Item Search
     
BG-Wiki Search
+ Reply to Thread
Results 1 to 16 of 16
  1. #1
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Ranged Critical Hit Rate Testing

    The relationship between dDEX and Critical Hit Rate was worked out on BG ages ago, but there have been many anecdotal reports over the years that ranged critical hit rate is not capped in situations when you would expect it. If I'm remembering right, Kirschy, Nightfyre, and Prothescar have all reported something to this effect.

    This issue was recently revived by Melphina from Alla, who demonstrates that Ranged Critical Hit Rate is uncapped despite having dDEX>90. Then he demonstrated that raising his AGI to 174 and ignoring DEX increased critical hit rate. I'm currently busy irl, so afk-testing something like this is perfect. Preliminarily I've come up with these questions to answer:

    Question 1) Does Ranged critical hit rate depend on dAGI (Player AGI - Monster ???) or is it purely a function of AGI?
    Rationale:
    Melphina's test shows that increasing AGI increases critical hit rate, but there are also some indications that dSTAT might not be involved, considering that 174 AGI is 77 AGI more than the monster's AGI (which should be only a few points lower than its DEX). That's quite substantial.

    Question 2) Is Ranged critical hit rate affected by Critical Hit Rate merits?
    Rationale:
    Melphina's first test shows an 11% critical hit rate. This is suspiciously low considering the 10% base, which means he may have been floored in his first case if Crit rate merits matter. Short test and probably worth doing.

    Question 3) Does range preference affect critical hit rate (True vs. Squarely vs. no message)?
    Rationale:
    SE varies Ranged Accuracy (I think) and Ranged Attack (for sure) with distance from the target depending on weapon type. It's not entirely unimaginable that they would also vary ranged critical hit rate.

    Question 4) What is the mapping of AGI or dAGI to Ranged Critical Hit Rate?
    Rationale:
    This is pretty much the final question and the one that will be the most useful to answer.




    Method:
    This can be tested using one of the ToAU chakrams (Mamoolbane, Trollbane, or Lamiabane) and Hpemdes. First, you gear yourself so that you hit for 0 damage all the time (fSTR = -1) with a D1 main hand (like Trainee Needle) and the chakram. This requires that your STR be 16 less than the target's VIT. Then you line up with the monster, melee it with a low delay weapon, and set up a script to throw a chakram all night while you parse it.
    Om'hpemde - Level 73-77
    Ul'hpemde - Level 68-72
    It's helpful to use /check so you know what level (and thus what stats) the Hpemde you're hitting has. /check gives "High Defense" when you have less attack than the target's Defense or "Low Defense" when you have 25% more attack than the target's defense, with a neutral message between those two points. I typically set my Attack at 315-318 and go around /checking Hpemdes when I want to isolate one of the level 77 ones.

    Here is an example script:
    Code:
    alias shot input /ra;wait 6;
    alias min shot;shot;shot;shot;shot;shot;shot;shot;shot;shot;
    alias five min;min;min;min;min
    alias thirty five;five;five;five;five;five;
    alias twohour thirty;thirty;thirty;thirty;
    alias tenhour twohour;twohour;twohour;twohour;twohour;
    You would save that in shots.txt in your scripts folder, run it in game with //exec shots.txt, and then start shooting every 6 seconds for 10 hours with //tenhour

    Tips:
    * Let the Hpemde walk away from you once to see how it paths, so you can be sure that it won't walk out of your /ra sight if it deagros.
    * Try to stick near a wall on land so you avoid uninvited Yovra/Phaubo attention.
    * You're unlikely to be able to keep Hpemde in "True" hit range while afk until they glitch out and stick to you no matter what. Their "oh hey I'm going to wander away now" AI glitches out after a few hours of re-claiming them and they just stick to you no matter what, at which point you could do this test. Preliminarily, stick to "Squarely" as your range.



    My plan at the moment is to parse against the lowest level Hpemde and then the highest level Hpmede, and then I'll put on all the AGI I can muster and re-parse against the highest level Hpmede. These tests take a long time because it's 1 shot every 6 seconds, so I don't expect this will get done soon.

    Anyone have other questions/possibilities?

