Item Search
     
BG-Wiki Search
+ Reply to Thread
Page 1 of 4 1 2 3 ... LastLast
Results 1 to 20 of 72

Thread: Test Instruments Thread     submit to reddit submit to twitter submit to tumblr

  1. #1
    An exploitable mess of a card game
    Join Date
    Sep 2008
    Posts
    13,252
    BG Level
    9
    FFXIV Character
    Gouka Mekkyaku
    FFXIV Server
    Gilgamesh
    FFXI Server
    Diabolos

    Test Instruments Thread

    Purpose: The purpose of this thread is to compile instruments for testing. This entails resources necessary (or useful) for creating experiments and strategies that have worked in the past (Example: Goldfish at Sea for critical hit rate). I'm currently looking at the best way to format this beyond Player Stats + Mob Stats + Old Strategies section. Unlike the Data compilation thread, it's best if we actually paste the information here for people to view (Unless it's a program like Bellsouth). We can reduce clustering through spoiler tags obviously.

    Discussion should precede with contributions to the original post and formatting suggestions.

    Player Stats
    http://bellsouthpwp2.net/k/i/kinemat...tatCalculator/

    Calculating Player Melee Accuracy:
    Spoiler: show
    Melee Accuracy is determined by your weapon skill, DEX, type of weapon, gear, and traits. It is used to determine your Hit Rate, which has a cap of 95% and a floor of 20%.

    H = 1 for onehanded weapons and 2 for two-handed weapons
    Weapon Skill less than 200: Accuracy = Skill + (1+H)*DEX/4 + Gear Accuracy + Traits
    Weapon Skill more than 200: Accuracy = 200 + FLOOR((Skill-200)*.9,1) + (1+H)*DEX/4 + Gear Accuracy + Traits

    http://wiki.ffxiclopedia.org/wiki/Accuracy_Bonus


    Calculating Player Melee Attack:
    Spoiler: show
    Melee Attack is determined by your weapon skill, STR, type of weapon, gear, and traits. It is used to determine your cRatio, which has a cap and a floor that also depend on your weapon.

    H = 1 for onehanded weapons and 2 for two-handed weapons
    Attack = Skill + (1+H)*STR/4 + Gear Attack + Traits + 8

    http://wiki.ffxiclopedia.org/wiki/Attack_Bonus


    Calculating Player Defense:
    Spoiler: show
    Defense is determined by your Vitality, level, gear, buffs, and traits. It is used to determine monster cRatio (when they attack you) along with monster Attack.

    Defense = 8 + Level + VIT/2 + Traits + Gear + C

    51 > Level, C = 0
    61 > Level > 50, C = Level-50
    Level > 60, C = 10

    http://wiki.ffxiclopedia.org/wiki/Defense_Bonus


    Calculating Critical Hit Rate:
    Spoiler: show
    Player Critical Hit rate is determined by your DEX, the monster's AGI, gear, and merits.

    Critical Hit rate = Merit Level + f(dDEX) + Gear% + 5%

    Code:
    dDEX = Your DEX - Target AGI
    
    f(dDEX[<1]) = 0%
    f(dDEX[0;14]) = FLOOR(dDEX/7)%
    f(dDEX[15;40]) = FLOOR(dDEX/10)% + 1%
    f(dDEX[41;50]) = (dDEX - 35)%
    f(dDEX[>50]) = 15%
    http://wiki.ffxiclopedia.org/wiki/Ta...tical_Hit_Rate


    Calculating weapon skill stats:
    Spoiler: show
    Weapon Skill damage is determined by a large variety of factors, outlined here:
    http://wiki.ffxiclopedia.org/wiki/Ca...n_Skill_Damage
    Some magical weaponskills don't actually follow the formula listed on this page, but have a more complicated form.

    In order to determine the weaponskill, you need to hold all other constant. For physical weaponskills, pDIF is a randomization factor applied at the end, which complicates the analysis. Magical weaponskills can skip most of Steps 1-6 because they generate a constant number at a constant TP.

