Miscellaneous Maths/Testing II

1910111214

Comments

  • Eld said:
    Sena said:
    I'm still not sure. Someone who knows more about limb damage would be more helpful.


    In other news, testing the effects of armour on denizen damage isn't as simple as I expected. I figured that if a denizen dealt X damage, armour would reduce the damage by Y% of X, with Y being equal to some percentage of the armour stat. Instead, it seems more likely that the denizen deals X±Y, and armour only reduces X, except that doesn't quite fit perfectly.

    For example, I've been testing with the King Stag. Damage ranges from 459 to 741 (600±23.5% fits perfectly, but that could be a coincidence) with no armour/resistance. With roughly 21% armour (it's not possible to see actual stats any more, as far as I know, so 21% isn't exact), damage ranges from 394 to 670. It's possible that those aren't the actual limits (and I'll continue testing to be sure), but the sample sizes are high enough that it's unlikely.

    If it was reducing the total damage, then I'm not sure how to account for the lower end being reduced by so much more than the higher end. On the other hand, if it is something like 600±23.5% or 459+(0 to 282) with armour only applying to the base (the 600 or 459), the higher end is being reduced slightly more than the lower end.

    It could be explained by the damage being 459+(0 to 282) with armour applying to both the base and random portions at different rates (~14% reduction to the base and ~2% reduction to the random portion). But that seems unlikely.
    Well, if those are the true limits, then there has to be some scaling to the variable portion, since the ranges with and without armour are different (282 without and 275 with). The only way I see to get the armour only applying to the base, with the variable portion unaffected, would be if the true range of damage values with armour were 6 points larger, and you were missing a few outliers. If, for example, the damage without armour ranges from 459 to 741, and with armour ranges from 391 to 673 (so 3 points extra on either end of the armoured range), you could have 600*(1-.11333) +/- 141, but a reduction of 11.3333% bears no particular relation to the nominal 21%.

    If you have all of the individual damage values stored (rather than just the max and min), it might be interesting to histogram them and see if the distribution looks pretty uniform or peaked around some central value, and maybe try fitting them with a simple Gaussian. Or even just skip the histograms and compare the first few moments of the distributions (mean, variance, skewness) with and without armour.
    Also, the King Stag has multiple attacks: I assume you're just testing with one? From a little quick collection, looking only at his goring attack (which I'm guessing is cutting damage, but not sure), and in dragon (so level 2 cutting/blunt resistance, whatever that means), in 43 attacks, I got an average of 462 with a standard deviation of 53. The distribution looks pretty consistent with a Gaussian, but it's too small a sample to conclude much along those lines from.
  • Both attacks are blunt, and seem to do the same damage. I was tracking them separately at first, and after 500+ attacks they had the same min/max and nearly the same average, with and without armour.
  • As for the distribution, 735 unarmoured attacks:


    1042 attacks with 38 armour (21% in the current display, ringmail+cavalry):


    Seems uniform enough to me.
  • Yup, sure does. And it would indeed be surprising if those ranges weren't correct. In that case, I don't see any way for the variable part to be unaffected, or for both to be reduced by the same percentage. Out of curiosity, though, what are the percentages for the mail and the shield individually?
  • Ringmail is 24/17 cutting/blunt (actual stats are 43/30), cavalry is 4/4 (8/8).

    Also, it seems I didn't have the full unarmoured range. After another 565 attacks (for a total of 1300), I have 7 instances of 745 damage. That doesn't help make anything clearer.

    Rather than just get another thousand or so attacks to be extra sure I have the complete ranges (or try and get every possible damage value, if there's a finite number of possibilities like with sipping, though I already have 120 different values for armoured compared to sipping's 40), I've been working on getting a runewarden alt up to a high enough level to survive the King Stag for testing, so I can try fullplate+tower.
  • I was hoping that seeing the effects of high armour would make it easier, but it doesn't. It does seem to indicate that armour is most likely applying to both the base and variable portion at different rates, at least.

    The average and the max seem to scale pretty linearly with armour. The min is a lot higher than expected at 124 armour though, and therefore the range is a lot lower. It could be that for the base portion I've hit a cap on armour reduction, but haven't hit a cap on the variable portion. I guess I'll have to test more armour values.

    A summary of my data so far:
    Unarmoured: avg 593.16, min 459, max 745, range 286, attacks 1300
    38 armour (ringmail+cavalry): avg 524.76, min 394, max 670, range 276, attacks 1042
    124 armour (fullplate+tower): avg 393.96, min 316, max 484, range 168, attacks 716
  • I've been collecting some numbers with just resistances, which I expected to be a bit more straightforward than armour, in case they shed any light. The samples are pretty small, unfortunately (basically just sitting there until someone comes along to do the quest). I wasn't sure how you determined the damage type, so I thought I'd check with toughness to see if anything changed.


