[Enhancement] Rawr.Enhance thread - Page 4

Levva · 03/05/09, 10:53 AM

Originally Posted by Santorayo

I've tried the newest version (2.2.0.3) from the beta and I'm scared. The relative stat values are so far off from my enhSim EP that I'm afraid that I'm using enhSim wrong all the time ^^ Especially haste is valued much lower in Rawr.

Is that v2.2.0b3 beta release version?

What you have to realise is that the Sim is tested and fairly reliable the Rawr.Enhance model is still under active development. You also MUST ensure you have every buff identical between the two if you are to do accurate side by side comparisons. So far I have not got to the stage of being able to do side by side comparisons as I am too bogged down in making sure I've implemented everything. Only after I'm sure all the bases are covered can I start looking to see that the calculations are as accurate as is possible in a closed form solution.

I've done a huge amount of updates since beta 3 release. In particular I have noted that I had not included any options for buffs involving hit, crit, haste. These have been added in the latest SVN release. If you want to test this out visit the thread I am maintaining on MMO Champion Shaman forum - *Some volunteer testers for Rawr.Enhance required* There you will find downloads of the latest dlls. Every time I have a new stable build I post the latest dll there so testers can test out the latest changes.

I would appreciate even more testers especially the really knowledgeable guys who can assist with the details of the individual calculations. The state of the model at present is such that its moving towards the stage where I can concentrate less on checking if every option has been covered ie: sanity checking and concentrate more on whether every last detail of a calculation is included. For this I need more knowledgeable testing.

Changes since beta 3

v2.2.0.32115 - Added support for lots of missing buffs (especially Misery/Imp.Faerie Fire for 3% hit buff & Draenei 1% hit buff)

v2.2.0.30266 - Fixed various Flametongue weapon checks so you only get FT weapon dmg if you have FT weapon imbue enabled

v2.2.0.29329 - Implemented Flametongue weapon bonus from Elemental Weapons, Implemented Bloodlust haste bonus, Lots of tidy up code + Ensured that all stats used in calcs are included in "Relevant Stats",
Added minRange to Int after discussion with Astrylian about stepping ranges, this should fix the issue with Int suddenly being valuable then with a small buff change not being valuable in the Relevant Stats Values graph.
Fixed ArP buffs not implemented in model, fixed expertise to no longer be stepped.

OnosKT · 03/05/09, 11:30 AM

Maybe I am missing something, I have looked over all the Rawr UI and can't figure out where I am supposed to find the relative stat values in order to test that.

Levva · 03/05/09, 12:36 PM

On the graph window at the top it says Chart : Gear > Head. Click here to change it to Relative Stats Values. ie: you are changing which chart to display.

One issue at present I am having discussions with the co-ordinator about is the base stat of relative stat values. At present he is treating them like pure dps increases and there is no single stat at 1.00 as a base stat. I am arguing that this makes it harder to use the RSV for their normal purpose, he is arguing that seeing that Agi adds 1.32 dps is better.

I am not convinced that RSV's can really be viewed in that light as they don't take account of removing one item and re-equipping another. We shall see where the debate ends up. I have already coded a change that would rebase things around AP = 1.00 and await his approval or otherwise to allow the change.

What is the views of this community on the usefulness or otherwise of AP=1.00 as opposed to AP=whatever dps it adds?

hozzer · 03/05/09, 2:43 PM

I have been of the mind for some time that AP or AEP or EP (whatever the preferred terminology is) should be expressed as something other than AP, probably DPS. When the DPS value of AP changes, it can only be viewed in terms of the decrease or increase in the values of other stats using the current system.

Malan · 03/05/09, 3:46 PM

The EP system as we currently use it is an abstraction. AP = 1 is just an abstraction, the 1 is an arbitrary value. So the guy you're in discussion with is probably right.

The sim ultimately produces both so it's not like the sim needs to change. For the Rawr audience, probably better to just provide the DPS equiv.

Levva · 03/06/09, 6:38 AM

I think we have settled on a compromise. The default will be to show raw values with a tick box option on the model to show AEP values.

Dorrinal · 03/06/09, 1:57 PM

Good compromise. Coming from the Mage model to look at upgrades for my Shaman, I would expect DPS values instead of AEP values. I imagine it's the same for most Rawr users.