    Steps for physical WS testing to remove the effects of pDIF: Skip if magical
    1) Change your subjob to Thief (unless it's your main, then sub SAM or something), bring a lot of STR gear, and bring relevant gorget and/or belt. Determine the skillchain properties using Sekka, if you don't know what they are. (Maybe make this another test section?)
    2) Bring an x-hit build that's convenient for getting 100TP while wearing a lot of STR and no neck/waist. If you're using a 1H weapon, it's going to be faster to use a ranged weapon to build TP. I recommend fulltiming movement speed gear if you can and not using food.
    3) Go to a starter zone, record your stats, and unload spellcast. Turn on your parser.
    4) Use your x-hit build to consistently get 100TP, and then SA->WS level -1 monsters (right outside the towns, the weakest of the weak).
    5) There are three pDIF values that are recognizable. Your max damage (pDIF=3.15), the frequency spike (pDIF=3.00), and the min damage (depends on weapon type). Normally the max damage is the easiest one to find, although the 3.0 frequency spike is a nice control. You want to keep doing SA->WS with constant TP until you're fairly sure you've found one of those values. Using your guesses for the other two is a nice way to confirm.
    6) Use the pDIF value and resulting damage value that you're confident in to factor out pDIF. This leaves you with Damage = (D+fSTR+WSC)*fTP.

    Determining fTP:
    7) Put on an appropriate gorget and/or belt (there should have been no gear in your neck/waist slot before) and repeat steps 3-6.
    8) This gives you another (D+fSTR+WSC)*fTP, except this time you know that it's either .1 or .2 more fTP than it was before (depending on whether you used both gorget and belt or just one or the other, while (D+fSTR+WSC) are the same.
    9) no-pDIF Damage from Step 6 above divided by no-pDIF Damage from Step 8 gives you ( fTP/(fTP+.1 or .2) ).
    Example:
    Step 6 / Step 8 = fTP/(fTP+.1) = .91
    fTP = .91 fTP + .091
    .09 fTP = .091
    fTP = 1

    Determining fTP at different TP values:
    10) Repeat steps 7-9, but instead of using a gorget, instead just the amount of TP you want to test. So you'd WS at 200 TP, find max damage, remove pDIF, take a ratio, and solve for your new fTP using the fTP you determined for 100TP in Step 9.
    Example:
    Step 6 / Step 10 = fTP[100TP]/fTP[200TP] = .5
    fTP[100TP] = 1 <-- from Step 9
    1 / fTP[200TP] = .5
    fTP[200TP] = 2

    Determining modifiers:
    11) You're going to need to be careful about which stats you're changing, but changing multiple stats (in one direction) at once or two stats (in opposite directions) is a good way to screen for modifier stats. Really be careful and keep in mind what you're doing. You will want to vary each stat you're testing by at least 12, and the more the better.
    12) First, take off some of the STR gear you put on earlier (but still be sure to keep fSTR capped: aim for ~4*(FLOOR(D/9)+10) for non-ranged)
    13) Check if this changes anything with steps 3-6. You can be a little more informal here, as it should be pretty obvious pretty quickly whether or not it mattered. Maybe a dozen weaponskills to be safe. This works as your test for whether or not STR is a modifier.
    14) Try varying stats using the gear you picked up in step 11. Keep repeating steps 3-6 informally. You will want to vary each stat you're testing by at least 12 (minimum mod I can remember is 10%), and the more the better.
    15) When you've found the stat types that you think are the modifiers, pile on as much of one of them as you can get (while holding the other constant) and write down your stats again.
    16) Repeat steps 3-6 and determine your new base damage with these stats.
    17) Repeat for all of your suspected modifiers.
    18) Now you should know the fTP of your weaponskill at 100TP, so you can divide it out of all the numbers you're working with and be left with (D+fSTR+Modifiers). You know D (the damage of your weapon) and fSTR (FLOOR(D/9)+8), so you can figure out how much base damage modifiers are contributing to the weaponskill.
    19) Do this for your base case (Step 6) and your Modifier+ cases (Step 16/17). Solve for the modifiers
    Example, Pyrrhic Kleos (20% STR, 30% DEX)
    Case A - 120 STR, 75 DEX from Step 6 : 38 Base damage from modifiers
    Case B - 80 STR, 75 DEX from Step 13 : 31 base damage from modifiers
    Case C - 80 STR, 120 DEX from Step 16/17 : 43 base damage from multipliers
    +40 STR resulted in +7 base damage, so (7/45)/.85 = .206 = ~20%
    +45 DEX resulted in +12 base damage, so (12/45)/.85 = .31 = ~30%


