Jump to content


Photo

SimulationCraft for Rogues


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

#21 drumbum

drumbum

    King Hippo

  • Members
  • 500 posts

Posted 17 January 2009 - 02:00 AM

2) Thanks for that catch! Fix committed.


It looks like you are clearing the combo points before calling the Relentless Strikes routine, so it's always going to believe there are zero combo points.

#22 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 17 January 2009 - 03:17 AM

It looks like you are clearing the combo points before calling the Relentless Strikes routine, so it's always going to believe there are zero combo points.


Bah! That's what I get for trying to code during a work meeting..... Fixed.

BTW..... Very thorough code review. I wish we were hiring. I'd offer you a job.

#23 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 17 January 2009 - 01:13 PM

I've added support for:

(1) Hemorrhage damage adder for physical attacks. Currently coded as being applied AFTER glance/crit and BEFORE armor reduction.

(2) Rogue Glyphs

(3) Tier 6 and Tier 7 Set bonuses. (It is unclear to me in exactly what order do Overkill, Slaughter from the Shadows, and Tier7-4pc get combined.....)

#24 drumbum

drumbum

    King Hippo

  • Members
  • 500 posts

Posted 17 January 2009 - 04:24 PM

The Hemo debuff should be added before crit, but I'm not sure if the Hemo debuff is also affected by glance.

#25 chalon

chalon

    Founder of the Chalonverse

  • Members
  • 1,717 posts

Posted 17 January 2009 - 10:56 PM

So, couple of interesting things I noticed regarding the interaction between Overkill and T7 4-piece:
1. The tooltips update to indicate the actual energy cost, and update both for the T7 4 piece and for Overkill. You have to re-mouse over it when the state changes, however. Cursory testing indicates the tooltip is accurate.
2. The energy cost seems to always be floored. For instance, Gouge costs 45 energy. 45 * 0.95 = 42.75. The energy cost shows up as 42, and is not rounded up to 43.
3. Overkill appears to be applied before the T7 4p is calculated. If it were applied afterwards, the Gouge cost would be 32. However, it's 33 which corresponds to: (45 - 10) * 0.95 = 33.25.

#26 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 24 January 2009 - 03:28 AM

Deleted out-dated contents....

#27 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13,510 posts

Posted 24 January 2009 - 04:14 AM

Well, you appear to be using a MH fist with Mutilate, which isn't going to work too well. Try rerunning the Mutilate numbers with a 1.4 dagger MH and see if you get something more plausible.

#28 DSmith13

DSmith13

    Glass Joe

  • Members
  • 4 posts

Posted 24 January 2009 - 04:20 AM

Also, you have MH deadly, OH instant. As of 3.0.8, it should be MH instant (and faster weapon) OH deadly.

#29 PessimiStick

PessimiStick

    Piston Honda

  • Members
  • 160 posts

Posted 24 January 2009 - 04:55 AM

In addition to Aldriana's comment (and that's a big one, heh), refreshing HfB at 5 seconds is pretty sloppy and worse than any good player would be doing. Also, you should be able to cut the CP used on Slice and Dice in the combat actions list to 4.

If I'm reading that output correctly, the White damage component of your Combat set seems to be a bit low compared to in-game results. I believe this is a pretty typical breakdown for something like PW: Wow Web Stats Wow Web Stats Wow Web Stats

#30 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 24 January 2009 - 05:11 AM

Well, you appear to be using a MH fist with Mutilate, which isn't going to work too well. Try rerunning the Mutilate numbers with a 1.4 dagger MH and see if you get something more plausible.


Ah... That's what I get for reusing the same Armory profile in another talent spec......

In addition to Aldriana's comment (and that's a big one, heh), refreshing HfB at 5 seconds is pretty sloppy and worse than any good player would be doing. Also, you should be able to cut the CP used on Slice and Dice in the combat actions list to 4.

If I'm reading that output correctly, the White damage component of your Combat set seems to be a bit low compared to in-game results. I believe this is a pretty typical breakdown for something like PW: Wow Web Stats Wow Web Stats Wow Web Stats


Perfect, thanks! I'll examine the WWS in detail.

Incoming EDIT of results post......

#31 PessimiStick

PessimiStick

    Piston Honda

  • Members
  • 160 posts

Posted 24 January 2009 - 05:30 AM

Can you explain the rupture output to me? For both specs I would expect similar numbers since the only real difference is the AP from Savage Combat and Find Weakness from Mut, yet the numbers in that row are very different for each spec.

