Dragoons: A Rotation Reborn

[pandabearcat]

Do Ayvar or Easymode still play?

[Jonnycbad]

I assumed BRD was ahead, or maybe used to be ahead...meh

MP based ability would be interesting given we have pretty high Piety left over from 1.0 which they never adjusted. I also thought from the live letter it seemed to suggest we're only getting new actions from job quests, which would be 2; 55 and 60, and traits woven in between. Given in most 10lvl ranges there's only 3 traits or so, I feel like we're likely to get 3 as well; likely buffs to things that aren't already buffed from traits, perhaps Chaos Thrust duration and Phlebotomize duration, and maybe cooldown buffs to Dragonfire or Elusive Jump, or a trait called High Jump or Double that increases Power Surge to +100% potency instead of 50% given most ff games with drg have a high or double jump trait.

[Jonnycbad]

A buff for phlebotomize from 18->30sec would be soooo welcome.

[Jonnycbad]

Dervy, is the rotation you have in your 2.45 video still relevant? In that video you alternate LS on jump and FT but is that better than doing the 2x FT during Blood with the first one buffed with LS? Your rotation in that video has b4b used on disembowel instead of right before chaos and not sure if it's still relevant or if the ht-ir-leg-dis-b4b-ct-ps-ph-jump-tt-vt-ls-ft-leg-ht-tt-spine-vt-dragonfire-ft is better or more updated than your 2.45 video. Please advise

[Dervy]

Everything is very wrong at the moment. The formula? Wrong. My weights on the blog? Wrong. 90% of BiS sets? Suboptimal.

The only thing I can now say, is DET is around 0.38-0.4, SS is 0.14-0.16 and CRT is most probably around 0.23. The next thing is that the more strength you have the more Determination scales.damage, and I have sufficient Data to back this up now.

Right now, I've switched from weighted sets to very... "loose" personal criteria :

- Do I have high Strength?
- Is my Determination high?
- If my strength is lower than i130, is my Determination retardedly high?
- Do I have 380-390 skillspeed for shenanigans?
- Have I met accuracy?

Hopefully all of this will be sorted out before Heavensward... Need more data!

EDIT: Sunny, wtf is going on here?...

Ability Damage DET vs STR:
http://puu.sh/hqVqZ/6104147b4c.png

Compared to Auto-Attack Damage DET vs STR...:
http://puu.sh/hqVtM/237cf912b5.png

I might need to redo all of my Auto-Attack data gathering... It's weird though, because all of my spreadsheets were direct copy&pastes from my ACT logs and all of the numbers I put into my graph are exactly the same. More data you reckon? Or is there some sorcery at play....

Also, at 660 STR, 345&334 DET, my parse numbers are so screwed up. Despite the 11 DET difference, the parsed damage logs are exactly the same (at 100 potency).

When I then parse at 150 Potency for those two skills, then divide by 1.5 to adjust back to 100 potency, the values are much higher, to the point that 345 @ 150/1.5 potency is higher than 355, and 334 @ 150/1.5 potency is higher than 345...

What do? Halp.

EDIT2: For now, I'm just using different determination values... kinda cheating... ah well. All of this Data I'll throw into my solver late on as well for some more accuracy.

[SunnyHirose]

MFW

So, two things.

One reason I got my action data from multiple skills (and also used data from crits, which have more precision) is because while I'm not an expert statistician, if my understanding is correct, it should reduce overall systematic rounding errors. Autoattacks are inherently harder to calibrate this way, because the only options you have are using a different weapon of the same iLvl (annoying / impossible) and buffs/debuffs (hard to track with existing tools, and the Right Answer is to write a program that tracks it, of course, but that's... like, a set of technologies I'd need to learn before I could program it). It stands to reason that it should even out some weird data points. I mean, it works for electron microscopes, right?

The other thing is that you may have run into a possibility I was afraid of: that there may be significant rounding voodoo throwing things off, or there just plain isn't a pure formula underlying all ranges of all stats (rather, partially or entirely LUT based calculations).

In the event that it's like a pure formula, though, we could entertain the idea that one point of a stat is not quite equivalent to the next while being linear in general--that is to say, there is error in x as well as in y. In that case, we ought to use an errors-in-variables regression (this should also reduce a bias in linear slopes that I suspect exists because formulas we've seen before tend to underestimate future gearsets).

So, there is yet hope, but anyway I am dreading having a look at auto-attacks because the level of error is a nuisance.

[Viridiana]

If y'all need more data sets (though they'd be from other jobs, probably), I'm not doing much until 3.0.

[Dervy]

Yup I also jumped on the Wildstar train as well. Was completely garbage. The story mode quests are awesome, especially the final batch of quests we just got in 2.55. Oooh baby. But if you wanted to get your ilvl up quickly, World of Darkness 24-man raid is your best bet.