    Horkval, no toughness:
      Attack 1 (gored): avg: 537.18, min: 417, max: 677, range: 260, attacks: 202
      Attack 2 (trampled): avg: 529.81, min: 417, max: 677, range: 260, attacks: 184
      Combined: avg: 533.67, min: 417, max: 677, range: 260, attacks: 386
    Horkval + toughness:
      Attack 1: avg: 524.92, min: 380, max: 677, range: 297, attacks: 244
      Attack 2: avg: 509.49, min: 380, max: 674, range: 294, attacks: 207
      Combined: avg: 517.84, min: 380, max: 677, range: 297, attacks: 451
    It's interesting that I saw 380s in both of those when I hadn't seen them in the earlier ones at all, but the average doesn't look to be reduced near as much as would be expected for toughness, so I'm assuming you're right that they're both blunt.
    Both combined (assuming blunt damage):
      Avg: 525.14, min: 380, max: 677, range: 297, attacks: 837

    That's pretty close to consistent with the damage being 600 +/- (0-150), with hork resistance reducing the 600 by 12-12.5% and leaving the variable part unchanged, but there's plenty of room for imagining patterns like that.
  • Unarmoured: avg 593.16  - 100%
    21% armour (ringmail+cavalry): avg 524.76  - 88.5%
    55% armour (fullplate+tower): avg 393.96  - 66.5%

    If we go with  ring of magus description (Example, a magical-type attack is delivering 1000 damage: 500 of it would go through unaffected, and 500 would be reduced a certain percentage by the power of the Ring.) those tests are close to % marks.
  • The Ring of the Magus description doesn't literally mean that half the damage is reduced and half is unreduced, that's just a simplified example. It's more complicated than that.

    With weaponry attacks for example, there are two parts to the damage, a health-scaling portion and a static portion. Only the static portion is reduced by armour, while the health-scaling portion is unreduced, which means that the more health health you have, the less effective armour is against weaponry attacks. In other attacks, the health-scaling portion might be reduced while the static portion is unreduced, or a certain portion of both parts might be reduced, or the entire attack might be unreducible (so it ignores armour entirely but not resistances), etc.
  • Demon fortify drains floor(CurrentHealth/20) of your health (boosted by Scimitar of Yen-Sorte, so Mhaldorians take 7% more damage from it than rogues) and heals your baalzadeen for floor(CurrentHealth/20)*5.
  • WessuxWessux Chattanooga
    I miss city bonuses. 
    Invest in a 9mm retirement plan.
  • Do any smiths have a full listing of what they crafted and after which items they gained a rank? I'm curious to see if I can figure out exactly how forging experience works.
  • edited January 2015
    Hmm, not sure where this should go. I bugged it a while back but have not heard anything. The last Bugfix Announce claimed the SETALIAS part was fixed at least, but does not seem to be. @Makarios @Tecton

    SETALIAS

    Here, the setalias command is still caught by retardation

    5418h, 4483m, 23630e, 17570w cexkdb setalias test1 kick carbine
    You move sluggishly into action.
    5418h, 4483m, 23630e, 17570w cexkdb
    Alias "test1" will now execute: "kick carbine"
    5418h, 4473m, 23630e, 17570w cexkdb

    QUEUE

    Here, things get weird. The QUEUE ADD command hits retardation, but if you have balance by the time it completes, the attack will go through straight away. If you do not have balance at that moment, then when it fires on balance you hit retardation again.

    5418h, 4473m, 23630e, 17570w cexkdb queue add eqbal test1
    You move sluggishly into action.
    5418h, 4473m, 23630e, 17570w cexkdb
    [System]: Added TEST1 to your eqbal queue.
    [System]: Running queued eqbal command: TEST1
    You cease holding your breath and exhale loudly.
    You leap into the air and launch a flying kick at Carbine.
    You connect!
    {Note no sluggishly into action when the queue fires}
    5418h, 4473m, 23605e, 17570w cekdb
    The Anthem shifts, protecting from electricity damage.
    5418h, 4473m, 23625e, 17570w cekdb queue add eqbal test1
    You move sluggishly into action.
    5418h, 4483m, 23625e, 17570w cekdb
    [System]: Added TEST1 to your eqbal queue.
    5418h, 4483m, 23625e, 17570w cekdb
    You have recovered balance on all limbs.
    [System]: Running queued eqbal command: TEST1
    You move sluggishly into action.
    5418h, 4483m, 23630e, 17570w cexkdb
    You leap into the air and launch a flying kick at Carbine.
    You miss.
    {But here both the QUEUE ADD and the firing of the QUEUE hit retardation}
    5418h, 4483m, 23605e, 17570w cekdb queue add eqbal test1

    SETALIAS is obviously still bugged, but I have no idea what is up with QUEUE. It should hit retardation when it fires the action, not before or both, right?

