Jump to content

Vilius

Player
  • Content count

    43
  • Joined

  • Last visited

Posts posted by Vilius


  1. On 9/4/2017 at 9:24 AM, Brynja said:

    The change OP mentioned was merged with the development-branch. 

    It wasn't implemented to the live servers (yet?) though, right?

    https://forum.elysium-project.org/topic/53637-development-update-08092017/

    Seems to be live now.  Unbelievable.

     

    @C0n0n I reviewed your links, my cursory findings:

    1. Agrees with my statement about 5% being the base miss chance

    2. Agrees with my statement about 5% being the base miss chance, but claims flat linear 0.04% penalty per difference in mob def vs wep skill (or 5.6% vs 63 mob at 300 wep skill) which is certainly wrong.

    3. I found nothing relevant here.

    4. I found nothing relevant here.

    5. Lots of conflicting claims in this topic, not sure what can be gleaned here.


  2. On 9/1/2017 at 8:19 PM, Trym said:

    Even tho these changes had no evidence there are no evidence that says the previous numbers were correct either.

    I believe the changes was made because they made more logic and not because of strong evidence

    There are 3 parts:

    1. Removing the bump

    At first I thought removing it made sense.  But the more research I did, I found more and more evidence that the bump was in place in at least TBC, and probably vanilla.  Main tests are summarized here: http://web.archive.org/web/20090820041711/http://elitistjerks.com:80/f31/t11885-rogue_warrior_weapon_skill_adjustment_discussion/

    Vanilla tests were not as high quality, but there is an absence of patch notes saying changes were made to this mechanic.

    However, of the three changes, removing the bump is the least offensive in terms of honoring the original mechanics.

    2. 315 endpoint

    5% was the original endpoint, and I think it was correct because it matches tests above, and it is equivalent to the equal level case.  5.6% makes no sense at all because 0.04% isn't even used in weapon skill vs miss scaling, either before the change, or after.  And even if it were, 5.6% pops up due to unequal defense vs wep skill, not equal defense vs wep skill.  It's a complete misunderstanding of other mechanics applied to something entirely different.

    3. 300 endpoint

    From what I could find, vanilla tests indicated over 8% but less than or equal to 9%.  They didn't do enough samples to nail down the next digit.  So anything from 8.1 to 9.0% could be valid from that alone.  However, 9% matches the TBC tests.  Until someone can give some evidence for 8.6%, or some other 8.X%, I don't see why we should believe that's the correct number.  Again, his endpoint of 8.6% was partially based on an incorrect endpoint at 315 of 5.6%.  If same scaling was kept, but 315 endpoint corrected to 5.0%, that would make his 300 endpoint 8.0%, which vanilla tests showed was incorrect.


  3. A change was recently accepted by the devs for how weapon skill affects miss rates:

    https://github.com/elysium-project/server/pull/759

    Some discussion on the subject can be found here:

    https://www.reddit.com/r/ElysiumProject/comments/6s4ae8/warning_a_baseless_change_to_hit_formulas_with

    While I agree the bump at 305 was strange and probably incorrect, he went beyond just fixing the bump and I have some serious concerns about the additional changes:

    1. Cephel is changing the 315 endpoint to 5.6% based on an incorrect understanding of how 5.6% is used elsewhere
    2. No evidence is given for the 300 endpoint being changed from 9.0% to 8.6%
    3. In fact no evidence is given for any part of the change
    4. He claims to have researched the issue extensively but has so far been unwilling to share links to that research
    5. Devs accepted the change with no evidence, which concerns me since any mechanics change can potentially affect gearing (and this one definitely does)

    My understanding of how this mechanic should work:

    1. 315 should be 5% because that makes weapon skill equal to mob defense, which is equivalent to the equal level case.  Before this change, this is what elysium code had in place.
    2. Unfortunately 300 wasn't sufficiently tested in retail vanilla.  What I recall from elitistjerks topics is no higher than 9% but not enough data to nail down the tenths.  So I don't know if this should be 9% or 8.6% or something else.
    3. 305 value is debatable (never seen a good info source)
    4. Piecewise linear between 300 to 305, and 305 to 315

    I've done some digging in elitistjerks web-archive but haven't found anything more concrete than that.  Does anyone have some solid evidence for how the scaling should be in vanilla?