What kind of in-game stats does that gear equate to? I don't have any spreadsheets on this machine so I can't easily compare the stats I'm getting from my own gear. I think it's pretty similar to what the rogues in my guild have though, so assuming that your data is based on as long of a duration as it looks like, the combat number seems to be getting close to what I'd expect from the game.

Also, the Evisc numbers seem a bit high, and Envenom very low, again assuming I'm reading the output correctly.

#32 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13,510 posts

Posted 24 January 2009 - 07:02 AM

Looks to me like you're getting too much yellow damage relative to the amount of white damage for Mutilate, and your Relentless Strikes regen seems high relative to the Focused Attacks regen - typically my Relentless Strikes regen is not significantly higher than my Focused Attacks Regen (within 10%, say). So just as an off-the-cuff guess: are you giving Relentless Strikes regen for both the Envenom and the SnD for a CttC Envenom? As it should apply only once in that case.

#33 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 24 January 2009 - 03:13 PM

Can you explain the rupture output to me? For both specs I would expect similar numbers since the only real difference is the AP from Savage Combat and Find Weakness from Mut, yet the numbers in that row are very different for each spec.

What kind of in-game stats does that gear equate to?

Also, the Evisc numbers seem a bit high, and Envenom very low, again assuming I'm reading the output correctly.


The gear_xxx values represent the "green mouse-over" numbers on your character sheet..... essentially, the contributions from gear alone excluding base values and talents.

The Rupture output was bugged..... but the real difference was that the Combat Rogue was waiting for a 5pt Rupture and the Assassination Rogue was going at 4pt. I've changed the Combat Rogue to trigger Rupture/Evis at 4ps and the damage seems more in line.

Well.... At least the Ruptures and Evis/Envenoms seem in line...... but the Assassination Rogue is hitting is triggering Envenom every 5.7sec while the Combat Rogue is hitting Evis every 12.0sec.

The Combo-pt generation seems to weigh too far in favor of the Assassination Rogue: Every 0.9sec vs 1.5sec. That just doesn't seem right.....

Looks to me like you're getting too much yellow damage relative to the amount of white damage for Mutilate, and your Relentless Strikes regen seems high relative to the Focused Attacks regen - typically my Relentless Strikes regen is not significantly higher than my Focused Attacks Regen (within 10%, say). So just as an off-the-cuff guess: are you giving Relentless Strikes regen for both the Envenom and the SnD for a CttC Envenom? As it should apply only once in that case.


I have a feeling that my combo-point generation is to high for Assassination, resulting in far too frequent Envenoms.... resulting in too much gain from Relentless Strikes.

Regarding damage calculation.... Perhaps my core formula is off:

    weapon_speed  = normalize_weapon_speed  ? weapon -> normalized_weapon_speed() : weapon -> swing_time;

    hand_multiplier = ( weapon -> slot == SLOT_OFF_HAND ) ? 0.5 : 1.0;

    power_damage = weapon_speed * direct_power_mod * total_power(); 
    
    direct_dmg  = base_direct_dmg + ( weapon -> damage + power_damage ) * weapon_multiplier * hand_multiplier;
    direct_dmg *= total_dd_multiplier();

"direct_power_mod" defaults to 1.0/14
"base_direct_dmg" is the constant "special" damage associated with some attacks (note that I do not multiply this by the weapon_multiplier)
"total_power" is the total Attack Power from all sources
"total_dd_multiplier" is the total direct-dmg multiplier from all sources (talents, buffs, debuffs, etc)

I use a target armor of 13000, a Sunder debuff of 3925, and a FF debuff of 1260.

Resistance calculation:
double action_t::resistance()
{
  if( ! may_resist ) return 0;

  target_t* t = sim -> target;
  double resist=0;

  double penetration = base_penetration + player_penetration + target_penetration;

  if( school == SCHOOL_BLEED )
  {
    // Bleeds cannot be resisted
  }
  else if( school == SCHOOL_PHYSICAL )
  {
    double adjusted_armor = armor() * ( 1.0 - penetration );

    if( adjusted_armor <= 0 ) return 0;

    double adjusted_level = player -> level + 4.5 * ( player -> level - 59 );

    resist = adjusted_armor / ( adjusted_armor + 400 + 85.0 * adjusted_level );
  }
  else
  {
    double resist_rating = t -> spell_resistance[ school ];

    resist_rating -= penetration;
    if( resist_rating < 0 ) resist_rating = 0;

    resist = resist_rating / ( player -> level * 5.0 );

    if( ! binary )
    {
      int delta_level = t -> level - player -> level;
      if( delta_level > 0 )
      {
        double level_resist = delta_level * 0.02;
        if( level_resist > resist ) resist = level_resist;
      }
    }
  }

  return resist;
}


Results post has been updated.

