SimulationCraft for Rogues - Page 4

dedmonwakeen · 01/28/09, 11:54 AM

Originally Posted by Leto

Does this mean that the "down time" would be inflated for specs with a higher number of attacks per second, since more attacks without the buff present when needed would occur, even if the actual time the buff is missing for each spec is the same?

Well..... Assuming that "fast-attack" spec maintained his higher number of attacks per second both inside and outside of "downtime" I imagine the "fast-attack" ratio would be comparable to the ratio of someone with slow attacks throughout.

Also, just a minor note for combat daggers, Vulajin's spreadsheet estimates that dual wound poison would be better than wound/deadly for it, since dagger hit frequently enough to outweigh the deadly dps.

I'm finishing up a post for the wiki right now, but I'll make sure the config file gets changed for future runs.

dedmonwakeen · 01/28/09, 1:48 PM

Here is data from my most recent run: SampleOutputMelee - simulationcraft - Google Code

Scroll down to the bottom for lots of detail.

Don't be too distracted by the Hunter numbers..... We're still tuning that class. Also, note that the "Gear Overview" chart shows that the profiles are not equally geared (in terms of the sum of stat values).

Here is the config file: http://simulationcraft.googlecode.co...runk/melee.txt

Scroll to the bottom to see the Rogues. Other classes were included in the raid for proper buff/debuffs.

xmod2 · 01/28/09, 1:53 PM

Getting a compile error in sc_rogue.cpp on 64 bit architecture.

g++ -Wall -m64 -O3 -c sc_rogue.cpp -o sc_rogue.o
sc_rogue.cpp:2909: error: no ‘void rogue_t::regen(double)’ member function declared in class ‘rogue_t’
sc_rogue.cpp: In member function ‘void rogue_t::regen(double)’:
sc_rogue.cpp:2917: error: ‘struct player_t::gains_t’ has no member named ‘adrenaline_rush’
make: *** [sc_rogue.o] Error 1

Looks like it's inheriting from player_t alright and none of the other class modules are giving me this error. I am getting the source straight from the svn trunk.

Running make BITS=64

edit: updated source tree works. Figures I'd do a check out during that 10 min window.

dedmonwakeen · 01/28/09, 2:12 PM

Originally Posted by xmod2

Getting a compile error in sc_rogue.cpp on 64 bit architecture.

g++ -Wall -m64 -O3 -c sc_rogue.cpp -o sc_rogue.o
sc_rogue.cpp:2909: error: no ‘void rogue_t::regen(double)’ member function declared in class ‘rogue_t’
sc_rogue.cpp: In member function ‘void rogue_t::regen(double)’:
sc_rogue.cpp:2917: error: ‘struct player_t::gains_t’ has no member named ‘adrenaline_rush’
make: *** [sc_rogue.o] Error 1

Looks like it's inheriting from player_t alright and none of the other class modules are giving me this error. I am getting the source straight from the svn trunk.

Running make BITS=64

Re-update. There was a 10min window there were the source tree was broken. Too much rushing on my part. Sorry 'bout that.

dedmonwakeen · 01/28/09, 4:10 PM

Okay...... Comparing WWS logs I've found several problems in the sim:

(1) Instant attacks with both weapons show up individually in the combat log, while I am combining them into one line..... which is confusing the reporting. (Mutilate, Stormstrike, ...)

(2) I haven't been engaging with Stealth nor using Vanish. That is two lost opportunities for Overkill. In these short PW fights (~160sec) 12sec represents a good chunk of time.

(3) Hunger for Blood is being refreshed far too early due to a flaw in my event rescheduling arch.

#2 and #3 will help close the energy-efficiency gap I see between simulation and wws.

Once I get all three of these complete I'll refresh the wiki page.

dedmonwakeen · 01/28/09, 5:43 PM

I've cleaned up the reporting on Mutilate (and Stormstrike), fixed the early refresh problems with HfB and SnD. I'm -not- taking advantage of Overkill yet.

Output: SampleOutputMelee - simulationcraft - Google Code

Config: http://simulationcraft.googlecode.co...runk/melee.txt

EDIT: New data complete. Contributions from Envenom/Eviscerate still seem low to my untrained eyes, but we're getting closer.

