EnhSim, DPS simulator - Page 3

Pitbuller · 09/08/08, 4:05 AM

Off-hand weapon dps ep is allways been very small. Around 3.5ep with 2.6s and Wf. With 1.3s and Ft it's only 2.2ep. Now it's actual so small that if we see fast one-hand caster weapon we could use it and dps will go up. Caster weapon dps also scale in wrath. [Cudgel of Saronity Justice]

Edit: I think it's just oversight and gonna be fixed.

Skiace · 09/08/08, 4:14 AM

Originally Posted by Pitbuller

Off-hand weapon dps ep is allways been very small. Around 3.5ep with 2.6s and Wf. With 1.3s and Ft it's only 2.2ep. Now it's actual so small that if we see fast one-hand caster weapon we could use it and dps will go up. Caster weapon dps also scale in wrath. [Cudgel of Saronity Justice]

I thought caster weapons were all supposed to be Main-Hand only. Is this weapon an oversight or are many caster weapons in Wrath One-Hand?

Pitbuller · 09/08/08, 6:22 AM

Another discovery:
I tested new naxx weapons and 1.8s caster weapon destroy 2.6s melee weapon in mainhand. Gap was over 150dps. Dps was same with douple flametongue or windfury/flametongue.

I also tested TBC weapons: [] and [] was pretty much par. Melee weapon win when calculted extra melee stats.

Spell damage potential is quite big now.

Edit: That's bug only lower ft/ft dps. But Wf/ft with caster weapon was same amount dps boost. So caster weapon still look like a viable choice.

tukez · 09/08/08, 6:27 AM

Originally Posted by Pitbuller

Another discovery:
I tested new naxx weapons and 1.8s caster weapon destroy 2.6s melee weapon in mainhand. Gap was over 150dps. Dps was same with douple flametongue or windfury/flametongue.

I also tested TBC weapons: [] and [] was pretty much par. Melee weapon win when calculted extra melee stats.

Spell damage potential is quite big now.

Be advised that I just discovered MH Flametongue bug which causes it to proc on OH attacks, not MH(good old copy&paste). So if you have ft/ft and faster weapon in off hand it will result in more DPS than it supposed to give.

tukez · 09/08/08, 2:35 PM

0.7.7
-Stormstrike PPM statistic is now a accurate. It shows the PPM of casts.
-Fixed MH flametongue bug.
-Flametongue now gives 96 spellpower.
-Option to use searing totem approximation removed.

enhsim - Google Code

The reason I removed the Searing Totem approximation is, that it just wasn't accurate anymore. That is because of the overall increase in spell damage, after I made that approximation.

I tried different event systems in hope of making it faster, but seems like my old system is still faster than heaps etc. Maybe I'll try to make up something faster by modifying my system, because I am not satisfied with speed currently. On my computer it is about 2400000 times normal. Used to be a lot more...oh well, that happens when you add a bunch of new procs.

Levva · 09/09/08, 6:49 AM

Sim looks great. One quick question. Any reason that Reverberation talent is not included as an option to test its effects?

Pitbuller · 09/09/08, 9:03 AM

Originally Posted by Levva

Sim looks great. One quick question. Any reason that Reverberation talent is not included as an option to test its effects?

From the config file:

average_shocks_per_minute	 10.0	//for example, put to 10.0 to have shock cooldown of 6 seconds

You can try bit more different rotations with this setting.
Example Fs, SS, LvB was suggested before anyone tested it with sim. It's very bad low dps rotion but without simulator it looked good and easy rotation. If there was just Reverberation talent to calculate shock cooldown you couldn't set this rotation for sim.

Reverberation is about 0.5% dps per talent point.

