A DPS Simulator 2 - SimFF

I hate Irony. At this point I'm just wrestling with the program rather than working out the implementation, which I hate.

There isn't even a good working sample with an AST tree.

I might just get the parse and walk the tree myself.

EDIT: OH MA GOD IT ACTUALLY RAN HALLELUJAH

i need a beer.

Quote:

---

Originally Posted by Dervy

Just to make sure I'm not being dumb (the heat makes me rather dumb and incapable of comprehension), 12% Chance with 1.5196% multiplier at base 354 Critical Hit Rating, right?

---

At 657 rating.

657?! k, yeah, I should of thought that, my bad.

That's... Very low. You checked that base Crit Damage is x1.5, or is it lower? If that's the case that it's 1.5 at base still, Crit scales damn awfully.

EDIT:

Quote:

---

Originally Posted by pandabearcat

EDIT: OH MA GOD IT ACTUALLY RAN HALLELUJAH

i need a beer.

---

Yayyyyyyy!!!!!!!! :D

Quote:

---

Originally Posted by Dervy

That's... Very low. You checked that base Crit Damage is x1.5, or is it lower? If that's the case that it's 1.5 at base still, Crit scales damn awfully.

---

I never bothered to check, honestly. Assumptions are totally a science thing, right? Right? >_>

Also, so far it's a minimum 1.5196x damage, it might be a bit higher. Based on some data points I need to confirm, it could go as high as 1.5216x damage, but the data point I have for that wasn't properly buffed so I had to extrapolate what it would have been buffed. Which is unfortunate, because it was closer to minimum damage than anything else I have so far. :(

By "it actually ran" I mean

1. It compiled
2. It stopped throwing null pointer errors

The good news is that while the work is extremely tedious (yay for making so many nonterminals) the work actually, well, works, so I am actually starting to be able to attach some code to the parsing. For example, so far it can load class skills/traits, add players/targets, and change attributes.

Next up is simply...writing the next 30 or so code stubs. Then it'll work.

I anticipate it'll be done by monday/tuesday.

Far far down the road someday, I may have it spit out actual C# code, so you can compile your rotations to make them run faster. In the meantime performance is the least of my concerns and I doubt anyone would actually care about this functionality so its not going to happen.

Also, please take a look at the script sample, and try to write up your own. I need people to get hands on so they can tell me if they need additional functionality from the language. Earlier is better.

Quote:

---

Originally Posted by pandabearcat

snap

---

So it's in "Please look forward to it." mode?

http://ku-so.hatenadiary.jp/entry/2015/06/20/105345

JP CRT% at level 60 is 5%. Very small sample-size, but very likely.

I need new damage formulas!

Also I put in the new DRG skills. Not sure if they work correctly...but whatever.

The engine has also been tweaked a little to help with multi-actor fights. Pretty much was necessary for pets to work as they are independent actors.

Still can't do multiple instances of dots however. I will need to completely rework how auras work to support that, whomp whomp.

Script now supports preapplication of auras (like disembowel).

IT WORKS!

It worrrrrrkssssss.

I expect it to be somewhat buggy at first, but it is working.

The only thing left to do is code up the UI, which I really don't feel like doing.

But it...works wonderfully (though a bit slow, who cares).

Hope to release a demo tomorrow once I get a somewhat UI working.

DoTs seem to scale *(1+SS*0.0002), from base 0, which is 1/3 of how DET scales.

I noticed that my WD formula is kind of the same, just a few adjustments here and there are needed, but it always calculates my DoT damage as if we have 0 Skillspeed. I also applied a -1 mod to the end of my damage formula and it seems to be slightly more accurate than normal, lol.

However, if Determination was butchered for WDs as well, which is what I also think is happening, due to this post https://www.reddit.com/r/ffxiv/comme...condary_stats/, then the scale of SS on DoTs is much greater than what we originally thought.

Quote:

---

Originally Posted by Dervy

DoTs seem to scale *(1+SS*0.0002), from base 0, which is 1/3 of how DET scales.

---

Just take your japanese blog: http://ku-so.hatenadiary.jp/entry/2015/06/30/120251

It looks pretty much to me like if SS reduces your GCD by X%, then it will simple increase your dot damage by these X%. (the simplest and most logical model actually - as it's the most linear thing possible)

Huh. That's a new post today then, why I never saw it.