Blizisyimhot · 01/29/09, 1:48 PM

Great work so far on the rogue piece. I appreciate the fact that you decided to model HAT since this simulator is probably the best tool to model the combo points generated based on the performance of party members. I have two things I wanted to ask.

Would it be possible to model 7/51/13 Combat Fist/Dagger Talent Calculator - World of Warcraft which I'm a fan of using a tighter cycle of 4SnD/5Rupture?

Can you explain how or who was in party with the HAT rogue? I noticed a party=3 for all the rogues. Does that mean only the rogues were grouped up together? Would it be possible to run it again with a full group of 5 say the two hunters, enhancement shaman, two rogues?

dedmonwakeen · 01/29/09, 2:07 PM

Originally Posted by Blizisyimhot

Great work so far on the rogue piece. I appreciate the fact that you decided to model HAT since this simulator is probably the best tool to model the combo points generated based on the performance of party members. I have two things I wanted to ask.

Would it be possible to model 7/51/13 Combat Fist/Dagger Talent Calculator - World of Warcraft which I'm a fan of using a tighter cycle of 4SnD/5Rupture?

Can you explain how or who was in party with the HAT rogue? I noticed a party=3 for all the rogues. Does that mean only the rogues were grouped up together? Would it be possible to run it again with a full group of 5 say the two hunters, enhancement shaman, two rogues?

Sure, I can add more specs...... better test coverage is always a good thing.

Your guess is correct: Normally, I allow the party designation to default to zero (I have no max=5 req), but in this case I threw all four rogues together in one party. My original testing had the HAT Rogue in a mixed (full) party, but it didn't seem to matter. I'll check again to be sure, but if you look under the "Procs" section you will see that the HAT Rogue is getting (on average) a CP every 0.5sec which would lead me to believe that he is energy-limited and not CP-limited.

EDIT: I was woefully mistaken: The HAT Rogue was not energy limited. Simply putting a 5th Rogue in the party was sufficient to dramatically increase his DPS. The extra Rogue was the Combat-Daggers-SerratedBlades spec you mentioned. For what its worth: This spec did not perform nearly as well as the VilePoisons Combat-Daggers spec, but the Rogue module is still in its infancy, so I'm not about to make any claims.... I'll update the links shortly.

Leto · 01/29/09, 3:04 PM

Originally Posted by dedmonwakeen

Sure, I can add more specs...... better test coverage is always a good thing.

Your guess is correct: Normally, I allow the party designation to default to zero (I have no max=5 req), but in this case I threw all four rogues together in one party. My original testing had the HAT Rogue in a mixed (full) party, but it didn't seem to matter. I'll check again to be sure, but if you look under the "Procs" section you will see that the HAT Rogue is getting (on average) a CP every 0.5sec which would lead me to believe that he is energy-limited and not CP-limited.

EDIT: I was woefully mistaken: The HAT Rogue was not energy limited. Simply putting a 5th Rogue in the party was sufficient to dramatically increase his DPS. The extra Rogue was the Combat-Daggers-SerratedBlades spec you mentioned. For what its worth: This spec did not perform nearly as well as the VilePoisons Combat-Daggers spec, but the Rogue module is still in its infancy, so I'm not about to make any claims.... I'll update the links shortly.

Just for clarification, the serrated blades spec uses a fist weapon main hand and sinister strike as the combo point builder.

Also, hunters are much better party members for a HaT rogue since their pets count as party members, and thus provide multiple sources of abilities that can crit per second.

dedmonwakeen · 01/29/09, 3:56 PM

Originally Posted by Leto

Just for clarification, the serrated blades spec uses a fist weapon main hand and sinister strike as the combo point builder.

Also, hunters are much better party members for a HaT rogue since their pets count as party members, and thus provide multiple sources of abilities that can crit per second.

Thanks for the clarification on the Serrated Blades spec..... I'll re-run again and post new numbers.

The sim does send pet crits to the HAT Rogue.... I keep expecting the Rogue to become CP-saturated..... but it isn't happening as fast as I would expect.