    Mob Stats
    Ul' Hpemde Stats (Masamune + Cymmina[INT]) [Level 68-72]:
    Spoiler: show

    Ul'Hpemde lvl72
    • VIT=75
    • AGI=67
    • INT=55?
    • DEF=292
    • EVA=280

    Ul'Hpemde lvl71
    • VIT=75
    • AGI=67
    • INT=55
    • DEF=287
    • EVA=275

    Ul'Hpemde lvl70
    • VIT=73
    • AGI=67
    • INT=55
    • DEF=282
    • EVA=271

    Ul'Hpemde Lvl69
    • VIT=72
    • AGI=65
    • INT=54
    • DEF=277
    • EVA=265

    Ul'Hpemde Lvl68
    • VIT=71
    • AGI=65
    • INT=53
    • DEF=271
    • EVA=260




    Tests for Monster Stats
    Determining VIT of a monster, Method 1:
    Spoiler: show
    Ingredients:
    * A 1 base damage weapon of your choice
    * Lots of different types of STR gear ranging from negative gear to positive gear.
    * If you are unable to make yourself crit for 0, you may need more negative VIT gear or to change to a lower STR job.
    * If you are determining the VIT of something with an ambiguous check message and level range whose level you can calculate from the xp it gives, you will need to be able to kill it (obviously).

    Instructions:
    1) Find the monster you want to test the VIT of, start beating it with your D1 weapon, and adjust your STR until you hit for 0 on Crits.
    2) Take this amount of STR and add 4 to it to find the monster's VIT.

    Theory:
    By using a D1 (rank 0) weapon, your minimum fSTR is -1, which means you can swing with a weapon that has an effective base damage of 0. Critical Hits give you a minimum pDIF of 1, so if you think of damage like (Base Damage + fSTR)*pDIF, then the only way to get a 0 damage crit is to have a (Base Damage + fSTR) of 0. fSTR of -1 occurs when your STR is 14 points lower than your target's VIT.

    http://wiki.ffxiclopedia.org/wiki/ST...rence_Function


    Determining the Evasion of a monster:

    Spoiler: show
    Ingredients:
    * A large variety of Accuracy/DEX gear and potentially different weapon types

    Instructions:
    1) Find the monster you want to test the Evasion of and /check it while naked or wearing all your Accuracy/DEX down gear.
    2) If it checks "low evasion" change weapons to one with lower skill and try again.
    3) Put on Accuracy/DEX gear until you find the point where it changes from neutral to "low evasion"
    4) Calculate your own Accuracy using the instructions above, and subtract 10 from it. That is the monster's Evasion before level correction.
    4b) Level correction for Evasion is (Monster's level - Your level)*4

    Theory:
    /check gives a "Low Evasion" message when your hit rate is 80% or better. Your Hit rate is 75% when the target's Evasion is equal to your Accuracy, and 2 points of accuracy gives 1% hit rate. So finding the Accuracy you need to hit the monster exactly 80% of the time means you've found the monster's Evasion + 10. Hit rate can only take integer values.




    Common Math Terms:
    Spoiler: show
    FLOOR(x,g) = round down the value x to the significance level of g. When not defined, assume you're dropping the decimal. That's commonly how it's used.