It's not exactly a 1:1 ratio. It looks like the higher your WD/STR/DET, the value of SSX% on DoT Damage will gradually take over the %Haste increase via Skillspeed.

http://puu.sh/iIedd/53b1024c3e.png

It's fairly close however. I really should take a measurement with equal skillspeed... Once I've done some more DET testing, I have a feeling that 0.0002 will increase to around 0.00025

EDIT:

Something else for you guys working on Crit http://www.bluegartr.com/threads/125...3.0-Bardshrekt

Check the first page for a new release Its finally out!

I just did a stress test for multiple trials for SimFF.

It is a bit slow, but I think it is as fast as can be expected.

A 10 minute trial takes on the order of 1-3ms to run.

10000 trials takes roughly 20 seconds. Memory usage does not seem to be a problem.

Welp I'm done for now. I'll probably add DRG cross class abilities sometime this afternoon or tomorrow. Right now sim is putting the old lvl 50 DRG at 530 dps.

Well, time to throw in a bit of crit info. All data is taken with 657 Crit, 689 AP, 260 Det, 12% Crit rate.

No venom numbers:

100 Potency--246-271, 373-412
150 Potency--370-409, 562-621
Auto--221-244, 336-371

Venom numbers:

100 Potency--295-325, 447-494
150 Potency--444-490, 674-745
Auto--265-292, 403-445

The 100 potency venom numbers trouble me. 325.5/294.5 is exactly 21/19, so there's no 21/19 spread that should give only 295-325, because a 326 should be possible if it's 325.5, or a 294 if it's lower than 294.5.

Honestly, though, the rounding errors here drive me crazy, which is why for a while I was working on a 500 potency fully buffed range, but that's ridiculously hard to get a range for when you're looking at ~300 values in the crit range. I was able to get a full non-crit range of 1793-1982, with an incomplete crit range of 2729-3012. Even if we assume that 3012 is the cap (which there's some room for it not to be), that's a x1.519677 multiplier.

TBH i don't need exact numbers. Even 5-10% variation is okay. It may affect BiS lists but tbh there isn't that much gear customization in this game anyway, and I am far more interested in getting people the correct rotations and general ballpark dps to shoot for than "omg det is 1% better for me than crit".

Also go download and break the simulator please. I am sure there are bugs and I need to find them to fix them

Your best bet for this kind of data retrieval will likely be BLM and DRG.
Fire 4 is a total of 504 potency, plus some other % damage buffs, but the effective potency is consistent and very high. It should provide values in the 2000-2500 range, which reduces rounding error to <0.5%.
For autoattacks your best bet is probably DRG, ignoring Heavy Thrust.

Quote:

---

Originally Posted by Viridiana

100 Potency--246-271, 373-412
150 Potency--370-409, 562-621


100 Potency--295-325, 447-494
150 Potency--444-490, 674-745

---

246*1.5075 = 370.845
373*1.5075 = 562.2975

295*1.5075 = 444.7125
447*1.5075 = 673.8525

Try get 200 potency and tell me if if a multiplier of 2.01 works.

Quote:

---

Originally Posted by Dervy

That's... Very low. You checked that base Crit Damage is x1.5, or is it lower? If that's the case that it's 1.5 at base still, Crit scales damn awfully.

---

Quote:

---

Originally Posted by Viridiana

I never bothered to check, honestly. Assumptions are totally a science thing, right? Right? >_>

---

These words, they taste so good!

70 WD, 370 AP, 267 Det, 354 Crit. . .

100 Potency--180-198, 260-288
Auto Attack--177-195, 258-284

Quote:

---

Originally Posted by Dervy

246*1.5075 = 370.845
373*1.5075 = 562.2975

295*1.5075 = 444.7125
447*1.5075 = 673.8525

Try get 200 potency and tell me if if a multiplier of 2.01 works.

---

Works reasonably well on my non-crit numbers. Don't have a full crit spread to try it against, though. 494-546 venomless, 592-655 with venoms.

My theory is that 100 Potency isn't an *1 Modifier. It's closer to 1.005, or 1.005/100 = 0.01005 per point of Potency.

10 = 0.1005
20 = 0.201
30 = 0.3015
40 = 0.402
50 = 0.5025
60 = 0.603
70 = 0.7035
80 = 0.804
90 = 0.9045
100 = 1.005
150 = 1.5075
200 = 2.01

Quote:

---

Originally Posted by Dervy

My theory is that 100 Potency isn't an *1 Modifier. It's closer to 1.005, or 1.005/100 = 0.01005 per point of Potency.