  4. Yes, +crit increases crit chance in roll 2 (this is also why the crit cap for yellow hits is 100%, completely different from the white crit cap).

    Mobs can only crit on white hits, so what levran said is technically true (since crit block doesn't occur on white hits)


  5. See Unit.cpp from source code.  White hits never critically block (there is code for this case but it never executes), but yellow hits can.  In fact yellow melee hits are implemented as a 3 roll system in the code.  First miss/dodge/parry is checked, if they fail then crit is rolled, and finally block is rolled.


  6. The only answer that matters is what the code says, and it works like this in the specific case of white hits from behind against level 63 mob:

    • Base chance your attacks are dodged is 5.6%

    • Each point of weapon skill above 300 reduces dodge chance by 0.04%

    • First 5 points of weapon skill above 300 reduce miss chance by 3% total

    • Above 305, each point of weapon skill reduces miss chance by 0.1%

    • Base miss chance is 9% for 2H/1H+shield, or 28% for dual wield

    • Therefore 2 ways to affect your crit cap: more weapon skill or more hit

    In summary, first 5 points of weapon skill (or weapon racial) is a pretty big deal, converting glance damage from 65% to 85%, and also equivalent to +3% hit, and 0.2% less chance to dodge.


  7. No, the formula in code is correct.

    missChance -= (skillDiff + 10) * 0.4f - 2.0f

    is equivalent to:

    missChance = missChance - ((skillDiff + 10) * 0.4f - 2.0f)

    For example:

    weapon skill = 300

    mob def = 315

    skillDiff = weapon skill - mob def = -15

    missChance = 5 - ((skillDiff + 10) * 0.4 - 2) = 5 - ((-15 + 10) * 0.4 - 2) = 5 - (-5 * 0.4 - 2) = 5 + 2 + 2 = 9

     


  8. 3 hours ago, Omgdontdie said:

    Assuming no overlap, that gives us 50% uptime. Realistically, its not uncommon that the procs overlap. You'd have to use a poisson distribution to estimate a realistic uptime if you account for overlaping.

    An expected value of 50% uptime is your best case scenario here. 77% is beyond inaccurate. 

    I agree with you in principle about Poisson being a better method than what Undertanker is doing, although I prefer to just do Monte Carlo sims since it's easy to add in non-uniform events (like no shield slam procs) and other irregularities.

    So I did just that for this one case of Quel'serrar proc uptime.  For now I didn't model dodge/parry/miss.  I assumed every 4th gcd was Shield Slam, and did runs for both SS can and cannot proc.  Other gcd's were assumed to be capable of proc.  In each case I did 1 million runs of 3 minute fights.  These are the results:

    2 PPM w/ SS procs: 57% uptime

    2 PPM w/o SS procs: 50% uptime

    1.75 PPM w/ SS procs: 49% uptime

    1.75 PPM w/o SS procs: 45% uptime

    Undertanker I understand your desire to simplify, but when you oversimplify you get very inaccurate results.


  9. Quote

    "Gains windfury", could simply be from windfury totem applying the buff to your weapon, different from gaining extra windfury attacks. Again, i dont play horde so not sure.

    "gains windfury totem" is just how the combat log displays that an extra attack has triggered from Windfury.  That message pops in, you get the extra hit, then it fades.  Sometimes the extra hit message can appear before the "gains windfury totem" message.  So far that execute case is the only case I've seen where you see the "gain" and "fade" message, but no extra hit occurs.

    Quote

    Also, extra attacks can happen on immune mobs, at least with Ironfoe. Ive had Ironfoe proc on banished/immuned mobs many times.

    Good to know, I'll have another look at the procEx code.


  10. Quote

    3. This could be yes, but im not sure. I might be able to explain your combat log though. From my understanding, execute has 2 components, and this complication makes it not trigger main hand chance on hits. That might explain why it triggers HoJ, but not windfury (since windfury is a weapon enchant). Hopefully the next dev update fixes this.

    Possibly, although strange that the log would show "gains Windfury Totem" but no extra hit occurs.

    Quote

    10. Like #8, not sure. I would think it would after the fix though. To counter Undertanker's point, I think the "immune" argument doesn't apply here. Based on my testing, I believe that the result of a hit/miss/crit etc is rolled and calculated and then afterwards the mob will be immune to the damage - so the attack still registers as a hit that does damage, but the target just happens to be immune to that damage. A little bit different than casting a spell/ability that does no damage to begin with.