    95% Confidence Interval = STD(data)*1.96; = The odds of the population mean falling within this interval is, shocker, 95%.
    STD(Data) = square_root( sum((x - average)^2) /n)

  2. #2
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Things that should be added to this:
    * How to check Monster INT (Nukes or Leaden Salute, I guess?)
    * How to check Monster AGI (Really not sure. I forget how we did this in the crit thread, if we did, apart from finding out the monster job/sub and back calculating.)
    * How to calculate and check Monster Defense (/check again? Set Attack = Defense?)
    * How to calculate and check Player Ranged Attack (Monster with known defense, calculate your attack based on the damage you do and pDIF formulas?)
    * How to determine which job combination monsters are (Figure out enough of their stats, check for job traits and MP?)
    * How to calculate monster stats in general (from their job/sub)
    * Confidence Interval calculations, along with some kind of narrative to make it more intuitive
    * Links to and explanation of how to set up kparser, along with the most recent memlock

    If you guys see anything wrong in anything I've written, or know a way it can be done better, please feel free to throw that in there. I'm done with this thread for the night, so have a ball! Delete things from my list in this post when you're done please.

  3. #3
    Masamune
    Guest

    For determining mob's stats, i have some data i already posted on wiki's corresponding discussion pages + a tool to "convert" my attack/accuracy /checks into VIT/AGI/Def/Eva Ranks for a given mob's family, which can be found under the "Mob Data" tab from my Excel dmg calc in sig.

    The idea is to use a "Mob's Families Stat Ranks" Table alongwith known Job Stat Ranks (same one used for player's stats) displaying rank letters ranging from A to F. Then from those tables could calculate ANY mob's stats in the game, by just inputting its type and its level.

    Note: i'll make another release of my dmg calc next week with a fix of the submentioned mob stats tool who got broken by inadvertance ;;

    Regarding Player Stats, i have an excel sheet with recorded Stats for all my jobs from lvl80 to 90 with all my available subjobs. i sent it to Motenten in hope he fixes his stat page, but unfortunately i never got an answer. I tried to find the new modifiers myself but i fails hard with Excel Solver...

  4. #4
    Chram
    Join Date
    Sep 2007
    Posts
    2,526
    BG Level
    7
    FFXI Server
    Fenrir

    Re: Stat calculator:

    Recovered from the dust bin, and went back to fix this problem that's been hanging over me for a while...

    Changes to the stat calculation formulas:

    Cor:
    Agi increased from C to B


    Rounding (to level 60):

    Each stat is composed of 3 portions: Race, Main job and Sub job. Each of these is calculated as a base value plus a scaling times (level - 1). Due to the fractional nature of the scalings, each of these will often end up with non-integer results (eg: 19.8).

    Prior to the last change, all values were floored before being added together. So, for example, an Elvaan 49 Dnc/Nin (one of Masamune's samples) would have Str values of 25.6 (race), 19.8 (main job) and 6.6 (subjob) for Str, which would be truncated to 25, 19 and 6, then added together for a total of 50. However the new value is 51.

    What occurs now is that the fractional values are allowed to contribute a bit more. Any element with a fractional value of at least 0.5 will add to a fraction pool. For the above, all three would thus add 0.5 to that pool, giving it a value of 1.5. If the total pool goes above 1, then that +1 is added to the total. We'd now have a str total of 51.

    Note that three high fractions such as the values above could, by themselves, add to over 2.0 worth for the fractional pool. Since the value has not increased by 2 points, we know that they aren't simply keeping all component fractions before flooring.

    Caveat: With the changes to the scaling values below, it's conceivable that the full fractions are kept, but that the /1024 fractions end up with less than full additional units. I can't (or rather, won't take the time to right now) verify this idea at this time, so will continue using the above formulation since it works pretty well.