Mana usage and how we can handle it:

With this build at 70lvl base rotation mana usage is around 13k. Totem droppin cost 39% of base mana this is 1044mana. We can expect 6k buffed mana pool at least.(Elixir of Major Agility + Elixir of Draenic Wisdom). If we expected get free mana bar every time shamanist rage is up this leave only (105/120 * 14000) - 6000 =~ 6000 mana negative. Jow is 4% of total mana, 100% proc and 4s Cd. If we get proc least one time per 5s this give us 5040mana. Now we are at negative 1000 mana. Mana spring give us 50mp5 * 21 = 1050mana. We can sustain basic rotation + totems infinite even without replenish effect or other outside mana sources.

Edit: Levva it's work just like you want. Just use reverse math.
0/5 = 120 / (6.0 + lag)
1/5 = 120 / (5.8 + lag)
2/5 = 120 / (5.6 + lag)
3/5 = 120 / (5.4 + lag)
4/5 = 120 / (5.2 + lag)
5/5 = 120 / (5.0 + lag)

Shock per minute don't actual tell how many shock sim use per minute. It's tell to sim what is max number per minute what is same as shock cd. If shock priority is set low then sim use far less shock than you have actual inputted.

Example with this config;

...
average_shocks_per_minute       10.00
...
stormstrike_cooldown            9.00
...
rotation_priority1              mw_lb
rotation_priority2              ss
rotation_priority3              fs
rotation_priority4              es
rotation_priority5              ls

Sim used only 9.9741 shock per minute not 10.

Malan · 09/09/08, 9:07 AM

So the new GUI front end is windows only?

Levva · 09/09/08, 9:19 AM

Originally Posted by Pitbuller

From the config file:

average_shocks_per_minute	 10.0	//for example, put to 10.0 to have shock cooldown of 6 seconds

You can try bit more different rotations with this setting.
Example Fs, SS, LvB was suggested before anyone tested it with sim. It's very bad low dps rotion but without simulator it looked good and easy rotation. If there was just Reverberation talent to calculate shock cooldown you couldn't set this rotation for sim.

Reverberation is about 0.5% dps per talent point.

Yes but it was also said that you change "average_shocks_per_minute" to reflect a deal of reality on how many shocks you can realistically get with latency, reaction times etc. ie: the AVERAGE not the IDEAL shocks per minute. Surely that is independent of the duration of the shock cooldown. I would have thought that the sim could easily include when a shock was off cooldown. At present without knowing the Reverberation value it cannot know the CD time period thus cannot simulate its impact.

I would think that it would make the sim easier to use to include a 5 pt Reverberation drop down and a warning if the "average_shocks_per_minute" was impossible given Reverberation. eg: 10.1 shocks per minute is impossible with 0/5 Reverberation. The max value with 5/5 Reverberation would be 12.0 shocks per minute.

My thought is that aside from any DPS increase it MIGHT be that offsetting the shock CD by 0.2 sec or 0.4 sec etc may open up extra GCDs or stagger timings sufficiently to prove useful. The only real way to tell is via a sim. If the sim doesn't support that you cannot tell.

tukez · 09/09/08, 9:58 AM

Originally Posted by Levva

Yes but it was also said that you change "average_shocks_per_minute" to reflect a deal of reality on how many shocks you can realistically get with latency, reaction times etc. ie: the AVERAGE not the IDEAL shocks per minute.

...

Maybe I should remove the average_shocks_per_minute and add shock_cooldown. If I implement lag in the sim I can remove the shock_cooldown entirely and add Reverberation. What do you think?

Meanwhile, just to clarify: average_shocks_per_minute sets the shock cooldown. As Pitbuller said, it does not mean that shock is used as soon as the cooldown is ready.

tukez · 09/09/08, 10:07 AM

Originally Posted by Malan

So the new GUI front end is windows only?

Yeah. VC++ is the only tool I know how to make GUIs.

• Raut · 09/09/08, 10:13 AM

Bleh. There's no Makefile in the source directory. Can we get that, please? Trying to figure out how to compile this on something else than Wintendo is hard enough.