    For auto attacks, the code has three separate proc flags: procAttacker, procVictim, and procEx (extra attacks).  procAttacker is set to "successful melee hit" before damage is calculated and before the outcome is rolled.  procEx is set to PROC_EX_IMMUNE if the target is immune (which isn't one of the values that causes extra attacks to trigger).  So I believe if a target is immune, weapon procs can happen but extra attacks don't.  I'm sure that Crusader procs on banished targets for example, so that matches what Undertanker has observed.  All that being said, Sunder Armor isn't handled by the immunity code so the bug we are seeing with Sunder not procing is a different issue.

    Quote

    Another comment in response to Undertanker's bonus on chain procs. I know extra swings can trigger other non-extra swing procs, but I have never actually seen something like fiery/maelstrom proc another fiery/maelstrom and this could be due to a lack of extensive testing. What he is posting seems to be based on what he believes should happen, but might not necessarily reflect what happens on this server. If you are adamant about testing this, I would recommend getting fiery, and see if you can get a double proc from one white swing.

    In the code there is a check done for procs to see if the trigger spell and proc spell are the same, and if so it disallows it.  If anyone has combat log evidence to the contrary I'd like to see it, always possible I missed something because the code is complex.


  11. I looked this up in the source code recently.  Assuming weapon skill is at least 300, for melee/ranged yellow attacks, miss chance is coded as 5% + penalty based on weapon skill and mob defense:

    If (mob def - wep skill) > 10, penalty = (mob def - wep skill - 10) * 0.4% + 2%

    Otherwise penalty = (mob def - wep skill) * 0.1%

    For attacks that require a weapon (e.g. HS), wep skill is set to your actual weapon skill including racial/gear bonuses.  But for attacks that don't require a weapon (e.g. Shield Slam), wep skill is set to 300.  Either way, the worst case miss chance is 9.0% so that's how much hit you need.

    Re: Band of Accuria, I understand the logic about fury warriors, but I've seen hunters use this: https://docs.google.com/spreadsheets/d/1gJT3mKYr5ru2VOvwb1EMf43BjSx52yZdJiHVZGi8J1o/pubhtml#

    BoA is listed there until Naxx patch.  Is there a better combo for hunters that gets them to +9% hit without using BoA in 1.7 and 1.9?


  12. Most of what you are saying seems right and matches what I've seen.  Couple things:

    #6 : I agree that's how they worked in Vanilla, but the devs' position is that they shouldn't work that way, and Shield Slam right now does not proc extra attacks for sure.  See https://forum.elysium-project.org/topic/48239-ironfoe-testing/

    #10: I agree that's how it should work, but right now Sunder isn't proc'ing extra attacks, and was unsure if it was also not proc'ing stuff like Maelstrom.  In topic linked above the extra attack bug is discussed and there's a fix that should be in the next development update.


  13. I'm trying to compile a bunch of different information on procs from the source code, forum posts, in-game tests etc.  Below I've listed some questions and in many cases what I *think* the answers are.  Other folks feel free to correct my thinking or add additional knowledge.

