[BP]Mortifer

I didn't mean saving the rng results, but rather using the same seed to more or less "reset" the rng to the same state in between tests, but as you said, since haste will use a different number of whitehits, the sequence won't be the same anyway.

But read about rng seeding for whatever programming language you are using, I believe that it could help you out with the issue of different simulations not giving the same results.

 

and

You are right. I used a previous version character file, that's why I got the errors.
Thanks alot for your help!

And

 

dr_AllCOM3

I just tested that. Rnd is indeed producing the same sequence every time (Kahorie is using something different and more random). It wouldn't be much work to implement.

The question is now, would using the same random numbers for each EP cycle improve the result?
If yes, when should a new sequence be generated?

DocsDK UI
Simulator

 

Foxx2405

If HB is on cooldown you shouldnt have to refresh diseases yet though. (15 sec duration vs 8 sec HB CD).

My guess is that the sim does recognize that HB refreshed FF, but that it does not use HB to reapply FF. I guess it can be modeled in though.

 

Afabar

No, if glyphed HB apply disease. There are two explenations. HB is resisted then it cast IT or there are no FU runes available when frost fever is about to fade.

On the question of the RNG, Doc if you want to implement it. I would reinitialize it every time you click on the EP button. Anyway the results could not be less accurate than they are today.

 

Dokus

Does this tool add AP from certain Talent points?
E.g. When i try out a specc that has Bladed armor. Does this tool adds and calculates the extra AP from that talent? or do i have to add manually the extra AP in the character.xml file?
same counts for Dark Conviction.

My guess is i have to add it manually, but not sure, thats why i ask.

 

Afabar

No the sim read your character file, modify it on the flight according to your talents and selected buffs.

 

leladax

Uhm, regarding that, does it calculate it assuming target is facing the player? 'Cause I also get impressively high EP on expertise.

 

Afabar

I don't know if it's the calculation of the EP that give a too high value. The sim compare expertise cap versus exp cap - 50, not just the gain of 50 exp. So the value seems very high where a GCD means a lot.
However, I'm surpised too by the high value because other spreadsheets never gave such number. Maybe those were wrong or it is my comparison that is not appropriate.

 

leladax

So, I'm going to assume Expertise is better than Strength 'till soft cap.

 

slant

I have several suggestions and enhancement requests for this very useful program.

- Sort abilities in the report descending by % of total damage.
- In the report, provide at least one decimal point precision. (example: obliterate is 20.4% of my total, not 20%)
- In the report, provide peak and average values for DPS and TPS.
- Report on runic power generated per 20s "cycle" and per second/minute.
- Support for individual fight length and number of fights. The simulator currently covers sustained damage over hours or days, but boss fights are actually individual 3-12 minute battles. Shorter fights exhibit different performance profiles due to the use of consumables, higher proportional bloodlust uptime, DPS cooldowns, short rampup due to starting at zero runic power, and so on.
- I'm sure this is already on the list, but accounting for trinkets with use and proc effects would be very useful, particularly when combined with DPS cooldowns like ebon gargoyle.
- Use multiple threads to take advantage of multi-core CPUs.
- The default priority for unholy should prioritize blood fever over frost fever due to the rage of rivendare talent.
- Support consumables in the buffs tab. (fish feast, flask of endless rage, haste potions)
- Support for various combat-impacting racial abilities. (racial weapon expertise, arcane torrent, berserking, blood fury, command, etc)

 

dr_AllCOM3

Spreadsheets, missed strikes and runes don't mix very well. I'm not surprised, that Exp+Hit are rated high in a tight BP rotation.
Edit: Unholy has a much lower value, I think because of the more relaxed rotation.


There's not much on the list right now, besides maybe improving EP.
You can just add fixed stats like consumables to the char file.
Multithreading would require a rewrite of large parts of the code. You can fake it by starting the sim multiple times.
Repeated short fights are interesting. I halfway implemented fight interrupts.

DocsDK UI
Simulator

 

Rurahk

These two options would be synergistic. Run a 5 min fight 100 times on multiple threads and average the results. This shouldn't require too much of a re-write, especially if shortcuts were used like using temporary files to collect results as opposed to having a master thread.

Other interesting statistics for multiple sims of short fights would be min, max, and variance. Very interesting would be a bell-curve-like representation of the dps plotted on a graph.

Something that would be a great addition would be some sort of integration with Rawr. Either implement the sim as a module or (probably more realistically) implement an import/export mechanism from Rawr's xml format to the sim's xml format. This would, in my opinion, greatly increase the usability of this program.

 

crusaderky

I, too, am trying a build based on glyph of howling blast, with <FrostFever /> at top priority. <IcyTouch /> is not even in the list, and yet I get quite a lot of IT even if my spell hit is capped.

I think that the problem stays in what the definition of a disease being "about to fade" is:

if you haven't got glyph of HB nor pestilence, it should be "if I don't shoot IT now, it will fade before the next F rune becomes available again"
if you've got glyph of howling blast, it it should be "if I don't shoot HB now, it will fade before the next FU runes becomes available again"
if you've got glyph of pestilence, it should be "if I don't shoot Pestilence now, it will fade before the next B rune becomes available again"

also, with glyph of pestilence you should be extremely conservative with the definition of "before" (if you fall 0.5s late, it's a disaster), while with goHB or unglyphed you really don't care if you sometimes stay 1s or even 2s without disease.

For example, in my goHB rotation, at a certain moment I've got 2 OB/HB to shoot, another OB/HB will be ready in 6 seconds, and FF is expiring in 5 seconds.
In this situation I shoot 2 OB, which I know that will proc at least 1 rime with a 28% probability, and if they don't, I'm staying without disease for 1 second (it's worth the gamble). If, instead, I see that if I shoot OB OB I risk staying 4 or more seconds without disease if rime doesn't proc, I do OB HB instead. I don't think there's a way to model that with the current priority syntax... correct me if I'm wrong.