    The above gives a perfect matchup for the 19 levels of sam/war (~30-49), 5 levels of dnc/nin, as well as a random sampling of other data provided for everying up to level 75. It appears to be cumulative with the segments calculated for race and main job between 60 and 75, and between 75 and 85 (each total being added to the fractional accumulator as +0.5 if 0.5 or higher on the fractional part).

    Levels over 85 diverge from this total. Implication is a new scaling value for levels 86+.

    Working through it, it looks like there isn't a change for 85+, but rather that the scaling for 76+ got changed from 0.35 to 0.39 (~400/1024).

    Also finding that some of the scalings for 60-75 aren't quite correct. Changing them to /1024 values with a couple tweaks seems to put everything in line, though I'm pretty sure it's not fully exact at this point. For the most part everything is exact, though, and the few that are off are generally only off by 1 point. I may do another revision at a later date.

    In any case, web site has been updated with the changes.

  5. #5
    Masamune
    Guest

    Aaaa nice to hear from you Motenten

    I was so sure you dropped the shit, glad to see i were wrong /psych.

    And yes, during my attempts to find those formulas myself, even if i failed, i got my "closest" results with /1024 format.

    Do you need more data for some particular job/sub/level combination ?

  6. #6
    Masamune
    Guest

    Quote Originally Posted by Byrthnoth View Post
    Things that should be added to this:
    1* How to check Monster INT (Nukes or Leaden Salute, I guess?)
    2* How to check Monster AGI (Really not sure. I forget how we did this in the crit thread, if we did, apart from finding out the monster job/sub and back calculating.)
    3* How to calculate and check Monster Defense (/check again? Set Attack = Defense?)
    4* How to calculate and check Player Ranged Attack (Monster with known defense, calculate your attack based on the damage you do and pDIF formulas?)
    5* How to determine which job combination monsters are (Figure out enough of their stats, check for job traits and MP?)
    6* How to calculate monster stats in general (from their job/sub)
    7* Confidence Interval calculations, along with some kind of narrative to make it more intuitive
    8* Links to and explanation of how to set up kparser, along with the most recent memlock
    I'll try to add methods i know for each points :

    For 1: easiest is Nuking Method, but cross-checking with ElementalWS is better for accuracy.
    For both methods, adding as much known multipliers like highest nukes + high MAB + HQstaff as you can ensures the accuracy of determined INT value, simply because the more flooring are involved the harder it gets to end up with an exact match to observed value, and also the higher base values are, the easier multiple multiplications in a row will add an error.
    Also, cross-checking with an ElementalWS is great especially in the case of mobs having MdB traits, elementalWS "equation" would be the 2nd alongwith the nuke one for a system having 2 unknown "INT" & "MdB" to determine. Personnally, i think Sanguine Blade is nice to use since it doesnot have a fTP type "Damage varies with TP", and a lot of jobs can use sword; but other WSs can work.

    1bis: to determine a mob's MND, i don't know yet an elemental WS having a dMND, but nuking with Banish3 works.

    For 2: 1st method and easiest is from evasion /check (and gets mob's evasion in same step). 2nd method is backcalculating from crit=f(dDEX) but is inaccurate if tester can't get a 40<dDEX<50, because @ dDEX<40, tiers covers large ranges of dDEX, meaning a lot of possible AGI values for one given measured critrate%, that method requires also large sample sizes.
    Examples of such calculations are shown on Colibri's discussion page @ wikipedia.org.

    For 3: same as for 2, easiest is defense /check (and gets mob's VIT in same step). Have to cross-check with BOTH "Low Def" AND "HighDef" tough, as sometimes can get multiple matching values.
    Another method, is backcalculating from crited ranged hits with highest baseDMG, using different str values; ranged WS + WSdmg% equips would be even better but requires building TP to 300% (as a lot of physical WSs have non-static fTP except @ 300%).
    3rd method is longest but would definitely add to our knowledge of calculating physical dmg: get as much normal hits @ pDIF ~around 1, then build a frequency chart to deduce from it the "spike" of frequency. The lowest dmg value corresponding to the leftmost edge of this spike should lead to fSTR => mob's VIT. That spike should cover 3-5 dmg values since that 2H update which added the "2ndary randomizer".