I'll create a ridiculously good scenario for the HAT Rogue and see if I can get his energy to flat-line.

EDIT: I can't get that darn HaT Rogue to flat-line. Relentless Strikes makes a 5CP Evis too cheap.

Leto · 01/29/09, 4:07 PM

Originally Posted by dedmonwakeen

Thanks for the clarification on the Serrated Blades spec..... I'll re-run again and post new numbers.

The sim does send pet crits to the HAT Rogue.... I keep expecting the Rogue to become CP-saturated..... but it isn't happening as fast as I would expect.

I'll create a ridiculously good scenario for the HAT Rogue and see if I can get his energy to flat-line.

I'd imagine energy capping would be tricky, considering how powerful the bugged HaT of last patch was. It basically had unlimited combo points, and rogues were practically able to hit eviscerate on every gcd.

Thinking about it more, I'm not sure you can energy cap with infinite combo point generation. Relentless strikes procs 100% of the time with a 5 cp finisher, making the cost of eviscerate 10 energy. You regen ten energy in the time of 1 gcd, so the energy used on each eviscerate is regained before you can use another.

edit You edited your post while I was replying, and you came to the same conclusion as I did

.

dedmonwakeen · 01/29/09, 4:21 PM

Originally Posted by Leto

Also, hunters are much better party members for a HaT rogue since their pets count as party members, and thus provide multiple sources of abilities that can crit per second.

At the gear point I'm modeling...... The Rogues are generating crits faster than the hunters.

Hmm.... I should be more specific: One Mutilate-spec and three Combat-specs were generating crits faster than two Survival Hunters, one Marksmanship, and one Beast Mastery.

Looking over the pet statistics for the Beast Mastery Hunter it is possible that 4x BM Hunters will generate more crits than the Rogues.

• Vulajin · 01/29/09, 4:31 PM

Originally Posted by dedmonwakeen

At the gear point I'm modeling...... The Rogues are generating crits faster than the hunters.

Hmm.... I should be more specific: One Mutilate-spec and three Combat-specs were generating crits faster than two Survival Hunters, one Marksmanship, and one Beast Mastery.

Looking over the pet statistics for the Beast Mastery Hunter it is possible that 4x BM Hunters will generate more crits than the Rogues.

I hate to ask a potentially very dumb question, but you are modeling that HAT only procs on ability crits, right? Poisons don't count, either.

dedmonwakeen · 01/29/09, 4:37 PM

Originally Posted by Vulajin

I hate to ask a potentially very dumb question, but you are modeling that HAT only procs on ability crits, right? Poisons don't count, either.

Well...... turns out the only one looking dumb right now is me. (And to decorate my dunce-cap further, l found out that my pets were not inheriting their party designation from their owner.)

So.... No poisons...... Can I generally say "no procs at all"? Just primary ability crits?

Leto · 01/29/09, 4:40 PM

Originally Posted by dedmonwakeen

Well...... turns out the only one looking dumb right now is me. (And to decorate my dunce-cap further, l found out that my pets were not inheriting their party designation from their owner.)

So.... No poisons...... Can I generally say "no procs at all"? Just primary ability crits?

Yes, only abilities. That is why rogues aren't good party members for other rogues. Although we attack very fast, we only perform an ability once every 3 or 4 seconds due to energy requirements, while hunters use every gcd and their pets use abilities rapidly as well.

Technically a HaT rogue would be a decent party member for another HaT rogue, but if they rely on each other for combo points, there may be streaks of weak combo point generation.

dedmonwakeen · 01/29/09, 4:46 PM

Okay..... Pets and hunters in same party....... HAT only proc'ing on special abilities that are not themselves procs.....

HAT Rogue + 4 Hunters
DPS=5976 (topping the melee list)
CP generated every 0.4sec

HAT Rogue + 4 rogues
DPS=3937 (bottom of the melee list)
CP generated every 1.1sec

Nothing new to you guys, obviously...... but at least confirmation my modeling is perhaps starting to tighten up.

EDIT: SampleOutputMelee has been updated. (HAT Rogue is with the 4 Hunters.)