 

mahal

I'm having a very difficult time getting the simulator to use pestilence the way I would prefer. It applies pestilence at the correct time, but then uses PS, IT immediately afterward. I would like the sim to only perform the first 4 lines of the rotation once, then after pestilence skip right to FSx2. Instead, it's simply going through the entire rotation every cycle. How do you get it to apply PS,IT only once?

<Rotation>
<PlagueStrike retry='1'></PlagueStrike>
<IcyTouch retry='1'></IcyTouch>
<BloodStrike retry='1'></BloodStrike>
<BloodStrike retry='1'></BloodStrike>
<FrostStrike retry='0'></FrostStrike>
<FrostStrike retry='0'></FrostStrike>
<Obliterate retry='1'></Obliterate>
<Obliterate retry='1'></Obliterate>
<Obliterate retry='1'></Obliterate>
<FrostStrike retry='0'></FrostStrike>
<FrostStrike retry='0'></FrostStrike>
<Obliterate retry='1'></Obliterate>
<Obliterate retry='1'></Obliterate>
<BloodStrike retry='1'></BloodStrike>
<Pestilence retry='0'></Pestilence>
</Rotation>

<Priority>
<BloodPlague></BloodPlague>
<KMRime></KMRime>
<KMFrostStrike></KMFrostStrike>
<FrostFever></FrostFever>
<Obliterate></Obliterate>
<BloodStrike></BloodStrike>
<FrostStrike></FrostStrike>
</Priority>

 

Fugazor

As EP calculations are not really working properly I decided to do what it does just manually. I noticed two interesting things:
1) Sometimes simulator suddenly loose some dps, for example you have X amount of results in 8340-8360 range and then you get 8300. It either fix itself with next calculation or, more often, application needs to be closed and reopen again to get good results again.
2) Similar bug often happen when you change stat values in character sheet. To be safe I learned to restart program every change like that.
Those two bugs most likely cause normal EP calculations to go wild often.

P.S. Simulation was set to 100h and there was hit and exp caps so I do not think it is RNG cause.

EDIT: I've got quite high haste value for 51/0/20 Blood (over 2 per 1 AP). I double checked results and compared 0% vs 10% haste (skill - number of "landed" increase): DS - 1%, PS - 1.7%, IT - 2%, HS - 1.1%, DC - 2%, Necrosis - 10.5%, BCB - 11.1%, melee - 10.5%, ghoul - 8.3%, DRW - 6.9%. I just don't really get gains of DS, PS, HS and if I increase haste to some crazy numbers (100%) for example DS keeps gaining as much as over 9%. Am I missing something here? I can hardly believe that lowering GCD of DC and IT by a little can boost other abilities that much.

 

Sadiera

I really want to know that too. Also the fact that most of the standard builds have 500-600 AP specified? Is the calculator created so the AP is what's added from gear and not your total AP?

 

leladax

Uhm, correct me if I'm wrong, but the EP simulation does not seem to regard stats like hit and expertise after their caps or soft caps according to their 'weaker' status at that point. I can't be getting hit being higher EP, expertise being second highest, and strength third while hit and expertise are both soft capped. It appears to calculate them - correctly - from a zero onwards standpoint according to their soft and hard caps. I don't know the internals so I wouldn't be surprised if personal stats in the simulator aren't even taken into account. Can anyone enlighten us? [edit: source code is available]

 

dr_AllCOM3

You take your paperdoll's stats and they include most of your talents. Just do as told in the readme .

DocsDK UI
Simulator

 

leladax

Is the rotation xml taken at all into account? I read in this thread that the simulator is not rotation based but priority based and the readme does not state how to use rotations.

 

crusaderky

One between rotation and priority is grayed out. The one that is not is the one that will be used. Priority and Rotation are mutually exclusive.

Actually, I too needed several time to figure that out. Maybe you could do something more intuitve? Like a Rotation/Priority mutually exclusive hitbox?


AP from bladed armor should NOT be included in your character XML. In fact, if you've got the bladed armor talent and you set your armor in the character XML to millions, you'll get millions of dps.

 

Fugazor

I wasn't using EP simulation (as I consider it too buggy) but normal simulation and manual calculations.

 

Kapaneus

Hey there, been trying to decide between my old blood spec and dw frost and I figured a simulator would be the easiest cheapest way to figure it out. So I plugged in my stats, spec, and gear for each and ran them through. There was a pretty large discrepancy between the specs (about 3k difference). Does this sound feasible or is it likely that I left something out of a rotation somewhere?

 

dr_AllCOM3

Kahorie's DK Simulator

1.0.0 Release notes

* Several small fixes (2T9 had no multipliers, Annihilation was affecting autoattacks, HB’s weird crit coefficient and more)
* Horn usage shows up in the results and a little result makeover

DocsDK UI
Simulator

 

dr_AllCOM3

(Link is in the first post from now on)

1.0.1 Release notes

* Several small fixes
* 60rp are saved when the Gargoyle/DRW is not on cooldown
* EP calculation is much better now
* (Negative EP results were due to lower than expected dps results aka RNG)

The EP values shouldn't go negative anymore or very rarely. You can blame RNG then. EP cycles use the same random numbers now, resulting in much more stable results. Additionally I formatted the output for better EJ posting. Just slap the table tags around it.
Unholy got a boost, because the Gargoyle is used more often now. Before it rarely got used, because the sim didn't save the rp for it. Also please test SS vs OB. I didn't touch those, but I think the results are different now.

Edit: Set h to 101+ for EP, I won't fix that now .

DocsDK UI
Simulator