    For 4: yes but gotta be under uncapped conditions, and controlled distance.

    For 5 and 6: Mob's stats are determined from an equation looking a lot like for player stats, with 3 components (RaceRank+MainJobRank+SubJobRank). Job and Sub stats ranks being known (same as for player), tester just have to play with RaceRank function and letters A to F until you get an exact match with values observed from /check. Guessing the mob's job is indeed from observed trait effects (intimidations, or MdB when doing INT checks).

    For 7: good point, as a lot of testers often post their results with low sample sizes without knowing to which proportions their test is "inaccurate", sometimes don't need high sample size to rule out an hypothesis.
    The formula is SQRT[ (Parsed% * (1- Parsed%)) / SampleSize ] that will give an ERROR%. That ERROR% means that the Parsed % value you got is within a so-called "confidence intervall" = [Parsed%-ERROR% ; Parsed%+ERROR%].
    For example, if you parsed a critrate ~10% but the sample size yields ERROR%~15% => your parsed crit%=10% could be ANY value within the intervall [-5% ; 25%], no need to be Einstein to see that intervall is way too large to get your parsed crit% any credibility.
    Another counter example: ruling out the hypothesis as crit-based WSs start @ 0% crit% at 100TP. 500ish WSs is often enough to prove the value 0% is NOT inside the confidence intervall.

    For 8: yea would be nice if there were some wiki, or better some "infoballs" when hovering mouse above certain labels. "TotalHits" got me confused often as it doesnot include misses or retaliations (or something like that i don't remember lol).
    For setting it up, well there s not much to do except to find the "Options" button to input the memlock XD (and only when it changed before Motenten updates it himself on site)

  7. #7
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Not exactly a testing instrument, but I've recently found a utility that *really* helps me sort through my chat logs when I'm looking for stuff. Also, it can be used for ghetto parsing of things that KParser doesn't cover:
    http://astrogrep.sourceforge.net/

    It lets you search for a word or phrase in a folder of text documents. So I could take a log that I'm interested in, put it in a new folder, and search for something like "Byrth retaliates for" to count my number of retaliations. At the moment, I'm using it to find every time I've ever fought Glavoid just as a curiosity. Some of my files have over 20,000 mentions of the word "Glavoid"... must have been the long farming sessions. xD

  8. #8
    Radsourceful

    Join Date
    Jul 2007
    Posts
    1,938
    BG Level
    6
    FFXI Server
    Bismarck

    Alternatively, install cygwin and use normal grep + assorted other tools. ♥

  9. #9
    Salvage Bans
    Join Date
    Jul 2009
    Posts
    964
    BG Level
    5
    FFXI Server
    Leviathan

    Or SFU/SUA for the non-shitty tools. GNU tools are bad enough already, and then the cygwin lib makes them even slower.

  10. #10
    TSwiftie
    Join Date
    Oct 2005
    Posts
    1,920
    BG Level
    6
    FFXI Server
    Fenrir

    I'm a big fan of Notepad++. Aside from being a great text editor for tons of stuff including scripting/coding, it's been absolutely amazing for searching my log files:

    http://img5.imageshack.us/img5/3888/notepadb.png

    Supports Regex, can just do a simple count, will display all the lines for you, find in a document, multiple documents, or an entire directory. It's just incredibly easy to use, and very intuitive. And it's free!

    http://notepad-plus-plus.org/

  11. #11
    First invited, last in the zone.
    Join Date
    Sep 2008
    Posts
    1,449
    BG Level
    6
    FFXI Server
    Lakshmi

    W7 actually will automatically let you search for a phrase inside a folder of text documents, without any external programs. Just include the folder in the list of indexed locations.

    It's how I update my macros when I get a new piece of gear that replaces an old piece.

    That said, notepad++ is pretty awesome. I love it's ability to copy/paste rectangles of text, rather than just continuous lines.

