at the very least be considered and maybe disproved. the second is, everything behaves completely separately (even more unlikely), even ccm only counting for the hand in which it is modded/auged. will think about those later

@bryan
where you put your ccm doesn't matter with the current result. i can't tell you though how cdb(caste) figures in, this remains to be tested in the future (no time atm, and not quite sure how to do it). i suspect it doesn't matter either, but ofc i cannot provide data (yet).
should i ever find some backup relics, i can do separate damage niveau testing maybe, on cata11 1x flameshards. eating an accuracy pill for 40% more cdb results in 2% less health on gremory, consistently. now, if 100% cdb from relic is missing, i should see some difference. but ofc, here's where the fun part starts, we're talking about 3-4kk relics which are more than quite rare ;/ i'm glad i just finished the first set of those for actual use.

[Identity]

#1 still bugs me, because the deviation is systematic

i'm still not sure what's going on here. will do some more tests and tell results later.


meanwhile, does someone know whether there were ccm(caste) (and cdb(caste)) mods in old hellgate london? i can't remember :/ wasn't that much into mechanics back then

[SuE]

i might have some plain demon cdb on jezzz which i can priovde for the test
i think you can use any crap claw and mod it with plain cdb and then do test with pills ?

[Identity]

yes, that will do

my problem though is, to look at cdb-mods, i need to make sure i crit all the time. so we need to definitely resolve critical chance first 

[Bryan]

Which reminds me.
Has the hard cap of 95% crit been verified?Because if it is so,we're looking at a constant 5% non-crit chance.You're the mathematician here,but on first thought I do believe it can affect results and the error margin.

[Harrowing]

On london, I know these things for a fact

#1 Crit chance and damage while dual wielding are actually different.  Each focus get its own source of damage with both mods/augs affecting the other focus.  Each works this way at 0.6 multiplier.  So each focus could actually independently crit from each other.  I have a very long detailed write-up on the subject that may or may not still be hanging around.  Written well back in 2008~09.

#2 Because each focus crits and has its own internal damage source you cant use it as a baseline for comparison.  Namely your base crit chance will always return a higher than expect result because each focus is separate meaning instead of E.g. 1 focus spell at 50% base in HG:G// In london (sp) you would have a 30% chance + another 30% at a critical hit respectively.  This turns out that is skews your baseline crit chance and damage.  So much so that you had to use the  crit-mini game + hp damage to determine whether or not you scored a Full crit  or only 1 focus happened.

#3 Like Bryan has already stated crit chance was for sure hard capped at 95%, because my current test evoker has 180% base crit chance through modded *mods* slotted in each of her Foci, yet she still doesn't always score a crit when the mini-game counter is up.

Finally, ever thought of just using a program like Fraps to record say 1,000 hits of firestorm and going back and counting the total # of crits?  Would surely cut down on the casting time needed to make all these tests.

[Identity]

you say you know for a fact, what was the source if i may ask? always interested in the details

hm... concerning the "full crit" vs "partial crit", i've never ever seen two distinct dmg niveaus in all my evoker's life. when i first read about this, i just blamed the flameshards/demonspine hitbox "glitch" (might not actually be a glitch, depending on how the skills were intended) where you would hit enemies twice with them (yes, this can be abused for insane amounts of dmg, and good evokers should practice ).

yea, already did some basic calc'ing concerning hands critting individually. doesnt fit the data though (can't reproduce calculations on the fly atm but i rejected it for that reason).

@firestorm
just tested vs occy (with demon glyphs, 20 and 23 crit, moloch pet -> 22/25 crit, total of 280 ccm among hands), then vs demons. the single projectiles didnt even crit versus the demons as often as they should, something's wrong with the skill. also, if you land a crit hit, it blinks for a moment. a series of blinkings would make the actual number indiscernible, even when using slomo after frapsing it.

[Harrowing]

you say you know for a fact, what was the source if i may ask? always interested in the details

hm... concerning the "full crit" vs "partial crit", i've never ever seen two distinct dmg niveaus in all my evoker's life. when i first read about this, i just blamed the flameshards/demonspine hitbox "glitch" (might not actually be a glitch, depending on how the skills were intended) where you would hit enemies twice with them (yes, this can be abused for insane amounts of dmg, and good evokers should practice ).

yea, already did some basic calc'ing concerning hands critting individually. doesnt fit the data though (can't reproduce calculations on the fly atm but i rejected it for that reason).