    1. Can extra attacks from WF proc things like HoJ?  No (I believe this behavior is governed by the m_doExtraAttacks flag)
    2. Do all of these work identically as in previous question: WF, HoJ, Sword Spec, Ironfoe, Thrash Blade?  Yes
    3. Can a melee hit cause both HoJ and WF procs from that single hit?  Yes.  The code seems to track a counter m_extraAttacks that decrements as extra attacks are executed.  But I'm seeing somewhat conflicting combat log evidence.  I have a log that shows a player hitting Execute, and messages for both HoJ and Windfury Totem pop up, but only one extra attack is shown in the log.  No other events nearby.  Perhaps this is just a combat log bug?
    4. Do all of these work identically as in previous question: WF, HoJ, Sword Spec, Ironfoe, Thrash Blade?  Yes
    5. Can abilities that require melee weapons but that do no damage (such as Sunder Armor) cause procs?  No, but this is believed to be a bug and slated for fix, see Ironfoe testing topic.
    6. Can abilities that require a shield (Shield Slam, Shield Bash) cause procs?  This one is tricky.  In vanilla evidence suggests that they could.  But early in 2.0 this was called out as a bug and removed.  From the Ironfoe topic, it seems that Elysium devs' philosophy is to keep the "fixed" behavior, so Shield Slam and Shield Bash shouldn't proc stuff.  But currently Shield Bash does, so this appears to be a bug based on their ruling.
    7. In two questions above, do extra attack procs function the same as weapon chant procs?  e.g. will Sunder be able to proc both WF and Fiery after the fix?  Might have to wait until after the fix to actually test.
    8. Can DMC: Maelstrom proc from any melee damage dealt including procs?  I've heard the answer to this is "yes" but I don't have one myself to test.
    9. If above question is yes, would that include all of the following or only some of these?  White hits, yellow hits requiring weapon, yellow hits requiring a shield, yellow hits requiring neither (e.g. Bloodthirst), Rend ticks, Deep Wound ticks, TF and other weapon damaging procs, Lifestealing/Fiery procs, rogue poisons, damage reflection (e.g. thorns, Essence of the Pure Flame)?
    10. Again for DMC: Maelstrom, can things that do 0 damage proc it, e.g. Sunder Armor?  I'm guessing no but I suppose this will have to be retested after the Sunder Armor fix mentioned above is implemented.
    11. Can abilities that hit multiple targets (e.g. Cleave, Sweeping Strikes, Whirlwind) cause procs on each of those targets, and if so are there any restrictions to those?

  14. Ironfoe > Quel for threat.  Quel is not a threat weapon, it's a mitigation weapon.

    Alcor's is only good for one fight and that is Vael, because you can spam HS at a much higher rate due to the speed of 1.3.  A big deal is made by some about Alcor's but other 1.3 daggers are nearly as good.  Finkle's is only like 5 tps less on Vael by my estimates.


  15. I think the RagePS cell is not counting rage gained from outgoing dodges?  Outgoing dodge/parry should reward 75% of damage that would have been dealt from a normal white hit *before armor reduction*.  Evidence from server code below.

     

    In Unit.cpp:

    In function CalculateMeleeDamage (lines 1571-1584):

    totalDamage becomes base phys damage dealt after armor reduction taken into account

    cleanDamage becomes amount of damage absorbed by armor

    Same function (lines 1662-1685):

    if hit outcome is dodge or parry, cleanDamage becomes base phys damage dealt before armor reduction

    In function RewardRage (line 654):

    cleanDamage is used to reward rage from outgoing dodge/parry, but multiplied by 0.75 first

    So outgoing dodge/parry reward rage based on 75% of damage that would have been dealt *before armor reduction*.


  16. Don't feel bad if dps pulls off you on multi-mob pulls (especially shamans spamming ES).  As mentioned above warriors are quite rage-starved especially while leveling, but most players don't understand this and just go ham immediately.


  17. I've never played an alliance toon, so WW axe may not be worth the effort with Bonebiter available for you at 34.  As an orc it was a no-brainer and lasted me 15 levels.

    Thermaplugg's axe is almost as good, but keep in mind it's a 20% drop rate, and there's no guarantee.  At least with quest rewards (WW axe and Bonebiter) it's guaranteed.  In general I ran each dungeon once or twice at most, and that set me up with gear pretty well but it's luck dependent.

    Skinning will probably make you more gold than alchemy while leveling.  On my server it was most profitable to make certain items from the leather and vendor those, or in some cases sell enchant mats, so alts or friends help.  I dropped skinning at 60 in favor of professions that would help more with raiding.


  18. Do your warrior quest at 30 (you'll need help for at least the last step) to get your Whirlwind Axe, which will last you until Bonebiter at 34 (earliest).  For horde they'd keep WW until Uldaman.

    For gold while leveling, grinding the right beasts is a good way especially if you have cooking and skinning.  For example, tigers in STV are excellent because their vendor trash is valuable, Jungle Stew recipe makes 2 with just one meat and no expensive spices, plus leather.  They are far better than basilisks in shimmering flats if you can cook.  Also way cheaper to cook your own food than buy it.  Herbs is great for gold especially at max level.  Alchemy won't make you much while leveling.


  19. I don't know about DoTs, but sorry you are wrong about breath.  Test it yourself, or watch any vid that shows tank's frame.  Here's one, you'll see that the tank's health plummets and rage spikes for every breath, e.g. at 1:00.  If you need further evidence, there is nothing in the server code that selects out breath attacks before calling RewardRage().

     

     

     

×