A DPS Simulator

[Tuldin]

I think it'd be a neat.

State: Ability Set
Transition/Policy: Available Abilities
Value: Damage

Buffs are just abilities, and skills with modifiers could be ignored because their benefits will impact value, and ultimately influence the policy as it converges.

I wouldn't mind taking it on and building off what you've started, if that's alright. ML and AI has been a passion.

[pandabearcat]

I think it'd be a neat.

State: Ability Set
Transition/Policy: Available Abilities
Value: Damage

Buffs are just abilities, and skills with modifiers could be ignored because their benefits will impact value, and ultimately influence the policy as it converges.

I wouldn't mind taking it on and building off what you've started, if that's alright. ML and AI has been a passion.

Sure, that sounds great! I would give you the source code but its going to be all revamped soon.

Tell you what, if you write up your code (or pseudocode), and just tell me what the sim needs to let you know and I will make sure those are available once I finish the engine.

The sim consists of...

• Player (contains stats, buffs, cooldowns, class)
• the Engine (checks to make sure rules aren't broken, such as GCDs and buff cooldowns)
• the Events (type of attack and their costs and damage, and crit values)
• and the Target (keeps track of debuffs, execute, DoTs and their values)

[Tuldin]

Good deal, i'll write something up this weekend and it out to you in the next day or two.

[Eein]

Hello Pandabearcat!

You look like you're making some great progress on your sim!
I'm the creator of the online sim Chocobro; I know people have their criticism and whatnot, but I wasn't aware of another member of the community working on something similar.

You can find my sim at:
http://chocobro.com/sim.php

It currently supports the same things your simulator does plus really in depth graphing of attempts, plus multiple iterations and some really early stat-weight generation.
I began working on a front-end GUI for adding specific gear in; (i used to import from lodestone, but due to slow update, I decided to remove the feature for now)
Once you sim, you can open the log at the bottom for a readout of procs, buffs etc.

I plan on providing custom priority rotations in the future, but my major hangup is due to a few quirks in damage calculation.

The key reasons why I haven't been focused on updating the sim heavily are:
The damage formula provided by valk is not accurate for varying gear levels. I've been working tirelessly on figuring out a working formula that is valid for all gear levels instead of just hacking away at valks formula to get something "close-enough"
DTR does not scale linearly. Heres some testing I did today: https://docs.google.com/spreadsheet/...rive_web#gid=0
(not really sure about DTR. its a mess)
ACC/Crit etc isn't 100% figured out yet. I have some rough estimates but I hear conflicting results regularly.

Since you're talking about revamping your source code, i'd be heavily interested in working on a software based version of a FFXIV sim.
These type of projects are always best left open-source and done through some sort of version control; just in case the developer quits etc, then someone else can pick it up relatively easily.
I started doing some event-driven work in C++ to gear my chops for a full blown software SIM and also messing with the QT framework for some easy GUI work.

In the future, i'd like to allow priority setups using tokenized strings for import similar how Simulationcraft does for world of warcraft.

For my javascript sim, my original intention was to use the front end to generate an array of functions, then parse those in order instead of hardcoding. If you've ever used EnhSim for Enhancement shaman in WoW, they basically had various sets of identified priorities that you could juggle around and it worked really well without an extremely large amount of code.

so you could move a GUI panel to create an order, for loop the list and generate an array of functions, then have a priority function like:
for (x = 0 ; x <= priority.length; x++){ ... } to search the priority based on array size. Not sure about C#'s internals regarding an array of functions, but depending on how you can use pointers, an array of function pointers could also be viable. You can probably use for in as well.

I'm going to hunt you down on excal and talk some code if you dont mind :P

You can usually find me in irc.quakenet.org #chocobro

EDIT: Just to note on your monk stuff for PB:
While in PB, you effectively have "no stance". So stance buffs dont apply like you would have all stances. The only thing PB is good for is getting greased lightning and damage buffs up as quickly as possible.

[Allyrion]

EDIT: Just to note on your monk stuff for PB:
While in PB, you effectively have "no stance". So stance buffs dont apply like you would have all stances. The only thing PB is good for is getting greased lightning and damage buffs up as quickly as possible.

Dragon Kick is a stance specific debuff but it will apply under PB. Hope that's not what you mean.
Bootshine, conversely, will not have its auto-crit. So your last statement holds anyway.