Ratak - US KT · 01/30/09, 4:31 AM

Originally Posted by dedmonwakeen

EDIT: SampleOutputMelee has been updated. (HAT Rogue is with the 4 Hunters.)

There looks to be still too big a gap between 15/51/5 and 7/51/13 with the first generating 179 combo points at 1.6 / sec and the second 94 combo points at 3.1 / sec. My expectation would be that a lot of numbers (like combo point generation) would be very similar between the two specs, with the first having more poison uptime and damage and the second off-setting this poison damage by Rupture etc.

Lokkun · 01/30/09, 5:12 AM

Gains:

...

Rogue_15_51_05:
adrenaline_rush=176.2
combat_potency=952.4
energy_regen=3538.7
relentless_strikes=679.0

Rogue_07_51_13:
combat_potency=957.2
energy_regen=3545.1
relentless_strikes=232.0

...

It looks like the Rogue_07_51_13 is not using Adrenaline Rush.

dedmonwakeen · 01/30/09, 8:11 AM

Originally Posted by Lokkun

It looks like the Rogue_07_51_13 is not using Adrenaline Rush.

Ah, good catch. Only one of the 3 Combat Rogues had adrenaline_rush in their profile.
I'll re-run the sim once I get to work.

In the meantime, I've updated the SampleOutputMelee wiki: Went into the wee hours of the morning last night running a microscope over the Mutilate Rogue output...... It is starting to match up with WWS parses......

Lokkun · 01/30/09, 9:34 AM

I may have found one reason why the Rogue with 7/51/13 provides such a low combopoint-generation, if this is the used configuration (I took it from .svn/trunk/melee.txt):

#
# Combat Rogue (Fist, Serrated Blades)
#
rogue=Rogue_07_51_13
party=3
level=80
talents=http://www.wowhead.com/?talent=f0xbZMgVbxbxoru0xRtxbc0h
actions=flask,type=endless_rage/food,type=great_feast/apply_poison,main_hand=wound,off_hand=deadly
actions+=/auto_attack/pool_energy/slice_and_dice,min_combo_points=4,refresh_at=5/tricks_of_the_trade,target=Rogue_15_51_05
actions+=/killing_spree/blade_flurry/rupture,min_combo_points=5/sinister_strike/adrenaline_rush
gear_strength=0
gear_agility=1100
gear_stamina=950
gear_intellect=0
gear_spirit=0
gear_attack_power=2350
gear_armor_penetration_rating=0
gear_expertise_rating=215
gear_hit_rating=250
gear_crit_rating=675
gear_haste_rating=280
# Not fair to use the same stats since Combat Rogue needs less expertise.
# Trade expertise for ivalue-equal amount of attack power
enchant_attack_power=160
enchant_attack_expertise_rating=-80
glyph_rupture=1
glyph_slice_and_dice=1
glyph_blade_flurry=1
relentless_earthstorm=1
tier7_2pc=1
tier7_4pc=1
main_hand=fist,dps=171,speed=2.6,enchant=berserking
off_hand=dagger,dps=156,speed=1.4,enchant=berserking
# Throw in a pair of Paladins.......  Eventually these static parms will be replaced with real actors.
blessing_of_kings=1
blessing_of_might=688
sanctified_retribution=1
leader_of_the_pack=1

I would recommend a Glyph of Sinister Strike instead a Glyph of Blade Flurry. That should boost the CP/Sec, but the difference between 15/51/5 and 7/51/13 is still too high, as Ratak already stated. I don't think that a missing SS-Glyph is the only reason for the difference between the two combat specs:

Rogue_15_51_05: combo_points=187.0|1.5sec
Rogue_07_51_13: combo_points=96.9|3.0sec

