1. FFXIV Reset Timers
    Last daily reset was 21 hours, 35 minutes ago / Next daily reset is in 2 hours, 24 minutes
    Last weekly reset was 1 days, 14 hours, 35 minutes ago / Next weekly reset is in 4 days, 19 hours, 24 minutes
+ Reply to Thread
Page 1 of 2 1 2 LastLast
Results 1 to 20 of 23
  1. #1
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Critical Rate Testing (And other stuff?)

    This advanced theorycrafting section isn't advanced enough; It needs more data dumps and ffxi style math theorycrafting without all the deepz mumbo jumbo.

    This thread is mostly going to be for dumping my critical hit testing, and with it should come data that can also be used for damage formula calculations. Dumps from other people are more than welcomed and I'll be editing the OP to include them. I'm a paranoid asshole when it comes to testing conditions though, so I'll be treating them warily.

    I'll be keeping the data in this post rather condensed for now, and the google doc linked will contain the more detailed data. The things I'm keeping track of personally are Class and level, wDMG, Str/Dex, Acc, Crit, Det, Move, Used, Hit, Critical Hit, Miss, Low, High, Critical Low, and Critical High. My testing procedure is rather simple: I just attack striking Dummies in Whitebrim. The parser I'm using is FFXIVAPP.

    One of the detriments to afk testing against dummies is that anybody can come along and put debuffs on your target. This can easily skew damage testing, but thankfully highs and lows can be accurately confirmed with a rather low sample size, so double checking is easy. Also thankfully, as far as I know, there shouldn't be any debuffs that would increase or decrease my critical hit rate.

    I also want to test if higher level enemies give you lower critical hit rates. Actual crit % should work in one of three ways: Either a given "critical hit rate" will always correspond to the same crit % regardless of mob level, or mobs of different levels will give you different crit %s at given critical hit rates, or a given critical hit rate will result in different crit %s depending on the delta between your level and the mobs level.

    Code:
    Test # | Class   | Stex | Crit | Det | Used  | C.Hit | 95% Range
    01     | 50 Bard |   ?  | 500  |  ?  | 5471  |  925  | 15.9 - 17.9
    02     | 50 Bard |   ?  | 541  |  ?  | 7514  | 1466  | 18.6 - 20.4
    03     | 50 Bard |   ?  | 547  |  ?  | 16802 | 3406  | 19.7 - 20.9
    04     | 50 Bard |  492 | 585  | 261 | 32633 | 7387  | 22.2 - 23.1
    05     | 50 Bard |  381 | 380  | 202 | 20466 | 1571  | 07.3 - 08.0
    06     | 50 Bard |  382 | 457  | 237 | 34106 | 4605  | 13.1 - 13.9
    07     | 50 Bard |  424 | 474  | 254 | 16383 | 2482  | 14.6 - 15.7
    08     | 50 Bard |  453 | 518  | 219 | 33649 | 5931  | 17.2 - 18.0
    09     | 50 Bard |  469 | 473  | 281 | 16178 | 2228  | 13.2 - 14.3
    10     | 50 Bard |  469 | 494  | 245 | 34290 | 5579  | 15.9 - 16.7
    Google Docs Link: https://docs.google.com/spreadsheet/...FE&usp=sharing

  2. #2
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Going to focus this post on information extrapolated from the data. Maybe eventually I'll even throw in some graphs and shit.

    With the conclusion of the fifth test, we have three rather reliable data points. A range of 1.2% at 547 crit, a range of 0.9% at 585 crit, and a range of 0.7% at 380 crit. All at 95% confidence. Assuming crit scales linearly, we can gather that you will gain 1% crit rate for approximately every 12~15 critical hit rate. (12.8125 to 14.8550, using 99% confidence on tests 4 and 5.)



    Y axis is crit %, X axis is critical hit rate, Low/High ranges are the low and high ranges using a 95% confidence interval. Higher sample size results in less space between the two lines. Pretty stuff is a two data point line from my lowest crit data point to my highest crit data point, using the measured/avg crit%.

  3. #3
    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

    I'll PM you the AHK request.

  4. #4
    Ridill
    Join Date
    Oct 2006
    Posts
    18,451
    BG Level
    9
    FFXIV Character
    Sath Fenrir
    FFXIV Server
    Cactuar
    FFXI Server
    Fenrir

    Would like to see all tests have the rigor of test 3. The range of the % really benefits from the larger sample size.