#34 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 25 January 2009 - 09:50 PM

Deleted out-of-date contents.

#35 koaschten

koaschten

    In the rear with the gear!

  • Members
  • 5,668 posts

Posted 25 January 2009 - 10:51 PM

Rogue_15_51_5:
18.6% : blade_flurry
25.0% : mongoose_oh
47.4% : moongoose_mh
97.4% : savage_combat
86.0% : slice_and_dice


That shouldnt be that low, should it?
How to get an Android Authenticator on your PC. (updated feb'11)

http://goo.gl/fackr


#36 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 25 January 2009 - 11:09 PM

That shouldnt be that low, should it?


Hmm....... It appears the AI is letting Slice and Dice fall off before refreshing. I'll add a "refresh_at" option.

#37 chalon

chalon

    Founder of the Chalonverse

  • Members
  • 1,717 posts

Posted 25 January 2009 - 11:19 PM

Few notes:
1. Blessing of Might is ~688 AP, due to the Imp. Blessing of Might talent

2. Berserking is a better enchant than Mongoose

3. If you want to look at best-in-slot gear, your gear values are quite low. That probably explains why your overall DPS number is so low. Should just grab the numbers from here: The World of Warcraft Armory

4. A Mutilate rogue shouldn't have any uptime on Savage Combat. Though it is probably safe to assume that someone else in the raid is providing that 2% damage boost.

5. How long was the fight durations? I didn't see that noted anywhere, though I'm probably just missing it.

#38 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 26 January 2009 - 02:49 AM

Few notes:
1. Blessing of Might is ~688 AP, due to the Imp. Blessing of Might talent

2. Berserking is a better enchant than Mongoose

3. If you want to look at best-in-slot gear, your gear values are quite low. That probably explains why your overall DPS number is so low. Should just grab the numbers from here: The World of Warcraft Armory

4. A Mutilate rogue shouldn't have any uptime on Savage Combat. Though it is probably safe to assume that someone else in the raid is providing that 2% damage boost.

5. How long was the fight durations? I didn't see that noted anywhere, though I'm probably just missing it.


1. Updated Blessing of Might

2. Added support for Berserking, assume proc rate same as Mongoose

3. I've switched to Aldriana's gear. Since I haven't added support for the melee trinkets yet, I increased the AP by ~400.

4. Fixed the reporting....

#39 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 26 January 2009 - 02:57 AM

5. How long was the fight durations? I didn't see that noted anywhere, though I'm probably just missing it.

If he used his standard raid input, then the fights are around ~280 seconds.
The sim is told to make the fight 300 seconds, so it estimates the health of the mob. This is done by adding up all damage until half of the 300 seconds have passed and assuming that the mob is now at 50% health.
Bloodlust is probably fired up at 35% so that fight are slight less then 300 seconds most of the time.
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#40 path411

path411

    Von Kaiser

  • Members
  • 93 posts

Posted 26 January 2009 - 06:48 AM

I'm not super familiar with SimulationCraft so correct me if I'm wrong.

Maybe I'm missing something, but was wondering:

Under attack stats, both specs have identical AP numbers, even though Savage Combat would increase the AP of the combat rogue by 4% over the mutilate rogue. This also seems to possibly be a problem for the talent Close Quarters Combat, as a combat build would have 2% more crit over a mutilate rogue from having 2 more points in this talent.

The expertise outputs however also seem strange. The mutilate rogue has 6.6 (I'm assuming %). While the combat rogue has 16.6. The 10 expertise gained from the weapon expertise talent is only 2.5%, which if the resulting number is a % would mean the combat rogue should be ending up at 9.1.
(Also a side note, 6.6 seems to be wrong itself, it seems like it is rounding up after taking fraction values of expertise. Where expertise will only improve your dps if it comes out to a flat expertise value.)

Lastly, your combat rogue is running with instant poison instead of wounding.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users