ShadowCraft for Mists of Pandaria - Page 9

Wally1169 · 01/24/13, 8:11 PM

Originally Posted by Indz

The problem is that haste,exp,crit in this specific gearsetup are extremly close together which produces a very long computation time for the reforger. The reforger is hosted on Google App Engine which is free, but has it's limits. I'm already in contact with Tini, who manages the reforger, to solve the issue but it will take some time and testing.

Thanks, I assume this applies to my question as well. I know you're doing this for "fun" (and I appreciate that) so no rush :-)

Pathal · 02/02/13, 9:33 PM

Regarding the phase based combat cycles idea.

https://github.com/dazer/ShadowCraft...c767254181b322

I pushed a rough draft to dazers repo. There's two things I want to mention first in the case that someone is curious about the design of it.

I'm still playing around with managing Killing Spree, and how to shuffle the morphing CD down the algorithm, I've almost decided on one but not yet. As a result, it's actually missing from the final damage breakdown, but it won't be too big a deal so long as you remember that the overal DPS is lower, and the damage breakdown %'s are higher.

The second was a weird issue. Apparently self.base_energy_regen was changing from a floating point 12, to somewhere around 5 million and change after compute_damage() was run once. I don't think we ever change that value in the compute_damage() method (and now determine_stats() method as well), in fact we almost never touch it except to read from it to determine energy_regen in the cycle calcs. Anyways, I reset the value every time we ran combat_attack_counts() to make sure try and make sure it's the right value each time. Haste and such are done in combat_attack_counts() method so it shouldn't be wrong, just ugly. Solved. It's been awhile since I said something that stupid, I suppose I was overdue.

• Aldriana · 02/03/13, 4:36 AM

At a guess: line 1452-4:

if ar:
    self.attack_speed_increase *= 1.2
    self.base_energy_regen *= 2.0

Assuming you pass through this section of code repeatedly (which is common for the inner loops of damage calculation) its pretty easy to see how this would result in vastly inflated energy regen.

tripcf · 02/03/13, 7:19 PM

Hey everyone.

I am looking for a tool to better optimize my gear, don't get me wrong I love using Shadowcraft, but right now I have so much gear it is hard to figure out which pieces give me the highest dps when including reforging.

If anyone has used Rawr, I am looking for something similar. Something that allows me to select that pieces I have, and then using that catalog of gear it calculates what gear set gives me the most DPS.

I have done some searching, here on EJ and elsewhere... but I can't find anything that really does this.

If you all can point me into the right directions I would really appreciate it!

Thanks!

Solodrow1 · 02/05/13, 11:43 AM

@ Tripcf

Have you tried askmrrobot? I use that tool. It allows you to swap gear and gives the ideal gem/enchant set up for the new piece and an estimated dps gain

fierydemise · 02/09/13, 1:58 PM

For handling combat energy capping what if instead of trying to determine how much energy is actually usable like you are doing currently we calculate attacks_per_second assuming no gcd and then cap the sum of evis, ss and rvs attacks per second at 1 or 1.25 with AR glyph. The current implementation allows nearly 2 aps when using stack_cds with t15_4pc.

Quick and dirty implementation (assumes anticipation), probably improvable by someone more familiar with python.

sum_aps = attacks_per_second['sinister_strike'] + attacks_per_second['revealing_strike'] + attacks_per_second['eviscerate'][5]
aps_cap = 1;
if ar: aps_cap=1.25
if sum_aps > aps_cap:
    ss_ratio = attacks_per_second['sinister_strike']/sum_aps
    rvs_ratio = attacks_per_second['revealing_strike']/sum_aps
    evis_ratio = attacks_per_second['eviscerate'][5]/sum_aps
    aps_overage = sum_aps-aps_cap;
    attacks_per_second['sinister_strike'] -= ss_ratio*aps_overage
    attacks_per_second['revealing_strike'] -= rvs_ratio*aps_overage
    attacks_per_second['eviscerate'][5] -= evis_ratio*aps_overage

Pathal · 02/09/13, 3:51 PM

That doesn't handle SS and GCDs spent on Slice and Dice (might encourage oddities with rupture too), but it's another method that could be more accurate with improvements.

fierydemise · 02/09/13, 4:30 PM

Fair point in SnD, I also realized that what I have there isn't quite right because GCD's spent on RvS should be fixed since we assume full RvS uptime. For handling SnD and RvS we should have two pools of aps, reducable (SS, Evis, Rupture) and unreducable (SnD, RvS). We sum over both pools to determine avs_overage but only reduce by the reducable pool ratio so methign like:

ss_ratio = attacks_per_second['sinister_strike']/reducable_aps
...
aps_overage = (reducable_aps+unreducable_aps)-aps_cap;
attacks_per_second['sinister_strike'] -= ss_ratio*aps_overage

Rupture can I think be dealt with by simply moving the rupture tick calculation to after the energy cap calculation. Rupture_ticks don't appear to be needed by anything else in the attack_per_second method so that should be safe. We could also add an option prioritize_evis_during_energy_cap which would set rupture to 0 before calculating the overage ratios.