  5. #5
    D. Ring
    Join Date
    Nov 2011
    Posts
    4,998
    BG Level
    7
    FFXIV Character
    Raldo Volca
    FFXIV Server
    Balmung

    What is "Stex"? Did you mean DEX?

  6. #6
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Thanks Yugl, for the AHK script. I've probably looked at that function dozens of times but never thought it did what I wanted lol. Now I'll be able to run tests when watching anime, reading forums, and while doing all the other menial bullshit that kept me from having xiv at the foreground.

    Quote Originally Posted by Sath
    Would like to see all tests have the rigor of test 3. The range of the % really benefits from the larger sample size.
    Yeah, the goal sample size I'm shooting for with each test is enough to only have a 1% range at 95% confidence. This should mean sample sizes of around 15k~20k depending on the crit rate.

    What is "Stex"? Did you mean DEX?
    Str/Dex, I was lazy and didn't like the formatting of the pseudo-table with it spelled out.

  7. #7
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Updated the google doc to include test 6. Managed to bring my accuracy low enough to cause misses since I was only attacking from the front, lol.

    Any requests for a crit rate value I should test at next?

  8. #8
    Ridill
    Join Date
    Oct 2006
    Posts
    18,451
    BG Level
    9
    FFXIV Character
    Sath Fenrir
    FFXIV Server
    Cactuar
    FFXI Server
    Fenrir

    Smaller intervals between crit values and then a graph would be nice (just take value as avg of min and max in interval)

  9. #9
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Added another test. And a crappy graph with no axis lables, courtesy of excel.

  10. #10
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Aaaand another test. Added another line to the graph also. Can honestly probably just use that line to estimate your crit% for now.

  11. #11
    Smells like Onions
    Join Date
    May 2014
    Posts
    1
    BG Level
    0

    Would it be possible to get a naked test to find out what the base crit rate is? If crit follows HP and what is proposed for parry and the like, the formula has some floor that may or may not be job specific.

    Currently your 12-15 Crit rate = 1% crit would predict that at 581 you'd have over 40% crit rate, which you clear don't. (If I'm, not completely misreading the data).

    e.g.
    HP = (85*JobConstant)+(14.5*(Vit-202)

  12. #12
    The Once and Future Wamoura
    Join Date
    Aug 2005
    Posts
    18,373
    BG Level
    9
    FFXIV Character
    Rocl Montaigne
    FFXIV Server
    Excalibur
    FFXI Server
    Bahamut
    WoW Realm
    Quel'Thalas

    I did some testing with curative spells on Scholar a bit ago, so this is 2.0 data but fwiw here are the results (it's a fairly small test, with 500 casts for what you want)
    Quote Originally Posted by Rocl View Post
    I heard a rumour that Adloquium's crit rate caps at 8%. I decided to do some critical tests to see how this all pans out, but I figure I'll post what I have now. Probably gonna do 500 Adloquiums and 500 Physicks cuz I'm bored as hell. Both tests will be done with 507 critical hit rate at Lv50 casting on myself.

    tests with 507 crit
    500 casts of Adloq, 86 critical cures; 17.2% crit rate
    500 casts of Physick, 78 critical cures: 15.6% crit rate
    500 casts of single-target Succor, 96 critical cures, 19.2% crit rate
    200 casts of Embrace, 46 critical cures, 23% crit rate [Eos]
    200 casts of Embrace, 8 critical cures, 4% crit rate [Selene]

    So no, there isn't a crit penalty on Adloquium.

    Since I'm equally bored gonna do some tests with the base crit rate of 341 and then see if crit rate affects pets (which I'm willing to be good money it does)

    tests with 341 crit
    250 casts of Adloq, 9 critical cures, 3.6% crit rate
    250 casts of Physick, 9 critical cures, 3.6% crit rate
    100 casts of Embrace, 6 critical cures, 6% crit rate [Eos]
    100 casts of Embrace, 3 critical cures, 3% crit rate [Selene]

    two hours later i got bored, so w/e
    most interesting thing is i'm pretty fucking certain Eos is #1 crit rate
    It's a super shitty test I did one afternoon, so it's not very useful, but eh, a decent starting place if nothing else. I had mainly wanted to nip the idea that Adloquium and Physick had vastly different crit rates in the bud before it spread further.

  13. #13
    E. Body
    Join Date
    Jan 2011
    Posts
    2,283
    BG Level
    7
    FFXIV Character
    Warren Castille
    FFXIV Server
    Balmung
    FFXI Server
    Ragnarok
    WoW Realm
    Cenarion Circle

    200 casts of Embrace, 46 critical cures, 23% crit rate [Eos]
    200 casts of Embrace, 8 critical cures, 8% crit rate [Selene]
    Am I missing something for 8/200 being 8%?