  • Arador said:

    Here, things get weird. The QUEUE ADD command hits retardation, but if you have balance by the time it completes, the attack will go through straight away. If you do not have balance at that moment, then when it fires on balance you hit retardation again.
    [   ]
    SETALIAS is obviously still bugged, but I have no idea what is up with QUEUE. It should hit retardation when it fires the action, not before or both, right?

    I believe this is working as intended.  Queue doesn't actually enter commands into a queue if you have the specified balance.  It shows you a message but all it really does is execute the command as if you didn't use the QUEUE command.  That's why it gets delayed by retardation.  However if you're off balance, it does use the queue system (which bypasses retardation as a system command), and it's handled correctly by performing the command when you regain balance, at which point is is delayed appropriately.

    The reason the log looks the way it looks is because if you attempt to queue things at any time when you already have the balance, it doesn't appear to actually enter a Queue (although the message says that it does).  Things are only stored in a queue if you don't have balance specified.  When you used test1 (kick) the second time, you were off balance, so did actually enter a queue
  • Except it hit retardation twice when I used it off balance. Please read that second kick again. I asked it to add the alias to the queue, it moved sluggishly into action, then it added the alias to the queue, then I regained balance, then it fired the alias which hit retardation again. 

    I understand what you mean by the first, though I do not think that is consistent with how the queue appears to work. You are adding to the queue, it just fires the action immediately. While the way it handles it in the first kick does have you coming out and kicking in the same amount of time, it is a very awkward way to handle it. 

  • edited January 2015
    Arador said:
    Except it hit retardation twice when I used it off balance. Please read that second kick again. I asked it to add the alias to the queue, it moved sluggishly into action, then it added the alias to the queue, then I regained balance, then it fired the alias which hit retardation again. 

    I understand what you mean by the first, though I do not think that is consistent with how the queue appears to work. You are adding to the queue, it just fires the action immediately. While the way it handles it in the first kick does have you coming out and kicking in the same amount of time, it is a very awkward way to handle it. 

    I see what you're saying, and it does appear to be hitting retardation twice (I actually didn't see the delay between the alias and the queue command in that log, I thought the queue command was immediate).  That's definitely not working as intended (99.9% sure of that at least).  Would paste-bin that and bug it with an explanation, or a link to your post here too (here).

    edit: 
    If that's just for demonstration, I get that, but obviously that's not very good use of queueing.  Personally, I don't think queueing should be used in retardation, at all, ever, so it's really a non-issue to me.  In all cases that I'm aware of, commands should be sent 1 second early, and one at a time, neither of which are accomplished using queueing.

    It would be interesting if BATCH mode were extend to affect queues though, but there'd be issues in determining which queue to run first if you have items in multiple queues (although again, that'd probably be so rare that it wouldn't matter).

    @Makarios, can we make BATCH mode make queued commands fire one at a time (in order), when set to OFF (batch mode sends curing commands all at once, non-batch mode sends them one at a time).  Could also just have queued commands automatically execute only one command at a time in queues, instead of firing them all at once (deleting everything but the last command).  Personally I don't really think we need to hard-code cookie-cutter easy-mode for retardation, but this method would still be 1-second slower on all commands than manually sending things 1 second early, so it's easiness in exchange for much slower actions.
  • Using a queue in retardation is obviously not the most efficient thing. You can hit faster of you anticipate balance. But that does not mean you would not want the system to work in a predictable and consistent manner.

     It makes scripting more consistent and is a safe way to start so that combat jitters and inexperience does not have you mashing at keys because every second feels like a year to you. 

    It lowers the barrier of entry for retardation fighting without lowering the barrier of possible efficiency for good fighters. 


  • Well I think it's clearly a bug, so I'd just bug it and call it a day.  I've probably bugged 100 things with curing/queuing, and they've all been fixed.
  • When using queuing and/or serverside aliases with multiple commands, it can make balance times appear significantly shorter (possibly up to half a second if your alias is doing a lot).