Levva · 03/23/09, 6:06 AM

New Rawr beta release available V2.20b5 - this includes the 3.1 talent trees and a feature I wrote to download possible item upgrades from Wowhead.com's ptr site.

Rawr - Release: Rawr 2.2.0b5

Please test out the Enhancement Shaman model and let me know of any issues.

Levva · 03/23/09, 12:59 PM

Plain melee dps calcs.

I am thinking that there is a factor missing in the melee dps calcs but not entirely sure I've got the right idea. The current code is :

float APDPS = attackPower / 14f;
float adjustedMHDPS = wdpsMH + APDPS;
float adjustedOHDPS = (wdpsOH + APDPS) * .5f;

float dpsMHMeleeCrits = adjustedMHDPS * chanceWhiteCrit * critMultiplierMelee;
float dpsMHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

float dpsOHMeleeCrits = adjustedOHDPS * chanceWhiteCrit * critMultiplierMelee;
float dpsOHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

float dpsMelee = (adjustedMHDPS + dpsMHMeleeCrits + dpsMHMeleeGlances + adjustedOHDPS + dpsOHMeleeCrits + dpsOHMeleeGlances)
* weaponMastery * (1 - damageReduction) * (1 - chanceWhiteMiss) * (1 + bonusPhysicalDamage);

Where dpsMelee is the total melee dps. The thing that strikes me is that wdpsMH & wdpsOH are just the dps values from the tooltip of the weapons. I fear that the effects of haste seems to be entirely missing from this calculation. However on adding it in it seems to vastly over estimate the melee dps.

Perhaps I'm just tired today but I'm sure I'm missing something really obvious.

tukez · 03/24/09, 9:40 AM

Originally Posted by Levva

Plain melee dps calcs.

I am thinking that there is a factor missing in the melee dps calcs but not entirely sure I've got the right idea. The current code is :

float APDPS = attackPower / 14f;
float adjustedMHDPS = wdpsMH + APDPS;
float adjustedOHDPS = (wdpsOH + APDPS) * .5f;

float dpsMHMeleeCrits = adjustedMHDPS * chanceWhiteCrit * critMultiplierMelee;
float dpsMHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

float dpsOHMeleeCrits = adjustedOHDPS * chanceWhiteCrit * critMultiplierMelee;
float dpsOHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

float dpsMelee = (adjustedMHDPS + dpsMHMeleeCrits + dpsMHMeleeGlances + adjustedOHDPS + dpsOHMeleeCrits + dpsOHMeleeGlances)
* weaponMastery * (1 - damageReduction) * (1 - chanceWhiteMiss) * (1 + bonusPhysicalDamage);

Where dpsMelee is the total melee dps. The thing that strikes me is that wdpsMH & wdpsOH are just the dps values from the tooltip of the weapons. I fear that the effects of haste seems to be entirely missing from this calculation. However on adding it in it seems to vastly over estimate the melee dps.

Perhaps I'm just tired today but I'm sure I'm missing something really obvious.

I think you need to reduce the adjustedMHDPS and adjustedOHDPS by the amount which is put to crits and glances.

float adjustedMHDPS = adjustedMHDPS * (1 - chanceWhiteCrit - glancingRate);

And the haste you mentioned needs to be added to the final calculation.

Levva · 03/24/09, 12:46 PM

Many thanks, I knew I needed the haste in there somewhere. However when I was trying, it just was massively screwing the dps calcs. I'd not noticed of course that I'd effectively double counted the MHDPS/OHDPS figures. I've modified the code as follows to separate out normal from crits & glances.

            float APDPS = attackPower / 14f;
            float adjustedMHDPS = (wdpsMH + APDPS);
            float adjustedOHDPS = (wdpsOH + APDPS) * .5f;

            float dpsMHMeleeNormal = adjustedMHDPS * (1 - chanceWhiteCrit - glancingRate);
            float dpsMHMeleeCrits = adjustedMHDPS * chanceWhiteCrit * critMultiplierMelee;
            float dpsMHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

            float dpsOHMeleeNormal = adjustedOHDPS * (1 - chanceWhiteCrit - glancingRate);
            float dpsOHMeleeCrits = adjustedOHDPS * chanceWhiteCrit * critMultiplierMelee;
            float dpsOHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

            float meleeMultipliers = weaponMastery * (1 - damageReduction) * (1 - chanceWhiteMiss) * (1 + bonusPhysicalDamage);

            float dpsMelee = ((dpsMHMeleeNormal + dpsMHMeleeCrits + dpsMHMeleeGlances) * hastedMHSpeed / unhastedMHSpeed +
                              (dpsOHMeleeNormal + dpsOHMeleeCrits + dpsOHMeleeGlances) * hastedOHSpeed / unhastedOHSpeed) * meleeMultipliers;

BTW have you an idea of timescale for a 3.1 EnhSim? The Rawr model has the haste, expertise (UR), flametongue & Armour Pen changes, along with the Feral Spirit glyph so its harder to check my model is working properly when I can't test it against your more accurate sim.