Question about the source: Where does "Sleep()" come from? Closest hit I know of is sleep(3)

       #include <unistd.h>

       unsigned int sleep(unsigned int seconds);

I had some trouble installing ZThread. It didn't like my environment variables. Solved that by using "CPPFLAGS=-fpermissive" as an option to configure.

tukez · 09/09/08, 10:18 AM

Originally Posted by Raut

Bleh. There's no Makefile in the source directory. Can we get that, please? Trying to figure out how to compile this on something else than Wintendo is hard enough.

Question about the source: Where does "Sleep()" come from? Closest hit I know of is sleep(3)

       #include <unistd.h>

       unsigned int sleep(unsigned int seconds);

I had some trouble installing ZThread. It didn't like my environment variables. Solved that by using "CPPFLAGS=-fpermissive" as an option to configure.

I build with VC++, so I don't have a Makefile for you. Why don't you use some IDE too which does these things for you?

That Sleep() you can safely remove. It is some legacy stuff left from multithreading testing and I will remove it from the source also.

Remember if you can't get the ZThread working you can disable it by using NOTHREADS define.

• Raut · 09/09/08, 11:18 AM

g++ -pthread -I/usr/local/ZThread/include *.cpp /usr/local/ZThread/lib/libZThread.a -o enhsim

There. I had to check with my coworker since it's ages since I manually handled static libraries. (*sob* NerdRep--)

IDEs are really a matter of taste IMO. I don't see why I would have to get, install and learn one to be able to do the above line. If nothing else, please include it in your compile_instructions as the file is a bit sparse on how to actually compile your program.

tukez · 09/09/08, 11:27 AM

Originally Posted by Raut

g++ -pthread -I/usr/local/ZThread/include *.cpp /usr/local/ZThread/lib/libZThread.a -o enhsim

There. I had to check with my coworker since it's ages since I manually handled static libraries. (*sob* NerdRep--)

IDEs are really a matter of taste IMO. I don't see why I would have to get, install and learn one to be able to do the above line. If nothing else, please include it in your compile_instructions as the file is a bit sparse on how to actually compile your program.

You asked for a Makefile, but as you noticed, you don't need one. I don't know what is more simple than to compile *.cpp. Only problem is ZThreads, but it depends on the compiler how to link it in.

I'll include that line as example in the instructions.

Levva · 09/09/08, 11:57 AM

Originally Posted by tukez

Maybe I should remove the average_shocks_per_minute and add shock_cooldown. If I implement lag in the sim I can remove the shock_cooldown entirely and add Reverberation. What do you think?

Meanwhile, just to clarify: average_shocks_per_minute sets the shock cooldown. As Pitbuller said, it does not mean that shock is used as soon as the cooldown is ready.

My preference would be something that's as close to simulating how the game works in reality as possible. Its not a whole lot of use simulating absolutely perfect conditions (zero lag, instant reaction times). If the sim is able to cope with reaction times & lag as variables, then this would be wonderful as we can simulate how the effect of slow reactions vs faster reactions would improve your dps. ie: you get more skilled at reacting faster and your dps goes up by X%. Similarly its useful to know the effects of lag.

The absolutely ideal world scenario is useful I suspect only at the level of comparison with real world. If for example the sim showed that a 100 ms lag/reaction time favoured a different rotation/setup vs a 0ms lag/reaction time then that would be useful info to know.

It may show a tipping point. ie: at x ms lag you are better switching to ABC setup over XYZ as lag will cause XXXX to give lower dps... etc. Then again it may be a straight line less lag = more dps with one fixed setup/spec. Having the sim provide this info would be ideal.

tukez · 09/09/08, 12:48 PM

I implemented lag as random time added to cooldowns requiring input, also on global cooldown. Here are the results, you tell me is it worth to have. Quick test shows the speed decrease to be about 5%. Maybe 10%.

Tests were made with the default BestDpsBuild.txt.

Levva · 09/09/08, 12:58 PM

A very straight line. If you've already implemented the code I'd keep it there and have a default lag parameter of say 50ms or 100ms. There is always going to be some lag. If there is an option then people who want ideal, unachievable dps in their sims can set lag to zero. Those that prefer to see achievable real world values can keep it in. The issue is if blizzard make any changes where the lag becomes a factor if the sim can handle the lag now we can flag that up to them instantly.