Edit: The difference is 90,1 combo pionts over the duration of the run. Rogue_07_51_13 had 116.2 sinister strikes during that period. Unfortunately I have failed in finding the crit rate of the sinister strikes. The only value I've found was 51.6% but it is stated as "crit per execute" and I have to admit that I am not aware of it's meaning. But from my experience this crit value can be achieved in 25man raids. So by calculating 51.6% as crit rate for sinister strike this leads to ~60 critical sinister strikes. A SS-Glyph provides a 50% chance on critical sinister strikes to award an additional combo point. So using a SS-Glyph would boost the combo points from 96.9 to ~126.9. Still missing 60.1 combo points. Accounting the 20 combo points Rogue_15_51_05 gains from ruthlessness, the total difference between the two specs should be more closer to 187 : 167. So all in all Rogue_07_51_13 is missing somewhere around ~40.1 combo points. If there is a mistake in this calculation please point at it. The only difference between the two specs is ruthlessness, so the combo point generation deficit of 7/51/13 should only be the amount of the benefit ruthlessness gives.

One other thing: I glanced at the "Gains" for Rogue_18_51_02 and it appears that Adrenalin Rush isn't used as well for the 18/51/2-setup.

Ratak - US KT · 01/30/09, 9:37 AM

Player=Rogue_07_51_13 DPS=4177.7 (Error=+/-1.8 Range=+/-332) DPR=238.1 RPS=17.5/17.3 (energy)
Core Stats: strength=124 agility=1418 stamina=1160 intellect=41 spirit=72 health=15699 mana=0
Spell Stats: power=0 hit=9.5% crit=14.7% penetration=0 haste=8.5% mp5=0
Attack Stats: power=4880 hit=7.6% crit=31.4% expertise=6.6 penetration=0 haste=8.5%

Armor Penetration effect of Serrated Blades seems to be missing - there should be a passive armor penetration reduction of 560 (i.e. -560) - or at least it's not reported above which could be misleading.

Drizzit · 01/30/09, 11:28 AM

I ran a few tests with SVN r1445, and found the following:

Melee.txt unchanged (for reference):

    Rogue_15_51_05:
        combo_points=190.5|1.5sec

    Rogue_07_51_13:
        combo_points=98.4|3.0sec

Changed the Blade Flurry Glyph to Sinister strike for Rogue_07_51_13:

    Rogue_15_51_05:
        combo_points=191.9|1.5sec

    Rogue_07_51_13:
        combo_points=112.0|2.6sec

Added eviscerate to the actions for Rogue_07_51_13:

    Rogue_15_51_05:
        combo_points=190.2|1.5sec

    Rogue_07_51_13:
        combo_points=169.9|1.7sec

In addition, there is a duplicate line for sanctified retribution in the rogue_15_51_05 configuration in melee.txt. This does not seem to affect DPS or CP generation.

Edit: Finally had the chance to look over the log. Due to there being no finisher other than Rupture and SnD, sinister strikes were used up to 2 times after 5 combo points had been generated. This resulted in lost combo points when the simulation was run, as no combo points can be generated when you're already at 5 CP.

Lokkun · 01/30/09, 12:31 PM

That is interesting! Just adding Eviscerate to the cycle, I wouldn't have expected an increase of combo point generation. Let's see, the action for Slice and Dice is:

actions+=/auto_attack/pool_energy/slice_and_dice,min_combo_points=4,refresh_at=5/tricks_of_the_trade,target=Rogue_15_51_05

This means (please correct me if I am wrong), that Rogue_07_51_13 is auto-attacking and pooling energy since a) combo points are 4 and b) the SnD-timer is remaining at least 5 seconds. That would mean that the Rogue pools energy over 100 if the SnD-timer is above 5 seconds, right? I think this would explain the issue that combo point generation is lower than expected. (If I am wrong with the interpretation of the action for SnD, you should just skip the following)

Instead the Rogue should refresh the SnD related to the Rupture timer and SnD timer. If no Rupture is is up, just SnD after 4 cp and start building for Rupture. I cannot provide math here, but experience values, since I have raided about 7 weeks with that spec. A good timing to hit the SnD (if you are already at 4cp) is, when rupture remains ~10 seconds. This value may vary +/- 1 second depending on gear and buffs. In addition do not let SnD drop below 5 seconds. But you should avoid pooling energy beyond 100 as a rogue. This will waste combo points.

