Jump to content


Photo

FeralbyNight (Cat & Bear simulation tool)


  • This topic is locked This topic is locked
1584 replies to this topic

#41 Drashian

Drashian

    Von Kaiser

  • Members
  • 36 posts

Posted 17 March 2009 - 09:29 PM

Practically, it's probably only good as a buffer against lag, yeah. Theoretically though, it does allow you to squeeze slightly more attacks into less GCDs if you don't mind letting a cooled-down ability sit around for up to a second. An example, with 3/3 Improved Mangle, from a quick script I wrote:

0.0 Faerie Fire (cools down at 6.0)
1.0 Mangle (cools down at 5.5)
2.5 Lacerate/Swipe
4.0 Lacerate/Swipe
5.5 Mangle (cools down at 10.0)
7.0 Faerie Fire (cools down at 13.0)
8.0 Lacerate/Swipe
9.5 Lacerate/Swipe
11.0 Mangle (cools down at 15.5)
12.5 Lacerate/Swipe
14.0 Faerie Fire (cools down at 20.0)
15.0 Lacerate/Swipe
16.5 Mangle (cools down at 21.0)
18.0 Lacerate/Swipe
19.5 Lacerate/Swipe
21.0 Last GCD over, restart the cycle (Mangle and Faerie Fire are lined up again)

Not enough time left at work to determine if this cycle actually produces more threat (since it's wasting parts of Mangle/FF cooldowns), and if it's not, to see if there is a cycle that does. If nobody else takes up the challenge I'll try next chance I get.

#42 Finito

Finito

    Glass Joe

  • Members
  • 12 posts

Posted 17 March 2009 - 10:09 PM

Does anyone have a tanking stats comparison similar to what was done in this thread with cat gear?

The link to the weighted Wowhead scale was extremely helpful for dps gear; it would be very handy to have one for tanking too. This might be annoying to TC with the bearshield changes, but there are fewer defensive stats than offensive so maybe it wouldn't be awful. I'm also not entirely sure how the 10% stamina nerf to heart of the wild works in terms of stamina gear scaling, since I'm not sure where the 10% factors in (additively or multiplicatively with the other multipliers).

#43 seatil

seatil

    Glass Joe

  • Members
  • 1 posts

Posted 18 March 2009 - 01:25 AM

I have some comments on the source code (if you're not a programmer you may wish to forego the rest of this post). I'm working from FeralbyNightv3_1_beta.cpp downloaded from the Google code linked to in the OP. I was unable to find anything more recent than March 10th and the subversion repository is empty.

My compiler flagged a couple of suspect if-checks at lines 2079 and 2121. The code in question is:

2079: if(enraged=1){
2080: enraged_l=1;
2081: }

It can be fixed by making 2079 read "if(enraged==1){".

I believe the intended behaviour is to check if the value of enraged is 1 and if it is, setting the value of enraged_l to 1. The actual behaviour is rather different: Set the value of enraged to 1, if-check always succeeds, set value of enraged_l to 1, regardless of the initial value of enraged.

Line 2121 contains the same error.

Also, the compiler is flagging a large number of implicit double => int casts. I think that most of these are deliberate and can be tidied up by just explicitly casting the values (int foo = (int) ...). Tidying the deliberate ones up would make it easier to find any unintentional ones. Your compiler should be flagging these -- if it's not then you'll need to enable the appropriate compiler warnings -- unfortunately I can't help with VC++ as I'm not familiar with that software.

#44 nightcrowler

nightcrowler

    Don Flamenco

  • Members
  • 448 posts

Posted 18 March 2009 - 01:07 PM

To seatil: thx for error finding. I'll change it now. As for the int > double, yes i should explicit cast it, i'll review it soon but it's not on my priority list. thanks.

For all: I've an hard time at work this day so I've not time to update the link and put on a bear list. I'll do it as soon as possible.

#45 Ugleestick

Ugleestick

    Glass Joe

  • Members
  • 1 posts

Posted 18 March 2009 - 09:20 PM

Hi, I think it's awesome what you're doing and I hope you get lots of support : ) However, I was looking at your DPS values for Str and Agi, and was wondering where you got your numbers?

I know currently Str > Agi by a good margin (Like, until you get 13KAP w/o savage roar and 30% crit, STR > Agi) but I'm not sure how much less important Str is going to be in 3.1 since savage roar is just raw damage instead of AP.

#46 Treescendant

Treescendant

    Glass Joe

  • Members
  • 2 posts

Posted 18 March 2009 - 09:34 PM

Hi, I think it's awesome what you're doing and hope you complete it and get lots of support : ) However, I was looking at your DPS values for Str and Agi. Where did you get those numbers? I know currently Str > Agi by a good margin (Like, until you get 13KAP w/o savage roar and 30% crit, STR > Agi) but I'm not sure how much less important Str is going to be in 3.1 since savage roar is just raw damage now instead of AP.