10 = 0.1005
20 = 0.201
30 = 0.3015
40 = 0.402
50 = 0.5025
60 = 0.603
70 = 0.7035
80 = 0.804
90 = 0.9045
100 = 1.005
150 = 1.5075
200 = 2.01

---

179.5 (minimum DB damage to match observations) divided by 1.005, times .3015 to get to 30 potency Mutilate dot makes 53.85, but I observed 53's. And they had to be low 53s, too, to be the full spread against the 58 maximum. And the 58 maximum only brings me to 195 instead of 198. This is with a small amount of SS boosting the Mutilate numbers, too.

Oh., i forgot to mention that my damage model minuses 1 at the end of the total calculation... So:

198 / 1.005 = 197.0149253731343
197.0149253731343 * 0.3015 = 59.4 - 1 = 58.4

179.5 / 1.005 = 178.6069651741294
178.6069651741294 * 0.3015 = 53.85 - 1 = 52.85

58.4/52.85 = 1.105014191106906

Then if we were to reverse it:

58.4 + 1 / 0.3015 = 197.0149253731343 * 1.005 = 198
52.85 + 1 / 0.3015 = 178.6069651741294 * 1.005 = 179.5

It's fairly accurate I guess. It doesn't fit all of my data points however... But it kind of makes a lot of it more simple to explain.

So, I'm stepping through crit values now, and it looks like it takes about 521 rating before you get back to a x1.5 multiplier for damage.

Quote:

---

Originally Posted by Dervy

Oh., i forgot to mention that my damage model minuses 1 at the end of the total calculation

---

But if you're going to minus something at the end anyway, wouldn't it be a simpler assumption that the potency numbers are a straight multiplier and that there's a non-1 subtraction at the end? >_>

I tried solving potency with my Dragoon data points (the old 2.X ones) and the model for my sample, but it gives me a larger variation of errors when I solve it this way without minusing 1 at the end of the potency calculation.

I hate this game sometimes. lmao

Doesn't ZAM have a way to extract data from gamefiles? simcraft does something similar and if I can get my hands on that datamining it would make this so much easier.

If I can grab spell data straight from the game I wouldn't have to write and rewrite all this code for abilities.

Yeah they should. Premium Virtue (Owner of XIVDB), owner of Garlandtools and two others from the Bluegartr community, one of which shares a similar name to a Community Rep, all can extract data from the game.

So could we simply ask them to dig into the game code and get out the exact damage/healing/potency... and every other equation from the game?

Quote:

---

Originally Posted by Dervy

http://puu.sh/iIedd/53b1024c3e.png

---

What's the 2nd table supposed to show? How is a base of 412 SS with a 2.5 GCD even possible?

No, no, you're right. I forgot to change it back. I was testing something else out before I took the printscreen, my bad. http://puu.sh/iIWrj/a4c263c46c.png Highlighted it in red to show that this was the baseline I'm measuring it from (couldn't reach 0 SS with that WD/DET), but it's +58 from the base of 384.

Also, it's supposed to highlight the difference in GCD from the baseline I measured, versus the impact DoTs made on damage.

So, I decided to step through some data points for Mutilate working towards SS scaling (even though I didn't get to SS scaling today, but oh well). I'll edit this post later with SS values, but for now I figured I'd give the points I do have. Everything is at 70 WD and 428 Skill Speed, because the only weapon I had on me with no SS is my Zenith, and that would be pointless. >_>

Varied Crit, 267 Determination:

Code:

---

Stats            Non      Crit
370 AP, 354 Crit: 53-58,  76-84
398 AP, 377 Crit: 57-64,  82-91
434 AP, 409 Crit: 62-69,  90-100
452 AP, 422 Crit: 65-72,  95-105
496 AP, 457 Crit: 72-79,  106-116 (not full spread?)
548 AP, 505 Crit: 79-88,  117-130
574 AP, 521 Crit: 82-91,  123-135
622 AP, 551 Crit: 91-100,  136-149 (not full spread?)
685 AP, 586 Crit: 99-110,  148-165
708 AP, 601 Crit: 104-115, 156-173 (Auto Attack 342-378, 516-568; Death Blossom 343-380, 518-572)

---

Varied Determination, ignored Crit, tried to match AP from before, still 428 SS 70 WD. Also ignored critical damage:

Code:

---

Stats            Mutil    AA      DB
398 AP, 339 Det: 57-64,  193-212, 194-214
434 AP, 343 Det: 63-70,  210-232, 212-235
452 AP, 360 Det: 66-73,  220-243, 222-244 (not full spread? pattern says 245 max)
496 AP, 385 Det: 74-81,  242-267, 244-270
548 AP, 401 Det: 80-89,  268-296, 272-300 (weird that there's a 4-point range here compared to 2-points elsewhere)
574 AP, 426 Det: 85-94,  283-313, 285-315
622 AP, 442 Det: 92-101,  307-339, 309-342
685 AP, 480 Det: 104-115, 340-375, 343-380
708 AP, 473 Det: 108-119, 351-388, 354-391
797 AP, 516 Det: 121-134, 398-439, 400-442

---

Varied Skill Speed with constant 267 Det. Very bottom set is 469 Det.

Code:

---

Stats            Mutil    Auto    DB
398 AP, 500 Sks: 58-65,  190-211, 193-213
434 AP, 522 Sks: 63-70,  207-230, 211-234
452 AP, 512 Sks: 66-73,  217-240, 219-242
496 AP, 546 Sks: 74-81,  238-264, 241-266
548 AP, 569 Sks: 81-90,  264-291, 266-294
574 AP, 580 Sks: 85-94,  276-306, 278-308
622 AP, 616 Sks: 94-103,  298-331, 301-333
685 AP, 574 Sks: 102-113, 331-366, 332-367
775 AP, 605 Sks: 115-128, 374-414, 378-417
              Full Gearset
862 AP, 623 Sks: 133-148, 428-472, 432-476 (469 Det)

---

Initial testing from ACT dev shows that base crit multiplier can be as low as 1.46, and low spellspeed values reducing dot tick damage to below 2.X levels. I don't know how accurate that is, but wanted to throw it out here.

Quote:

---

Originally Posted by Dervy

but it's +58 from the base of 384.

---

Baseline is 354 - but that's what you used anyway.

Your GCD reduction just seems off:
2.443/2.5 = 0.9772 => 0.0228
2.435/2.5 = 0.974 => 0.026
2.423/2.5 = 0.9692 => 0.0308

Quote:

---

Originally Posted by Waliel

and low spellspeed values reducing dot tick damage to below 2.X levels.

---

Well, from what I've done so far and what I know of the game mechanics, that's impossible. There's no way that Skillspeed goes into a negative multiplier somewhere that reduces potency for DoT ticks. If you Calculate damage for X Potency, it will be X damage, or basedmg*potency, no less. It would make no sense for Skillspeed to be calculated this way.

The only two possible outcomes is that X% GCD Reduction increases DoT damage by X%, or Skillspeed is scaling damage from either 0, or 354 SS, as their base value.

Where is his work? Cause I need to take a look at this. All of my work so far has indicated that DoT potency is higher than what my damage model calculates at 354 Skillspeed, meaning, there's something affecting it already from below this numerical value.

EDIT: Skillspeed definitely scales damage from base potency of 0 with my new model. That is for certain. And Determination got absolutely butchered. That's hilarious.

DET vs WSs:
2.5: 0.00062
3.0: 0.0001495

DET vs AA's:
2.5: 0.0014
3.0: God knows.

AA's are completely all over the place. Idk why my WD*STR*DET model isn't working for AA's. I removed the -1 at the end of the formula, added it back. Double checked my AA's and the values I used....

I also checked with Ravahn. The guy who he was originally working with to do SS's messed up all of his data and forgot to add a lot of control variables to isolate SS.

Okay, here it goes. SquareEnix looks like they may of changed more than just the stats themselves. It appears there's no longer a base damage value which my old model used to plus on top of the Strength modifier, after it's been multiplied by 0.075.

This change to damage changes a lot of things. For example, the Strength Coefficient is now 0.011, and the Determination Coefficient is 0.0001368, which makes Determination scale almost 4.5 times less than 2.5X for Weapon Skills.

So, here's the Updated Models, which, are a WIP:

Weapon Skills:
((Potency*0.01005)*(WD*0.0400573+1)*(STR*0.1104171)*(DET*0.0001368+1)*BUFFS)-1

Auto-Attacks:
((WD*0.0319937+1)*(STR*0.1387779)*(0.0001513)*BUFFS)-1

Skillspeed Mod for DoTs:
(1+(skillspeed-354)*0.0001285) - Note, this fits into the Buffs section.