  2. #2
    The Syrup To Waffles's Waffle
    Join Date
    Jun 2007
    Posts
    5,053
    BG Level
    8
    FFXIV Character
    Cair Bear
    FFXIV Server
    Excalibur
    FFXI Server
    Fenrir

    Without trying to over-complicate things before this even gets started, just based on your questions I imagine it is worth considering that range preference and AGI could interact dynamically with ranged critical hit rate instead of just independently if they even do at all. It wouldn't be my first assumption by any means, just figured it may be something to keep in mind.

  3. #3

    Edit: Nvm, I'm tired and fail at reading.

  4. #4
    CoP Dynamis
    Join Date
    Jun 2011
    Posts
    284
    BG Level
    4
    FFXI Server
    Fenrir

    I also still think it's worth testing whether or not Ranged WS use the same critical hit rate formula as regular ranged attacks. For whatever reason (lol coding) ranged WS could be forced to use the dDEX formula.

  5. #5
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Stats:
    Spoiler: show
    Level 90 DNC / 1 WHM
    55 STR
    84 DEX
    63 VIT
    77 AGI
    91 INT
    61 MND
    75 CHR


    Level 68 Ul'Hpemde (confirmed via /check) vs. a Taru 90DNC/1WHM
    5105 hits -> 566 crits
    11.09% +/- 0.86% (95% confidence interval) Crit rate
    : Started at "True" and ended at "Squarely"

    Level 77 Om'Hpemde (confirmed via /check) vs. a Taru 90DNC/1WHM
    5311 hits -> 516 crits
    9.72% +/- 0.8% Crit rate
    Likely 10%
    : Squarely the whole time (3.1 from Distance)

    Level 77 Om'Hpemde (same as the previous one, now stuck to me) vs. a Taru 90DNC/1WHM
    4834 hits -> 471 crits
    9.74% +/-0.84% Crit rate
    Likely 10%, meaning that it was probably floored again and True vs. Squarely did nothing. This does not exclude a possible interaction of range and dSTAT because dSTAT was obviously floored here.
    : True the whole time (2.0 from Distance)

    Level 77 Om'Hpemde (same as the previous one) vs. a Taru 90DNC/1WHM with 92 AGI "Squarely"
    5445 hits -> 657 Crits
    12.06% +/- 0.87% Crit rate
    12% Critical Hit Rate. 3.1 distance the whole time, which is "Squarely" for Mamoolbane.

    Level 77 Om'Hpemde (same as the previous one) vs. a Taru 90DNC/1WHM with 92 AGI "True"
    5445 hits -> 657 Crits
    11.96% +/- 1% Crit rate
    Very likely 12% Critical Hit Rate. 2.0 distance the whole time, which is "True" for Mamoolbane. Given that I cannot exclude small dSTAT changes based on this test because I don't know if I'm on a critical point, I chose to end it here. Still, I'm going to default to doing all my future tests at "Squarely" distance because it very likely doesn't matter and Squarely is easier.

    Level 77 Om'Hpemde (same as the previous one) vs. a Taru 90DNC/1WHM with 81 AGI
    1260 hits -> 117 crits
    9.28% +/- 1.6% Crit rate
    Because 10% is likely the floor and we exclude 11% at 95% confidence, this AGI value is actually 10%. Hooray for luck.

    Level 77 Om'Hpemde (same as the previous one) vs. a Taru 90DNC/1WHM with 83 AGI
    3319 hits -> 339 crits
    10.21% +/- 1% Crit rate
    Likely 10%, though 11% has not been excluded at 95% confidence. The Om'hpemde ended this test for me (SS below). I'll go back to it some day if necessary.

    Level 77 Om'Hpemde (same as the previous one) vs. a Taru 90DNC/1 WHM with 114 AGI (Stormsoul x2, Breeze Pearl x2, AF3+2 hands, Utsu neck, Scouter's Rope, Desultor Tassets)
    4636 hits -> 607 crits
    13.1% +/- 0.97% Crit rate (still running)
    Likely 13%. The fact that I'm seeing such a minor increase in critical hit rate with a large AGI jump (+22 AGI = +1% crit rate) implies that I was probably early in the 12% tier above and am late in the 13% Tier now. It also implies that the crit tiers here are freaking huge, considering that I'm increasing dSTAT by 22 and only seeing a 1% increase.