Leto · 01/30/09, 12:32 PM

Originally Posted by Lokkun

I may have found one reason why the Rogue with 7/51/13 provides such a low combopoint-generation, if this is the used configuration (I took it from .svn/trunk/melee.txt):

I would recommend a Glyph of Sinister Strike instead a Glyph of Blade Flurry. That should boost the CP/Sec, but the difference between 15/51/5 and 7/51/13 is still too high, as Ratak already stated. I don't think that a missing SS-Glyph is the only reason for the difference between the two combat specs:

Rogue_15_51_05: combo_points=187.0|1.5sec
Rogue_07_51_13: combo_points=96.9|3.0sec

Edit: The difference is 90,1 combo pionts over the duration of the run. Rogue_07_51_13 had 116.2 sinister strikes during that period. Unfortunately I have failed in finding the crit rate of the sinister strikes. The only value I've found was 51.6% but it is stated as "crit per execute" and I have to admit that I am not aware of it's meaning. But from my experience this crit value can be achieved in 25man raids. So by calculating 51.6% as crit rate for sinister strike this leads to ~60 critical sinister strikes. A SS-Glyph provides a 50% chance on critical sinister strikes to award an additional combo point. So using a SS-Glyph would boost the combo points from 96.9 to ~126.9. Still missing 60.1 combo points. Accounting the 20 combo points Rogue_15_51_05 gains from ruthlessness, the total difference between the two specs should be more closer to 187 : 167. So all in all Rogue_07_51_13 is missing somewhere around ~40.1 combo points. If there is a mistake in this calculation please point at it. The only difference between the two specs is ruthlessness, so the combo point generation deficit of 7/51/13 should only be the amount of the benefit ruthlessness gives.

One other thing: I glanced at the "Gains" for Rogue_18_51_02 and it appears that Adrenalin Rush isn't used as well for the 18/51/2-setup.

The difference is likely that 15/51/5 performs more finishers, and thus gets more ruthlessness procs.

It looks like the pooling logic accounts for some of it as well.

dedmonwakeen · 01/30/09, 12:42 PM

Originally Posted by Drizzit

I ran a few tests with SVN r1445, and found the following......

Thanks.... I've integrated your changes into the trunk: r1447

I'll regenerate the wiki page shortly.

EDIT: A bunch of posts happened while I was typing....

Originally Posted by Ratak - US KT

Armor Penetration effect of Serrated Blades seems to be missing - there should be a passive armor penetration reduction of 560 (i.e. -560) - or at least it's not reported above which could be misleading.

Ah..... I have it modeled as a "dynamic" buff so it is not getting included in that report. I'll clean that up.

Hmmm.... The report is only including the "percentage" version of penetration which is why Serrated Blades was not displayed.

Originally Posted by Lokkun

That is interesting! Just adding Eviscerate to the cycle, I wouldn't have expected an increase of combo point generation. Let's see, the action for Slice and Dice is:

actions+=/auto_attack/pool_energy/slice_and_dice,min_combo_points=4,refresh_at=5/tricks_of_the_trade,target=Rogue_15_51_05

This means (please correct me if I am wrong), that Rogue_07_51_13 is auto-attacking and pooling energy since a) combo points are 4 and b) the SnD-timer is remaining at least 5 seconds. That would mean that the Rogue pools energy over 100 if the SnD-timer is above 5 seconds, right? I think this would explain the issue that combo point generation is lower than expected. (If I am wrong with the interpretation of the action for SnD, you should just skip the following)

Instead the Rogue should refresh the SnD related to the Rupture timer and SnD timer. If no Rupture is is up, just SnD after 4 cp and start building for Rupture. I cannot provide math here, but experience values, since I have raided about 7 weeks with that spec. A good timing to hit the SnD (if you are already at 4cp) is, when rupture remains ~10 seconds. This value may vary +/- 1 second depending on gear and buffs. In addition do not let SnD drop below 5 seconds. But you should avoid pooling energy beyond 100 as a rogue. This will waste combo points.

The pool_energy action is not all that clever: It will "execute" (read: do nothing for 1sec) if the next action in the priority list limited ONLY by energy......

So...... SnD if "rupture expires in 10sec" OR "SnD expires in 5sec" ?