GCD Calculation:
I read through a few sources about GCD on Bluegatr, Reddit and the JP one and now for GCDs I'm using:
2.50256-(0.01*(SS-354)/26.5)

The reason for 2.50256 is because the first 6 values of Skillspeed are valued differently. The Japanese use the 2.50256 number to by-pass the modifier for the first 6 values.

Of course, my Damage Models only work for the Dragoon and are still a WIP, so a few numbers will change.

Also, I spoke to Clorifex (Garlandtools guys) and he mentioned that the game uses whole numbers in every formula he's encountered so far. Oddly:
1 / 0.0400573 = 24.96426889
1 / 0.1104181 = 9.056486629
1 / 0.0001368 = 7308.289367

There's a good probability that the game actually Divides WD/STR by whole numbers, rather than multiplying by long decimals.

So, when I rounded these values, I got this:

((POTENCY*0.01005)*(WD/25+1)*(STR/9.05)*(DET/7308+1)*buffs)-1

This only gives me a 0.2 variation error over my model which multiplies values by decimals.

Total cumulative variation error is 3.8871, which is good.

EDIT: Updated AAs

Quote:

---

Originally Posted by Dervy

((POTENCY/0.01005)*(WD/25+1)*(STR/9.05)*(DET/7308+1)*buffs)-1.

---

Shouldn't that be Potency/100? Or at least *.01005, if you insist on keeping the .01005 idea you have going? :p

On that note, how accurate is the model if you do Potency/100, STR/9, and change the -1 at the end to something else?

Of course, I meant (Potency*0.01005) :P

I changed it to Potency/100 and STR/9 and the variation dropped another 0.2, still kept the -1 however.

I then made it so it solves for -X. This value just changed to -1.5, but variation dropped by further 0.5.

So, it would of been ((Potency/100)*(WD/25+1)*(STR/9)*(DET/7308+1)*BUFFS)-1.5, total error of 3.1816

Interestingly, this model is even accurate when I tested for 72WD, 841STR, 505DET. Which, is good. At least I know that Determination is hard-valued now at DET/7308, or DET*0.0001369 and scales from 0.

Having said this... I'm curious as to how this model reflects my "Job Coefficients".

EDIT: Errors are all of the errors from my data points added up together. So the more DataPoints I have, my Variation Error will increase.

Quote:

---

Originally Posted by Dervy

So, it would of been ((Potency/100)*(WD/25+1)*(STR/9)*(DET/7308+1)*BUFFS)-1.5, total error of 3.1816

---

Yay progress! Also, dunno if you saw, but I updated my post from yesterday to add in the SS numbers for my NIN.

I'll go have a loot at it now ^^

I'm actually kinda glad I can use a Damage Model in this format now. It makes it so much easier to remember, and it looks so much more cleaner.

This is all a WIP remember, so a few of the values will eventually change over time, as I add more and more Data Points to my solver.

EDIT: Fixed it. I messed up the AA_Delay calculation, lmao. Everything's perfect!

Interestingly, DET's weighting is 0.120, SS is 0.097 due to the DoT clipping, but, CRT with the same old 1.5% DMG mod gives me a weighting of 0.363. I'm iffy about this value however until we've got a working Crit Dmg scaler and someone has figured out a new Crit Formula for us to work with.

Interestingly, DET's weighting is 0.120, SS is 0.097 due to the DoT clipping, but, CRT with the same old 1.5% DMG mod gives me a weighting of 0.363. I'm iffy about this value however until we've got a working Crit Dmg scaler and someone has figured out a new Crit Formula for us to work with.[/QUOTE]

Lol yup it seems we are back to ilvl 80 ilvl 90 gear back then just by looking at WIP for stat weights. It's funny. Keep up the good work guys.

Off topic, hey dervy do u plan on updating the how to dps drg page with latest information about drg abilities and openers soon?

Quote:

---

Originally Posted by Dervy

Where is his work? Cause I need to take a look at this. All of my work so far has indicated that DoT potency is higher than what my damage model calculates at 354 Skillspeed, meaning, there's something affecting it already from below this numerical value.

---

Well, that's the thing. He didn't show any work and just said that. No idea if there's any truth to that and, as I said, just wanted to throw it out here to see your thoughts.

If yall added playerlevel as a modifier to the formula I'm hoping you could find parallels between 50 and 60, and we won't have to go through this process every time people level. Cause I'm pretty sure thats how it scales somehow.