RemJay24 · 02/09/13, 10:34 PM

I am aware Shadowcraft isn't perfecting Sub 100%, but I am seeing a 1% DPS decrease if I choose my cp builder as Hemo only. I can't see how dropping backstab yields a 1% drop. The dps decrease nearly matches the 2p bonus increase (since the 2p benefits backstab and not hemo). Is this one of many flaws with Sub in shadowcraft - or am I doing something completely wrong? (I wanted to see how much of a drop it would be to ignore Backstab.)

On that note; might Sub get some love from SC in the near future? (possibly come 5.2)

Thanks!

Pathal · 02/10/13, 2:10 PM

@fierydemise
I uploaded an implementation of that. I tried a shortcut so I didn't have to pass phase lengths through and manage GCDs that way, not sure if it's the best solution at the moment since reaction time and latency should be a factor. I broke free GCDs down into portions of a second, subtracted the GCDs from SnD from 1s (multiplied by the GCD size in the case of the AR glyph), and took the remaining portion of the second and figured out how many evisc cycles fit in that, then divided that by evisc_cycles/remaining_gcds if it was over.

Seems to be coming together overall though.

@RemJay
I talked about this a couple months ago, actually. I'm sure Sub could be done better. I haven't really touched it since Mists' release though, since there were very few high end players using it, and I couldn't afford to use it on progression to gather logs to see where things might be going wrong (if they were / are). If I can find some time between Combat stuff and 5.2, maybe I'll be able to smooth it out. Not a promise, it's a "maybe I'll find time to might be able to fix something".

I'm not playing much right now, so I would have to debug using other's logs and there really aren't many good ones out there.

Indz · 03/06/13, 6:41 AM

Shadowcraft is now updated for Patch 5.2.

Please note some things:

Items
- mostly thunderforged Items are missing. If there are other Items missing like trinkets/weapons, please let me know.

General
- the server is under heavy load so all is a bit slow

Engine/Patch
- Switching between the Patches only changes the UI, so you can't compare the old values with the new. It always uses the most recent 5.2 Engine.

Trinket Ranking
- I can't say if the ranking is correct or wrong. I think Pathal can tell you more about it. You can always check https://github.com/dazer/ShadowCraft-Engine/.
- Please also note that the trinkets are ranked with an suggested best reforge or no reforge. There is one trinket with expertise and one with hit so you need to plug the trinket in, hit reforge all and watch dps.

Itemset Ranking
- It's possible that your T14 Items are ranked very high in comparising to the new items. Example: You are wearing 3 set items excluding the helm. Now you want to change your helm. The T14 helms are now ranked very high because the T14 4pc EP bonus is added to the item. You see something like this "4372.0 base / 143.5 reforge / 2259.7 gem (Match gems) / 3240.6 set".

If you are having other issues or found a bug please let me know.

Pathal · 03/06/13, 6:58 AM

RoRO might be high, Talisman and Bad Juju might be low. I need to find some logs of actual use before I can feel truly confident.

I don't remember Combat being slow (I think Nextormento benchmarked it as faster than it was beforehand), but I can take another look tonight to see if I can speed it up.

Pathal · 03/07/13, 11:46 AM

With the introduction of Heroic Thunderforged pieces, we have to think a little more about BiS and EP weights.

Technically, BiS means the best set of gear to achieve best possible DPS. With Heroic Thunderforged pieces, that means the item would most likely not be a normal heroic piece (usually with the exception of some tier pieces). Due to the statistical likelihood of being full Heroic Thunderforged, which as I understand it is rather low even for the most seasoned raiders, I think there should be a consensus of some sort here. Making sure everyone's on the same page will alleviate potential confusion in discussion.

So, for practical reasons, I'm thinking we should construct our BiS list and EPs with heroic only, not heroic thunderforged pieces.

Nalinaa · 03/07/13, 4:50 PM

Not to dismiss what you said Pathal, but in my 25 man on tuesday we were getting, what seemed to be, at least 2 thunderforged pieces per boss, and you can get them from coins.

But I agree that BiS list should be plain heroic gear, because most of the time BiS lists are used for getting EP values for the tier and to give plays an idea of which item they should get for which slot, and what off piece we want from tier.

edit:removed a part that wasn't correct.

Schmoopy · 03/07/13, 7:13 PM

Originally Posted by Nalinaa

Not to dismiss what you said Pathal, but in my 25 man on tuesday we were getting, what seemed to be, at least 2 thunderforged pieces per boss, and you can get them from coins.

But I agree that BiS list should be plain heroic gear, because most of the time BiS lists are used for getting EP values for the tier and to give plays an idea of which item they should get for which slot, and what off piece we want from tier.

edit:removed a part that wasn't correct.

RNG can be in and against your favor and a sample size of 1 isn't very sound for statistical analysis.