I think the reason Agi outperforms Str in 3.1 is due to the Primal Gore talent allowing the periodic damage from Rake, Lacerate and Rip to crit. Since Agi increases our crit chance it becomes the more sought after stat.

#47 Thaeryn

Thaeryn

    Piston Honda

  • Members
  • 174 posts

Posted 18 March 2009 - 10:56 PM

It's also due to the change in Savage Roar. Before, it scaled directly from your AP, so stacking Strength was the best way to take advantage of it. Now that it's just a flat damage increase, it scales from both the crit and AP we get from Agility. This combined with the bleed critting puts it on top pretty easily.

#48 ramenchef

ramenchef

    Piston Honda

  • Members
  • 166 posts

Posted 21 March 2009 - 02:03 AM

Your armor value seems high for a pre-inspiration 3.1 value. Or is this 3.0.9?

#49 nightcrowler

nightcrowler

    Don Flamenco

  • Members
  • 448 posts

Posted 21 March 2009 - 05:31 AM

Your armor value seems high for a pre-inspiration 3.1 value. Or is this 3.0.9?


After. Pre 3.1 is around 38k.

#50 Malazaar

Malazaar

    Don Flamenco

  • Members
  • 336 posts

Posted 21 March 2009 - 09:21 AM

It's also due to the change in Savage Roar. Before, it scaled directly from your AP, so stacking Strength was the best way to take advantage of it. Now that it's just a flat damage increase, it scales from both the crit and AP we get from Agility.


SR improved crit scaling before, the only real change is primal gore greatly increasing agility scaling and the SR change decreasing ap/str scaling by ~3 % (including the glyph).

#51 nightcrowler

nightcrowler

    Don Flamenco

  • Members
  • 448 posts

Posted 21 March 2009 - 01:28 PM

I've done a lot of work on the code, now it should be easier to read and quicker. I've made some changes on cat algoritm and changed some gear (trinkets). I've runned bear simulation. Now I only had to finsh the addon. Main page, tables and links updated.

#52 tlbj6142

tlbj6142

    Piston Honda

  • Members
  • 119 posts

Posted 21 March 2009 - 09:27 PM

Bear questions....

Why does haste have a negative value? Doesn't more attacks mean more SD procs?
What exactly does "Night's Choice" mean? IOW, why did you inflate some values over those in the first column (like hit)?

#53 Deathwing

Deathwing

    Bald Bull

  • Members
  • 1,126 posts

Posted 22 March 2009 - 01:17 AM

I'm guessing haste has a negative value because it increase parry gibs.

#54 nightcrowler

nightcrowler

    Don Flamenco

  • Members
  • 448 posts

Posted 22 March 2009 - 03:25 AM

Night's Choice is my personal stat choice, if you want to know how is build is the following:

70% Stamina-equivalent tankpoints + 30% ap-equivalent TPS. After that I've inflated by 20% hit an expertise and reduced by 10% ArPen. Reason for those changes: hit it's the best threat stat (after expertise) because make it stable, we all know that missing 2 hit at the begining of the fight or missing a taunt is really really bad. Expertise because parry gibbs can shoot you but obviously an averaged out fight can't weight that. ArPen was reduced because is the only stats affecting only threat, with boss hitting for less all other aggressive stats increase their TP value except ArPen.
Haste has a negative effect due to increasing parry gibbs and a positive effect due to increasing shield chance. The border line is at 52 expertise. (If you are expertise capped 1 haste = 3 Stam-EQ_TP).

Chocing gear for tanking is not like chocing it for cat. DPS is 1 number and you should maximize it. For tanking & healing the situation is more complicated and more subject to flavor (for example I personally like to gem for stam/exp) that's why I've put TP, DPS, TPS in separate collumns and then suggested my value, but you can build your own starting from them.

Edit: changed 2t8 bonus with the right proc chance (2%).

#55 Allev

Allev

    King Hippo

  • Members
  • 545 posts

Posted 23 March 2009 - 04:15 PM

From another thread, I thought it was worth bringing the discussion of the topic here...

Here is an example report from my sim (best in slot and raid buffed, 2T7+shred glyph + SR glyph):

Cat (crit% vs. a boss (-4.8% added)): 51.6857%
Cat (AP): 10738
Expertise: 25
Hit: 8.02074%
Armor Penetration: 24.5205%
Haste (from equipment): 10.6648%


Boss base Armor (before penetration and debuff): 11900
Boss modified Armor (after penetration and debuff): 6826.37
Boss Damage Reduction (after debuff and penetration): 30.9461%
Weapon Speed Reduction % (Total with Windfury): 27%
Missed % (Total): 0%
Dodged %: 0.248855%

5 minute fights (10k simulations)

with rip glyph:

DPS: 6888.68 +/- 2.11081
White cat: 2030.19, Yellow cat: 4858.49
White: 29.4713%
Mangle: 5.20451%, Avg dmg: 4589, #/cycle: 2.13419),
Shred: 23.4303%, Avg dmg: 8003.44, #/cycle: 5.509),
RIP: 20.8473%, Avg dmg: 44621.2, #/cycle: 0.879185)
Rake: 16.7622%, Avg dmg: 12987.9, #/cycle: 2.42863),
Ferocious Bite: 4.28438%, Avg dmg: 15324.9, #/cycle: 0.526092, Average Energy Usage: 35)
Savage Roar uptime: 98.0435%
Average Cycle Time: 27.3172 seconds
Mangle debuff uptime: 91.937%, Mangle-RIP debuff downtime: 7.20694%

with mangle glyph:

DPS: 6802.07 +/- 2.10665
White cat: 2031.44, Yellow cat: 4770.63
White: 29.865%
Mangle: 3.59395%, Avg dmg: 4564.29, #/cycle: 1.47221),
Shred: 25.9634%, Avg dmg: 8004.4, #/cycle: 6.0646),
RIP: 19.1706%, Avg dmg: 38298.9, #/cycle: 0.935882)
Rake: 18.0811%, Avg dmg: 13035.3, #/cycle: 2.59342),
Ferocious Bite: 3.32593%, Avg dmg: 15329.1, #/cycle: 0.405664, Average Energy Usage: 35)
Savage Roar uptime: 98.5194%
Average Cycle Time: 27.4871
Mangle debuff uptime: 93.9575%, Mangle-RIP debuff downtime: 4.7852%


So, with that output, SR uptime looks good-- but Rake uptime is very low. I know there's logic in there that's supposed to hold off on shredding when mangles and rakes are expiring (specifically "(mdebuff_timer>=((rake.energy-(10+en-shred.energy))*10))&&(rake_timer>=((rake.energy-(10+en-shred.energy))*10))".

Is this code working effectively though? I'd think we can get better uptimes than that if we're purposely waiting for energy. (For reference-- the default mangle profile in SimulationCraft achieves similar uptimes without waiting on energy for rakes-- which means, we should be able to do better).

#56 tlbj6142

tlbj6142

    Piston Honda

  • Members
  • 119 posts

Posted 24 March 2009 - 05:25 PM

For tanking & healing the situation is more complicated and more subject to flavor (for example I personally like to gem for stam/exp) that's why I've put TP, DPS, TPS in separate collumns and then suggested my value, but you can build your own starting from them.

So, do you hope to gain avoidance, crit and hit from gear, enchants (+agi/stats), food, potions and gem +sta/exp?

#57 CiC

CiC

    Glass Joe

  • Members
  • 2 posts

Posted 25 March 2009 - 09:33 AM

Armor Penetration: 24.5205%

someone can tell me how much ArPenRate it is to get the 24%?
and does it mean that ArPen will get one of the best stats with upcoming changes?

#58 Huggme

Huggme

    Von Kaiser

  • Members
  • 28 posts

Posted 25 March 2009 - 10:02 AM

someone can tell me how much ArPenRate it is to get the 24%?
and does it mean that ArPen will get one of the best stats with upcoming changes?


15.395 ArPenRating = 1% Armor Penetration (correct me here if wrong, took it from feral numbers)
It seems like ArPen is going to be very strong stat, semiquoting nightcrowler:

Stats value
[TABLE]Stat | points
Armor Pen (shred) | 28.4
Agility | 26.9
Armor Pen (RIP) | 26.8[/TABLE]


So with shred idol, ArPen is the best stat with the gear selection nightcrowler used in his sim.
With rip idol, ArPen is very close to agi.

Keep in mind that the more ArPen you have, the better the stat becomes. So if you have almost none on your gear now, agility is better. 24% is pretty high, I personally have around 11% on my PTR char.

You get to know more if you read night's post, he explains ArPen further in Discussion section. Or you could read Feral numbers thread, where Wulfmanjack posted how good ArPen is with Ulduar loot, and how it gets even better if you gem for ArPen:

Gemmed for armorpen the value goes even higher
Armorpenetration Rating: 1,73
Agility: 1,60


Edit: Sorry to Nightcrowler for ruining his name

#59 SquidHC

SquidHC

    Glass Joe

  • Members
  • 1 posts

Posted 26 March 2009 - 01:11 AM

Do you have an ETA for this add-on? I am very interested to see how it is going to work =D

#60 Fasc

Fasc

    Von Kaiser

  • Members
  • 27 posts

Posted 26 March 2009 - 02:39 AM

Do you have an ETA for this add-on? I am very interested to see how it is going to work =D


With Nightcrowler getting himself a rather lengthy temporary ban, I'm not sure what the ETA for the add-on or updates to the current models/simulations will be. I too am anxious to see it in action since most of his information seems to be very good theorycrafting.

If anyone is in contact with him (he has his profile set to not receive e-mails or accept private messages) and could either relay information or post up a location where he is continuing this research and development, it would be greatly appreciated.




1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users


    Bing (1)