  12. #12
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Enmity Testing Paradigm (testing ambiguous enmity, like an Atma):
    1) Set yourself and a friend up so that you take 0 damage from whatever monster you're testing with and net-0 Enmity. Composure Phalanx/Stoneskin before the pull is ideal for this.
    2) Pull a monster (200 CE, 900 VE). Wait 15 seconds (for the VE to decay) and have your friend attempt to generate 200 CE (two Threnodies).
    3) Now the monster should go to your friend, assuming you each have the same amount of Enmity.
    4) Put on whatever Atma, gear, etc. you want to test.
    5) Do an action that should have fixed Enmity, like Etude for 160 CE.
    6) Have your friend do an action with a similar amount of Enmity, and then cast Cure 1 for 0 HP (1CE/cast) until he catches up to you and the monster shifts hate.
    7) Verify it by casting Cure 1 once for 0 HP and bouncing the monster back to yourself.
    8 ) Sum up the Enmity you each should have, based on (http://kanican.livejournal.com/21727.html)*, and figure out how much Enmity you actually got from your test action.

    *I'd avoid using things like Dispel, Sleep, and Sleep II. I don't think he has updated that since the Enmity nerf on those spells. Dispel is a real loss for this kind of testing, but Sleeps have never been good for it just because you can't tell who is at the top of the mob's hate list when he's asleep. =p

  13. #13
    Kaeko
    Join Date
    Mar 2006
    Posts
    1,974
    BG Level
    6
    FFXI Server
    Odin

    Quote Originally Posted by Byrthnoth View Post
    I'd avoid using things like Dispel, Sleep, and Sleep II. I don't think he has updated that since the Enmity nerf on those spells. Dispel is a real loss for this kind of testing, but Sleeps have never been good for it just because you can't tell who is at the top of the mob's hate list when he's asleep. =p
    The values are updated past the RDM spell enmity nerf so you can use them. I think the main page should still have a "last updated" date on there to give you a better sense if there are future enmity changes later. Glad to see someone continuing the research~

  14. #14
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Heh =p Yeah, I found the RDM main page after I said that. The "Enmity Table (Full)" page still has pre-patch values.

    Random aside, but could you possibly project a level 90 coefficient for Healing Magic CE? All jobs that get Cure also get Tranquil Heart now, so the only job I can test the trait potency on is WHM with Cure V unless I have that value.

    Edit: Also, do you mind if we pull all your Enmity values onto BGwiki some day? I doubt we'll get around to it soon, but it'd be a good resource to have.

  15. #15
    Kaeko
    Join Date
    Mar 2006
    Posts
    1,974
    BG Level
    6
    FFXI Server
    Odin

    My old issue with putting stuff on the wiki was I would ninja edit stuff a lot and didn't want 2 sources saying conflicting things. Since I don't update anymore it would probably be easier to have everything on a wiki as opposed to my own site. If you're willing to do the work I have no problems at all.

    The enmity table (full) probably should be taken down - haven't updated that one since like '08 since I realized putting every value on there would exceed the size limit LJ allows. I would look strictly at the job tables.

    For CE on cures, the projected pattern (if there is no change) would be an increase of 1 in the denominator for each increase in level. So at 75 it's 40/55, 76 it's 40/56, and so on. That would project lvl90 being 40/70 CE per 1 HP cured, or a 0.5714 ratio. This is a pretty big extrapolation though.

  16. #16
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Hmm.. Looking at your table, the denominator only increases by 6 for every 10 levels. (http://kanican.livejournal.com/30142.html)

    Still, how does this sound?:
    1) Venom Potion, RDM/BRD with 0 Enmity (-300HP)
    2) I pull a level 0 mob with Dispel (201 CE)
    3) WHM mule cures me for 300 HP (Something like 155-156 CE if the extrapolation worked, including Tranquil Heart, or 196 if it hasn't changed since 75)
    4) Cure 1 for 0 HP until the monster switches to the White Mage.