  14. #14
    The Once and Future Wamoura
    Join Date
    Aug 2005
    Posts
    18,373
    BG Level
    9
    FFXIV Character
    Rocl Montaigne
    FFXIV Server
    Excalibur
    FFXI Server
    Bahamut
    WoW Realm
    Quel'Thalas

    HUMAN ERROR. Yugl just PM'd me that ACT is/was shoddy counting cures and that's what I did the majority of the test with so RIP trying to contribute to the zeitgeist

    last time i try to do anything math related, back to reading books and putting too much thought into video game storylines

  15. #15
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    Quote Originally Posted by Tazmaliah View Post
    Would it be possible to get a naked test to find out what the base crit rate is? If crit follows HP and what is proposed for parry and the like, the formula has some floor that may or may not be job specific.

    Currently your 12-15 Crit rate = 1% crit would predict that at 581 you'd have over 40% crit rate, which you clear don't. (If I'm, not completely misreading the data).

    e.g.
    HP = (85*JobConstant)+(14.5*(Vit-202)
    I'm assuming it's 5% at 341, which would mean that (581-341)/~14 = ~17; 17+5 = 22. There's definitely a linear gain, it's just has a biased starting percentage so you can't do a straight division without any subtraction to get a crit%. I'll see if I can do a nocrit+ test today.

    Rocl, while those healing tests are low sample size they show that it's likely 'critical hit rating' results in the same crit% for both attacks and heals.

    I haven't used ACT for any of these tests yet, since I like the more easily readable format of ffxivparser for this type of stuff. Maybe I'll try running them both at the same time today and see what happens lol.

  16. #16
    The Once and Future Wamoura
    Join Date
    Aug 2005
    Posts
    18,373
    BG Level
    9
    FFXIV Character
    Rocl Montaigne
    FFXIV Server
    Excalibur
    FFXI Server
    Bahamut
    WoW Realm
    Quel'Thalas

    Hi if you want to hear another retarded bag of failure, those tests were done w/ XIVAPP. I just default to saying ACT since that's what I use now but in the year of our lord 2013 I used XIVAPP. gj me 8/200 = 8%

  17. #17
    Yoshi P
    Join Date
    Aug 2006
    Posts
    5,141
    BG Level
    8
    FFXIV Character
    Dead Gye
    FFXIV Server
    Lamia
    FFXI Server
    Ragnarok

    So I started doing some det/dex tests... and I ended up doing a 16k sample test with 473 crit; and in the OP you can see I have a 16k sample test at 474 crit already done.

    07 | 50 Bard | 424 | 474 | 254 | 16383 | 2482 | 14.6 - 15.7
    09 | 50 Bard | 469 | 473 | 281 | 16178 | 2228 | 13.2 - 14.3

    If I use a 99% confidence margin instead of 95% then the ranges are 14.4-15.9 and 13.1 - 14.5 for tests 7 and 9 respectively. This either means I've narrowed down the 473/474 critical hit rate value to 14.4-14.5% critical hit rate or that critical hit rate -> crit% works like a step function and I stumbled upon one of the threshholds. If you cross reference the graph I made and posted though, you'll notice the 'pretty stuff line', that's just a straight line from the avg lowest crit% test I did to the avg highest crit% test I did, crosses very very very close to the low range of test 07, so we probably just narrowed down that crit% to a ~0.1% range.

    I might as well do a base crit test after test 10 so I can get a better looking graph. Or just make assumptions and redo the pretty line using 5% at 341.

  18. #18

    Quote Originally Posted by Rocl View Post
    Hi if you want to hear another retarded bag of failure, those tests were done w/ XIVAPP. I just default to saying ACT since that's what I use now but in the year of our lord 2013 I used XIVAPP. gj me 8/200 = 8%
    I love you, don't ever change.

  19. #19
    Salvage Bans
    Join Date
    Feb 2014
    Posts
    989
    BG Level
    5
    FFXIV Character
    Harold Saxon
    FFXIV Server
    Ragnarok

    Is there any recent data that has indicated crit having large gaps between breakpoints and having diminishing returns?

    Or is it idiots trying to justify their builds without data?

  20. #20

    No data that I have ever seen or heard of has shown crit rate to be anything but linear.

    Unless someone's got actual data of their own, they're talking out of their ass.