    Presumably because it delays output on your end of the commands being sent but not (or not as much) the serverside balance calculations.
  • Did a bit of testing of blademaster and dragon battlerage, mainly checking to see whether either damage or rage gain are affected by stances or bands. Turns out damage is not, and rage gain is. Not sure if anyone has posted damage numbers for various battlerage attacks anywhere, but I couldn't find any, so here's mine:

    Blademaster:
    Leapstrike 437
    Spinslash 1420
    Headstrike 327 (unboosted)

    Red Dragon:
    Overwhelm 583
    Dragonblaze 1721 (total over 4 hits)
    Flamebath 327 (unboosted)

    Wasn't able to test the damage boost for headstrike and flamebath, but the fact that they're the same suggests that that type of attack might have standardized damage. Does anyone know if that's the case, and if so, what the boost is?

    For rage gain, I just whacked a practice dummy a bunch of times and recorded the rage gain for each hit and the balance time for that attack, and calculated the rate as the mean gain over the mean balance time.  Numbers:


    Arash 
    Avg gain: 3.761 +/- 0.011
    Avg balance: 1.849 +/- 0.005
    Avg rate: 2.034 +/- 0.006

    Sanya
    Avg gain: 3.950 +/- 0.014
    Avg balance: 1.943 +/- 0.005
    Avg rate: 2.033 +/- 0.006

    Mir
    Avg gain: 4.983 +/- 0.017
    Avg balance: 2.450 +/- 0.006
    Avg rate: 2.034 +/- 0.006

    Thyr
    Avg gain: 3.564 +/- 0.015
    Avg balance: 1.753 +/- 0.006
    Avg rate: 2.033 +/- 0.008

    No band:

    Sanya
    Avg gain: 4.530 +/- 0.023
    Avg balance: 2.232 +/- 0.006
    Avg rate: 2.030 +/- 0.008

    Mir
    Avg gain: 5.547 +/- 0.029
    Avg balance: 2.754 +/- 0.008
    Avg rate: 2.014 +/- 0.008

    Thyr
    Avg gain: 4.120 +/- 0.019
    Avg balance: 2.062 +/- 0.009
    Avg rate: 1.998 +/- 0.009
    So rage gain is definitely adjusted to account for both stance and band, although it looks like there may end up being a little bit of an overall bonus for the band. Some of the numbers for the unbanded case look a little wonky, though, so I might double check them.
  • Rage gain is just strictly dependent on the balance time of the attack, so anything that affects that will affect the raw rage gain per attack (but not the overall rate at which rage is gained).
  • Antonius said:
    Rage gain is just strictly dependent on the balance time of the attack, so anything that affects that will affect the raw rage gain per attack (but not the overall rate at which rage is gained).
    Yeah, that was my assumption, and the testing was mainly just to verify as much, which it basically did. But I could easily imagine there being some rounding effects in the adjustments to the raw gain that would result in the sort of ~1% level differences I seem to be seeing in the rates. Even if they're real, these aren't big enough differences to care about in terms of actual gameplay, just the kind of little details that I like to poke at for potential insight into the underlying calculations.
  • So, I have most of the weekend free. What aspects of the new class would people like me to test first?

    Things that I can test on my own without help (also without artefacts) will get done much faster.
  • Here's my personal list of stuff on Terminus I want to figure out.

    Terminus:

    - Hunting

    Trusad - how much does it raise crits?

    Tsuura - how much damage resist vs. denizens?

    - Longevity

    Ukhia - what is the % chance to remove clot willpower cost?

    Qamad - what is the willpower regeneration boost?

    Fliennad - does it only affect tree, or will it make your moss and boar work instantly too?

    - Augmentation

    Mainaad - how much is the damage boost? Does it stack in a weird way with the artie scythe?

    Balateth - how much is the speed boost?

    Tah'maal - how much fire resistance? How much does it reduce cooldown of rebirth?

    Mainaas - how much cutting/blunt resist? Does it stack in a weird way with the Tsuura reduction from denizens only?

    - Conquest

    Tooros - how much does int/collar affect the damage?

    Trointeth - how lon does it disrupt shrines?

    Lament - how long does this last? If I use lament and someone kills 20 people, do they really gain 20 stats?

    Environmental -

    Bahar - does it work in room too?

  • How strong is Bahar in terms of other ranged attacks? Does anyone know? 

  • What damage type are the scythe attacks if they aren't cutting?

    Aka will +magic damage arti be viable? 
  • Nevren said:
    What damage type are the scythe attacks if they aren't cutting?

    Aka will +magic damage arti be viable? 
    They appear to be psychic.
  • Awesome, thanks! 
Sign In or Register to comment.