Results 1 to 10 of 52

Hybrid View

  1. #1
    Player
    Lastelli's Avatar
    Join Date
    Jun 2015
    Location
    Gridania
    Posts
    937
    Character
    Lastelli Sungsem
    World
    Ragnarok
    Main Class
    Monk Lv 90
    My two cents: using a simplistic deviation measure between arbitrarily chosen percentiles in order to estimate the difficulty of a job is meaningless if we don't include any additional rDPS gain due to utility, especially if we're talking about NIN since optimal use of TA and SW is what really makes a great NIN.

    And this still wouldn't solve another important issue: since most hc parties do not include a MNK and MNK has a little more than half the number of parses that NIN has, we can reasonably assume that there's a higher % of "skilled" NIN than "skilled" MNK on fflogs parses. So we should also check the skewness of the distributions of mnk and nin parses and this still wouldn't give the whole picture.
    (0)

  2. #2
    Player
    SpeckledBurd's Avatar
    Join Date
    Oct 2016
    Location
    Ul'dah
    Posts
    715
    Character
    K'ahli K'uhla'tor
    World
    Balmung
    Main Class
    Monk Lv 100
    Quote Originally Posted by Lastelli View Post
    My two cents: using a simplistic deviation measure between arbitrarily chosen percentiles in order to estimate the difficulty of a job is meaningless if we don't include any additional rDPS gain due to utility, especially if we're talking about NIN since optimal use of TA and SW is what really makes a great NIN.

    And this still wouldn't solve another important issue: since most hc parties do not include a MNK and MNK has a little more than half the number of parses that NIN has, we can reasonably assume that there's a higher % of "skilled" NIN than "skilled" MNK on fflogs parses. So we should also check the skewness of the distributions of mnk and nin parses and this still wouldn't give the whole picture.
    Sorry, but that's not how statistics work (as has already been explained to address all of these arguments already).


    The percentiles aren't "arbitrarily chosen" they're the median and a maximum that can be reasonably attained without heavily catered parses that make up the majority of the 99% and Max percentiles across all fights. That is literally what a median player will be capable of compared to an expert, it's the Median and Maximum points you would see in a Min-Q1-Q2-Q3-Maximum spread. Ninja's performance, again, isn't magically independent of its own Trick Attack, it benefits from it as well and that impact will be reflected within the data. And there isn't an issue with there being less Monk parses than Ninja parses, as far as statistics are concerned once you've got N=20 in a sample the deviation between the minimum to Q1 to median to Q3 to Maximum isn't going to change in any significant way. It only matters in the case of an incredibly small sample size, not a large one and a second larger one.
    (0)
    Last edited by SpeckledBurd; 09-15-2017 at 08:35 AM.