I play with usually 150-250 ms lag as normal on my 8Mb ADSL connection, lower than that is very rare. I am aware that Oceanic players often have significantly higher lag.

If the sim can give values we can readily replicate in game it becomes more meaningful, and we can trust the output or even find specific bugs where the real values differ from the sim. If the sim gives values that we never seem to achieve (due to lag) it makes it harder to accept the sim is accurate and harder to notice the more obscure bugs.

tukez · 09/09/08, 1:47 PM

0.7.8
-Lag is taken into account. Use lag_min and lag_max to set the lag. It is in milliseconds.

enhsim - Google Code

tukez · 09/09/08, 2:28 PM

Might be a good idea to include threat in the simulation. You would input the TPS the tank does and then we can include Windshock in the rotation.

Pitbuller · 09/09/08, 2:54 PM

Originally Posted by tukez

Might be a good idea to include threat in the simulation. You would input the TPS the tank does and then we can include Windshock in the rotation.

Very good idea.
Threath might be big problem. I calcuted threath from dps what sim show and it was horrifed 2323. Might be hard find tank who do 2112tps at sunwell level. Tanks get TPS boost but I doubt that it that big.

Ps. I want bloodlust too.

Rouncer · 09/09/08, 2:59 PM

I was wondering if there were any plans to intergrate your sim into a Rawr model?

Using their interface to pull info from the Armory on all the items, as well as specifying buffs and debuffs involved, and then inputting it all into your simulator to run the actual numbers?

(sorry if this question is retarded from a technical side, I don't know anything about programming. My main used to be a mage and nothing compared to Rawr.mage for ease of use and I miss it.)

Originally Posted by Pitbuller

Very good idea.
Threath might be big problem. I calcuted threath from dps what sim show and it was horrifed 2323. Might be hard find tank who do 2112tps at sunwell level. Tanks get TPS boost but I doubt that it that big.

Ps. I want bloodlust too.

Tanks are currently running 3-4k tps in the Beta, from what people have posted. Don't forget that they are baking the Salv threat reduction into the tanking abilities.

tukez · 09/09/08, 3:23 PM

Originally Posted by Pitbuller

Very good idea.
Threath might be big problem. I calcuted threath from dps what sim show and it was horrifed 2323. Might be hard find tank who do 2112tps at sunwell level. Tanks get TPS boost but I doubt that it that big.

Ps. I want bloodlust too.

Ah..yes, I've forgotten about the bloodlust. It is a bit problematic though..I am not sure what to do with the cooldown exaclty. Should people be able to choose from 5 or 10min cd or should it have something to do with the combat length.

tukez · 09/09/08, 3:31 PM

Originally Posted by Rounced

I was wondering if there were any plans to intergrate your sim into a Rawr model?

Using their interface to pull info from the Armory on all the items, as well as specifying buffs and debuffs involved, and then inputting it all into your simulator to run the actual numbers?

(sorry if this question is retarded from a technical side, I don't know anything about programming. My main used to be a mage and nothing compared to Rawr.mage for ease of use and I miss it.)

Well, I think it might be possible to make a GUI in Rawr and then run enhsim.exe from Rawr, like I do it with my GUI. I don't know how to do it with Rawr, though. Rawr is C# which I'm not familiar with and I don't know is there something in Rawr which would make this impossible. I have not even tested Rawr so I don't know much about it.

Maybe someone wants to make better GUI for EnhSim!

Pitbuller · 09/09/08, 4:01 PM

Bloodlust is simple:

One or two shaman?

If one then uptime is (combat length / 10) and then round it up next integer and divide it by combat length - last bloodlust partial loss divided by combat length.

If two then uptime is (combat length / 5) and then round it up next integer and divide it by combat length - last bloodlust partial loss divided by combat length.

Then you can even stack bloodlust with other cooldowns.