Uhm... Yeah http://sim.ffxivguild.net/ but it's garbage. Your old sim for the bard is much better imo. T0rin and other players also quit. I wish they stayed Poor Summoners haven't had a thread for over a year.

If you want to talk about a remaking a Bard simulator, contact reddit.com/u/MrYaah (he might post here at some point). He has a very good Bard Model which we worked on together for quite some time and I think he's in the process of making a simulator in Python.

EDIT:

If y'all need more data sets (though they'd be from other jobs, probably), I'm not doing much until 3.0.

Bard stuff could be quite useful, just to check whether the JP assumption that Dragoon and Bards share the same Job Coefficient is accurate or not.

EDIT 2: These is my result so far using my "Occam's Razor" approach. No rounding. Still need more Data Samples to account for even greater accuracy, but I think I'm drawing closer and closer to something we can start working with. It's the only expression thus far which has given me the most desirable results.

http://puu.sh/hs7GZ/58be85360f.png

http://puu.sh/hs7J2/1aa1ff155a.png

AC: ((WD*0.0403713+0.9236585)*(AP*0.0445705+2.645907)*(DET*0.0011454+1.8473665))

AA: ((WD*0.0403713+0.9236585)*(AP*0.0445705+2.645907)*(DET*0.0022385+1.5882988))

http://www.filedropper.com/detvsbasestrtotals

And yes, the WD*AP*DET expression even works with 0 WD parses. If people could go through it and check for mistakes and/or have any suggestions to improve on, I'll greatly appreciate it. Remember, only works for the Dragoon and maybe the Bard.

What I, or someone, needs to work on is how this bloody game rounds numbers...

[pandabearcat]

The problem with the old dps simulator is that the code was really not very polymorphic and couldn't really handle what I had planned for putting in stat weights (it just did pure power conversions, w/o stats and damage). It also didn't have a very robust priority parsing. Also you don't want to see what the event handling was like, there was a lot of spaghetti code. Was originally made for just bards. Monks and dragoons weren't too hard to add, but trying to make it work for casting was a chore, esp once things started taking more/less time than GCD.

If I do go back to it I would likely start over, maybe with a different language, either one that could be hosted online or something faster like C++. I don't think I could ever work with python, indents for braces? No way >.< Python's lamda functions would be useful though.

I should also add that iirc the simulator's primary purpose was to define "optimal" rotations, something which I figured people would have done by now since its been 2 years, but I guess people are still arguing about things? Back then when the game first came out there were a lot of people with terrible playstyles. So the simulator was made to settle those arguments. I couldn't really care less about stat weights, I get little enough gear as is.

[Viridiana]

EDIT:
Bard stuff could be quite useful, just to check whether the JP assumption that Dragoon and Bards share the same Job Coefficient is accurate or not.

I'm bad at checking for edits! BRD data coming up soon. . .

Gear: Wave Bow (42 WD) and Race Specific Equipment
Stats: 42 WD, 202 Determination, 307 AP, 2.8 weapon Delay

Heavy Shot (150 Pot) 153-170 normal, 229-254 critical
Straight Shot (140 Pot) 142-157 normal, 212-235 critical
Auto Attack (2.80 delay) 77-86 normal, 116-129 critical

Only using Straight Shot, Heavy Shot (Bloodletter too just for more data points), and Barrage. Looks like after the +Damage% buffs the 150 potency skill is almost twice the 2.80 auto attack.

Second round:

Stats: 42 WD, 202 Determination, 413 AP, 2.8 weapon delay

Heavy Shot 197-218 normal, 297-328 critical
Straight Shot 183-202 normal, 274-303 critical
Auto Attack 100-111 normal, 151-166 critical (150 auto attack refused to happen >_<)

Third round:

Stats: 42 WD, 246 Determination, 413 AP, 2.8 delay

Heavy Shot 202-223 normal, 305-336 critical (pretty sure 304 is minimum, but wasn't happening)
Straight Shot 188-207 normal, 281-310 critical
Auto Attack 105-116 normal, 157-174 critical

Fourth round:

Stats: 46 WD, 233 Determination, 315 AP, 3.28 delay

Heavy Shot 168-185 normal, 253-279 critical
Straight Shot 157-174 normal, 236-260 critical
Auto Attack 102-113 normal, 153-170 critical

Final set for now:

Stats: 46 WD, 303 Determination, 562 AP, 3.28 delay

Heavy Shot 288-319 normal, 434-478 critical
Straight Shot 269-297 normal, 409-442 critical (should be about 405 minimum)
Auto Attack 183-202 normal, 274-303 critical.