Results 1 to 10 of 292

Hybrid View

  1. #1
    Player
    Chrono_Rising's Avatar
    Join Date
    Jun 2017
    Posts
    922
    Character
    Gulvioir Muruc
    World
    Gilgamesh
    Main Class
    Dark Knight Lv 80
    Quote Originally Posted by Lyth View Post
    Quietus:
    Replacing Unleash:
    50 blood = 180n – 150n
    50 blood = 130n
    Lyth, am I missing something is 180n-150n not 30n?
    (1)

  2. #2
    Player
    Lyth's Avatar
    Join Date
    Jul 2015
    Location
    Meracydia
    Posts
    3,883
    Character
    Lythia Norvaine
    World
    Gilgamesh
    Main Class
    Viper Lv 100
    Quote Originally Posted by Chrono_Rising View Post
    Lyth, am I missing something is 180n-150n not 30n?
    You're right! I'll go back and fix that after writing this. I must have been half asleep or something.

    But it's even more hilarious, now that I look at it. If Quietus replaces Unleash, you have a net gain of 30n. If Quietus replaces Stalwart Soul, you have a net loss of 28n. The average net gain per combo is 2n. So until you perform Quietus a sufficient amount of times (on the order of 15-29 times), every other Quietus is going to bring your cumulative total into being a net loss. As an example, if the first GCD you replace is Stalwart Soul, it goes:

    -28n, 2n, -26n, 4n, -24n, 6n, -22n, 8n, -20n, 10n, -18n, 12n, -16n, 14n, -14n, 16n, -12n, 18n, -10n, 20n, -8n, 22n, -6n, 24n, -4n, 26n, -2n, 26n, 0

    Hey! We broke even. Only took 29 times.

    Of course, final values may change and all that. It's still amusing.

    Quote Originally Posted by Izsha View Post
    ...
    Hey Izsha.

    A lot of people like using this sort of GCD averaging approach because it seems like a simple way to do it. I personally try to avoid it for a few reasons.

    First, any time you end up comparing two sequences with a different number of GCDs, you're effectively comparing two "fights" of different lengths. There's a problem with this. In practice, differences in fight length often result in big differences in dps. If we take the fight length as fixed, then you don't really get to pick which combo step you finish on. There's a set amount of GCDs, and you truncate the combo where it ends.

    Another reason is because averaging causes you to lose information. If I know what's happening GCD by GCD, I can always go back and calculate the average. Let's take a look back at the earlier part of this post as an example. If I average, I just see a 2n potency gain over time. But what's actually happening? We oscillate between gain and loss on a GCD by GCD basis. It also allowed me to pick out a more dramatic sequence of events. I could have started by truncating Unleash, and I would have broken even on my cumulative gains after (only) 15 steps instead of 29. Not that you're going to be using Quietus 15 times in a fight.

    The reason why I have blood on both sides of the equation is because I'm using a cute bit of algebra to figure out the potency value of blood. And that's probably the most important reason why I used this approach. Because in mathematics, the most elegant solution is also the most fun solution.

    Quote Originally Posted by Chrono_Rising View Post
    ...
    Oh, I rounded.
    50 blood = 550 - 400 - 20 blood
    50 blood = 150 - 20 blood
    70 blood = 150
    70(50/70) blood = 150(50/70)
    50 blood = 50(150)/70
    50 blood = 107.142857143
    1 blood = 2.14285714286

    So if you plug that back in, you get:
    550 - 400 - 20(2.14285714286) = 107.142857143

    Of course, I'm not overly interested in that 0.142857143 potency, so I just chucked it. Likewise, the average net potency for Bloodspiller should be 202.380952381 instead of 202, but I didn't like the look of that 0.380952381 either, so I fed it to my pet rabbit.

    The reason why I've calculated these separately is because the value of blood actually varies from GCD to GCD, based on which combo action is getting truncated from the sequence. There might be some notational issues floating around in there, I'll go back and run through it again, but it looks like the overall result is correct.
    (1)