    With a little more confidence:
    1) Yes
    2) Yes
    3) Probably no boost for being in the sweet spot, though dSTAT modification has not been rigorously excluded
    4) Something like dAGI/10 works right now (Player AGI - Monster AGI)
    ...but it doesn't confirm any of them. I need to run control tests!

  6. #6
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Well, first time this has ever happened:


    GM MPK! GM MPK!

  7. #7
    The Syrup To Waffles's Waffle
    Join Date
    Jun 2007
    Posts
    5,053
    BG Level
    8
    FFXIV Character
    Cair Bear
    FFXIV Server
    Excalibur
    FFXI Server
    Fenrir

    That is indeed peculiar.. o_o

  8. #8
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    Another great disturbance in the force, as if an afk taru cried out in anguish and then was suddenly silenced:
    Spoiler: show


    I don't know what they changed, but they changed something.

    I have more data to post, but an interesting sample will be finished tomorrow so I might as well wait.

  9. #9
    Relic Horn
    Join Date
    Sep 2011
    Posts
    3,153
    BG Level
    7
    FFXI Server
    Leviathan

    It distinctly has something to do with your attacks. I was 'engaged' with an Om'hpemde for two days straight casting songs on it with no issue

  10. #10
    An exploitable mess of a card game
    Join Date
    Sep 2008
    Posts
    13,258
    BG Level
    9
    FFXIV Character
    Gouka Mekkyaku
    FFXIV Server
    Gilgamesh
    FFXI Server
    Diabolos

    Duration between beginning attacks and attacking?

  11. #11
    Relic Horn
    Join Date
    Sep 2011
    Posts
    3,153
    BG Level
    7
    FFXI Server
    Leviathan

    That's what I was thinking. Alternatively, it may be the result of swinging too quickly after a deaggro window. Neither of these would likely show up in the ways we've been using them for most of their existence

  12. #12
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    On the contrary, I've spent several hours autoattacking with the exact weapon I was using under the exact same condition as the second instance of this when I was doing delay constant testing. You'll note that my script had run out and I was no longer /ra'ing.

    The first example triggers an attack following a /ra for 0 damage (2 seconds before he misses for the first time)
    The second example triggers an attack following a normal hit for 0 damage (2 seconds before I parry for the first time)

  13. #13
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1


    Spoiler: show



    As you can see from the data table, some of the values have not reached 95% significance. The column that is labeled "Variance" is actually 95% confidence interval width, my bad! Still, the basic story I'd propose is:

    Ranged Crit Rate = FLOOR( (AGI-AGI)/1000, 0.01 ) + Base rate + Merits + Gear

    It looks like the slow part of the dDEX growth, but it never gets a steep transition. Also, it could be AGI-DEX (without changing the equation) or really AGI-anything (if you changed the equation). I've just proposed AGI to mirror the melee critical hit equations.

  14. #14
    Chram
    Join Date
    Sep 2007
    Posts
    2,526
    BG Level
    7
    FFXI Server
    Fenrir

    What is AGI-MAX? Looks like it's supposed to be AGI-STR?

    But yeah, I'd agree it looks like +1% crit rate per +10 (Agi-Agi). Even the two you marked as "NO" are within margin of error for that equation.

  15. #15
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    I just used the max stat calculated from Masamune's DRG/DRG projection, but we know the hpmeses do not have dragon killer so it doesn't really work.

    The values are within margin of error. I just used the closest value for the projected real value. It is also worth noting that AGI = DEX for this job combination.

  16. #16
    BG Content
    Join Date
    Jul 2007
    Posts
    19,967
    BG Level
    9
    FFXI Server
    Lakshmi
    Blog Entries
    1

    I did two more points at 161 and 162 AGI, but I had to use a second Athos's piece to get there. They were 21% and 22% crit rate respectively, so 18% and 19% without the set bonus. This is consistent with (AGI - AGI)/10 still. At the moment I am doing one more sample at 151 AGI to confirm another boundary. I figure there are two tests left that I still want to do:
    1) Set one of my mules up on DRK spamming Absorb-AGI with Liberator 99 while the other one is exactly on a boundary. If I have him sub RDM and Haste himself/wear gear, I bet I can get the recast down to 30 seconds (which would mean I can maintain about -30 stat and hopefully see a 3-4% crit rate difference if it is AGI-AGI.
    2) Figure out the absolute most AGI I can scrape together in one set and do it to see if there is a steep slope/cap to this somewhere.