[Eein]

Dragon Kick is a stance specific debuff but it will apply under PB. Hope that's not what you mean.
Bootshine, conversely, will not have its auto-crit. So your last statement holds anyway.

Are you positive? The earlier testing i had done didn't allow this interaction. Some retesting and clarification would be great.

[Allyrion]

Are you positive? The earlier testing i had done didn't allow this interaction. Some retesting and clarification would be great.

Absolutely sure. I've used it myself quite a bit (though I prefer to use it as a dps cooldown).
There are also discussions all over the monk threads here about PB use which include Dragon Kick for this reason. Was your testing during the beta?

I liked Almalexia's analysis of it, though you may find something out with your own testing.

I've been tinkering with effective PB use and I guess this is as good a thread to post as any.

I can squeeze in 5 attacks while PB is active. While I definitely want to use 3 Snaps for GL3, what of the other two? Is it better to DK > Twin > Snap x3, or Snapx5? Maybe Twin and Snapx4?

Snap x5: 900 potency
DK > Twin > Snap x3: 150 + 154 + 648 = 952 potency
Twin > Snap x4: 140 + 792 = 932 potency

But wait! Snap won't put me in Opo form during Perfect Balance and my next DK won't apply the debuff. Will this impact DPS?

DK > Twin > Snap x3: 150 + 154 + 648 = 952 potency
DK > Twin > Demolish > Fracture > ToD > DK > ... = 165 + 140 + DoT + 110 + 44 + 165 ... = 624 potency

1576 net potency

DoTs don't benefit from DK's bonus, so it's not a factor to consider. But direct damage does. Can we squeeze more damage out of the opening rotations by using DK at the end of PB?

Twin > Snapx3 > DK : 140 + 594 + 165 = 899 potency
DK > Twin > Demolish > Fracture > ToD > DK > ... = 165 + 154 + DoT + 120 + 48 + 180 ... = 667 potency

1566 net potency

The difference is not significant, but the first option is more intuitive.

On review, I realize I didn't incorporate the +% damage from each step of GL--not that it matters, since more +% damage is going to pay off more anyway. But just for completeness sake:

DK>Twin>Snapx3 looks like this: 150 > (140+10%) 154 > (180+20%) 216 > (180+27%) 228.6 > (180+34%) 241.2 = 989.8 net potency
Twin>Snapx3>DK looks like this: 140 > (180+10%) 198 > (180+17%) 210.6 > (180+24%) 223.2 > (150+31%) 196.5 = 968.3 net potency

So the initial difference is about 31 rather than about 53. Still insignificant, but now it's Mathemagically™ insignificant!

[Eein]

Absolutely sure. I've used it myself quite a bit (though I prefer to use it as a dps cooldown).
There are also discussions all over the monk threads here about PB use which include Dragon Kick for this reason. Was your testing during the beta?

I liked Almalexia's analysis of it, though you may find something out with your own testing.

Thanks for clarifying Yes my testing was in beta. I'd go into more detail, but I'll refrain from derailing the thread with off-topic.
<3

[pandabearcat]

Hi all, I started working on planning the rewrite but it was confusing me so likely the coding won't happen until Thursday at the earliest.

I also found a bug in which BL had a 12 sec cd rather than a 15 sec one, which drops bard pps by ~1.3%. Oh well, no big difference in the others, though I dropped DRG priority as it was simming worse than rotation and I don't have time to figure out why.

DRG Rotation     avg potency: 144912.10862 PPS: 201.2668175277778
BRD (no DE)      avg potency: 147679.25512 PPS: 205.11007655555554
MNK (no filler)  avg potency: 155706.09278 PPS: 216.25846219444446

[pandabearcat]

I've been...slacking XD

And doing holiday shopping...

Anyway, I did have time today to look at the stuff, and rather than doing what I promised instead I decided to take a first look at the "realtime" dps of the classes.

First comes Monk, I should have the rest in a few hours once I finish lunch.

Here is the "normal" monk opener with PB available:



As you can see it hits average DPS at around 15 seconds in

Here is "uhoh" monk without PB, aka when you have to disconnect from the boss and start all over.



Its awful. You don't hit "average" dps until the 30-45 second range.