@firestorm
just tested vs occy (with demon glyphs, 20 and 23 crit, moloch pet -> 22/25 crit, total of 280 ccm among hands), then vs demons. the single projectiles didnt even crit versus the demons as often as they should, something's wrong with the skill. also, if you land a crit hit, it blinks for a moment. a series of blinkings would make the actual number indiscernible, even when using slomo after frapsing it.

With regards to hellgate london.  I can manipulate each items spawning damage attributtes as far back as using reanimator and 1.21b modding.

Basically what I did was create prisms with inherrent 25%, 50%, 75%, 100%, 125%, 150%, and finally 250% crit chance.  It shows up in the same column as where the 8% chance does for lenses.  Secondly When I control the prism I can look at the total damage done to a modded shulgoth monster @ level 5 with my level 5 evoker (shulgoth had 5000,000 hp for these tests).  Basically any damage, crit chance, crit damage, armor property were all moddable by me for the purposes of my testing.  Of which took a long time .

I came up with the theory on the first test when my crit result was much higher than expected.  more than 20% so.  So I started testing hp damage numbers @ various crit levels.  In these tests there were 4 distinct levels of hp damage corresponding on whether or not had an active "crit".

This kind of method can be used for HG:G currentlly because you have a huge damage variance in most of the focus items.  Mine were modded to do 25~25 damage.

Someone a while back thought this exact thing was happening in HG:G, Sadly I had to correct him because the current method doesn't have the #'s to hold up to that theory.

And now I am going to spend the day trying to find my old post on the details.

[Harrowing]

Which reminds me.
Has the hard cap of 95% crit been verified?Because if it is so,we're looking at a constant 5% non-crit chance.You're the mathematician here,but on first thought I do believe it can affect results and the error margin.

I had a Spec'ed MM with at minimum 108% crit chance while in sniper mode against demons.  It still wont critical all the time in harrow deep on the current HG:G.  So yes there is a hard cap of 95%

@ firestorm just tested vs occy (with demon glyphs, 20 and 23 crit, moloch pet -> 22/25 crit, total of 280 ccm among hands), then vs demons. the single projectiles didnt even crit versus the demons as often as they should, something's wrong with the skill. also, if you land a crit hit, it blinks for a moment. a series of blinkings would make the actual number indiscernible, even when using slomo after frapsing it.

Could it actually work in a PVP setting where the target is static and you could view each shots individual damage?  That is actually how I did all my recent venomus spirit testing.

I briefly came back to try and understand what could make situation #1 actually happen, and for the life of me it has me boggled too, because no viable current formula for dual wielding evokers can explain it.   And really it has me truely boggled, because its not a statisical outlier,  I have a 1,600 Sample size with the same basic results.

Oh and For all those that wonder.  CDB is cummulative at a 0.6 multiple from both hands  So 1500% in each becomes (3000)* 0.6 = 1800% CDB  I verified this with pvp damage results using only crit hits and a 2,200 sample size.  Very specific on cdb damage (which hand you have all your cdb mods doesn't matter as long as each hand has at least 1 cc+CCM+CDB mod/augment)

[Identity]

Oh and For all those that wonder.  CDB is cummulative at a 0.6 multiple from both hands  So 1500% in each becomes (3000)* 0.6 = 1800% CDB  I verified this with pvp damage results using only crit hits and a 2,200 sample size.  Very specific on cdb damage (which hand you have all your cdb mods doesn't matter as long as each hand has at least 1 cc+CCM+CDB mod/augment)

i've got a few problems here.

1.) pvp is not pve. mechanics might be implemented differently. also, players aren't mobs. same argument.

2.) what skill was used? this might be the biggest problem.

3.) deducing cdb-values from dmg seen is tedious. i'd need more information to judge that, but at first sight that looks more like guessing than actual results.

3.) the mod/augment for cc/cdb to "enable" the focus' internal properties is wrong. at least in pve. i'm 100% sure.

4.) the 0.6 is what i suspect, just not with base counted twice. that means, if you have, say, 1000 cdb without weapons and using two glyph-cages, you have 1000 + (200 + 200) * 0.6 = 1240 cdb. ofc, i dont have proof for that either. it's just what i currently think happens there.