    It's not as "clean" as just doing the test without Tranquil Heart at all, but it should work.

    Verified it using the above test. It follows the trend in your table (6 per 10 levels, 9 per 15 levels) The coefficient for level 90 is (40/64), or 5/8.

    300*N*.91 + 44 CE = 201 + 13 CE
    300*N*.91 = 170 CE

  17. #17
    Kaeko
    Join Date
    Mar 2006
    Posts
    1,974
    BG Level
    6
    FFXI Server
    Odin

    Quote Originally Posted by Byrthnoth View Post
    Hmm.. Looking at your table, the denominator only increases by 6 for every 10 levels. (http://kanican.livejournal.com/30142.html)

    Still, how does this sound?:
    1) Venom Potion, RDM/BRD with 0 Enmity (-300HP)
    2) I pull a level 0 mob with Dispel (201 CE)
    3) WHM mule cures me for 300 HP (Something like 155-156 CE if the extrapolation worked, including Tranquil Heart, or 196 if it hasn't changed since 75)
    4) Cure 1 for 0 HP until the monster switches to the White Mage.

    It's not as "clean" as just doing the test without Tranquil Heart at all, but it should work.
    Ah sorry it's been too long. The pattern is like +1, +1, +0. So you just follow that. The reason I used to post exact modifiers was so people could see the pattern.

    As far as method, looks ok but I have no idea what tranquil heart is.

  18. #18
    BG Content
    Join Date
    Jul 2007
    Posts
    16,848
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Quote Originally Posted by Kaeko View Post
    Ah sorry it's been too long. The pattern is like +1, +1, +0. So you just follow that. The reason I used to post exact modifiers was so people could see the pattern.

    As far as method, looks ok but I have no idea what tranquil heart is.
    Tranquil Heart is a Job Trait SE added that reduces enmity from healing magic. All jobs that get healing magic get the trait starting at a low level. In your opinion, is Enmity applied as a percentage or a /256 fraction?

    At the moment, I'm getting this as my Enmity equation:
    Floor( Floor([Cure HP] * [Cure Modifier])*(-Enmity%) )

    So Curing myself for 390 gave:
    Floor( Floor(390*5/8 )*.93 ) = Floor( Floor(243.75)*.93 ) = Floor(225.99) = 225 Enmity

    Curing myself for 300 gave:
    Floor( Floor(300*5/8 ) *.93 ) = Floor( Floor(187.5)*.93 ) = Floor(173.91) = 173 Enmity

    Both results would have been one higher if there weren't two floor steps.

    Edit: By the way, Red Mage seems to get a lower tier of Tranquil Heart than White Mage. Hence the .93 instead of .91.

  19. #19
    Sea Torques
    Join Date
    Dec 2007
    Posts
    556
    BG Level
    5

    Has anyone attempted to valid the use of enspells for checking magic accuracy bonuses? Even if enspell trials are spread out across mobs of different level it can be controlled for statistically. The main issue is whether any results based on enspells can be extended to other spell types and a lesser one concerns the nature of enspell damage outcomes in terms of resists (e.g., 3 vs 4 levels of resist).

  20. #20
    Masamune
    Guest

    pretty great idea there CDF, using enspells instead of nuking would indeed get tester lot higher sample size.

+ Reply to Thread
Page 1 of 4 1 2 3 ... LastLast

Quick Reply Quick Reply

  • Decrease Size
    Increase Size
  • Remove Text Formatting
  • Insert Link Insert Image Insert Video
  • Wrap [QUOTE] tags around selected text
  • Insert NSFW Tag
  • Insert Spoiler Tag

Similar Threads

  1. What needs to be tested thread
    By Yugl in forum FFXI: Mathy Parts Redux
    Replies: 173
    Last Post: 2016-05-24, 00:31
  2. Test Data Compilation Thread
    By Yugl in forum FFXI: Mathy Parts Redux
    Replies: 11
    Last Post: 2011-08-19, 13:33