Oh and did you see the suggestion for allowing EnhSimGUI to import config from clipboard? This would allow direct import of Rawr.Enhance & ShockAndAwe exports and thus make the whole process a lot quicker for the end user. This is mainly due to removing the need for them to have to cut n' paste in a text editor. So the end result would be a lot more user friendly.

If you are happy to implement that and you need me to make specific tags on the import to verify clipboard contains valid data just let me know and I can standardise the header across both modules.

tukez · 03/24/09, 9:04 PM

Originally Posted by Levva

Many thanks, I knew I needed the haste in there somewhere. However when I was trying, it just was massively screwing the dps calcs. I'd not noticed of course that I'd effectively double counted the MHDPS/OHDPS figures. I've modified the code as follows to separate out normal from crits & glances.

            float APDPS = attackPower / 14f;
            float adjustedMHDPS = (wdpsMH + APDPS);
            float adjustedOHDPS = (wdpsOH + APDPS) * .5f;

            float dpsMHMeleeNormal = adjustedMHDPS * (1 - chanceWhiteCrit - glancingRate);
            float dpsMHMeleeCrits = adjustedMHDPS * chanceWhiteCrit * critMultiplierMelee;
            float dpsMHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

            float dpsOHMeleeNormal = adjustedOHDPS * (1 - chanceWhiteCrit - glancingRate);
            float dpsOHMeleeCrits = adjustedOHDPS * chanceWhiteCrit * critMultiplierMelee;
            float dpsOHMeleeGlances = adjustedMHDPS * glancingRate * .35f;

            float meleeMultipliers = weaponMastery * (1 - damageReduction) * (1 - chanceWhiteMiss) * (1 + bonusPhysicalDamage);

            float dpsMelee = ((dpsMHMeleeNormal + dpsMHMeleeCrits + dpsMHMeleeGlances) * hastedMHSpeed / unhastedMHSpeed +
                              (dpsOHMeleeNormal + dpsOHMeleeCrits + dpsOHMeleeGlances) * hastedOHSpeed / unhastedOHSpeed) * meleeMultipliers;

BTW have you an idea of timescale for a 3.1 EnhSim? The Rawr model has the haste, expertise (UR), flametongue & Armour Pen changes, along with the Feral Spirit glyph so its harder to check my model is working properly when I can't test it against your more accurate sim.

Oh and did you see the suggestion for allowing EnhSimGUI to import config from clipboard? This would allow direct import of Rawr.Enhance & ShockAndAwe exports and thus make the whole process a lot quicker for the end user. This is mainly due to removing the need for them to have to cut n' paste in a text editor. So the end result would be a lot more user friendly.

If you are happy to implement that and you need me to make specific tags on the import to verify clipboard contains valid data just let me know and I can standardise the header across both modules.

You have haste calculation the wrong way. Should be unhastedMHSpeed / hastedMHSpeed.

I changed haste(30%) and armorpen(25%), are they correct? What about expertise, I read somewhere you would get more expertise too?

I implemented that copy from clipboard functionality. Just use the same format as with config files. EnhSimGui checks does the text start with "race" and is the character after "race" whitespace character. I also did the item ID thingy.

Have you thought about doing the whole GUI with Rawr? I guess you can't get the EnhSimGUI implemented in the official Rawr, but Rawr is open source right?

Rouncer · 03/24/09, 9:19 PM

The 30% to haste is only to melee so that boost won't affect the GCDs of spells (haste never affects GCDs of melee attacks anyway) or casting times with say MW4 active.

Expertise being boosted was a bug. We just have the additional expertise from the new version of Unleashed Rage.

Armor Penetration Rating is 25% more effective.

Levva · 03/24/09, 9:46 PM

Originally Posted by tukez

You have haste calculation the wrong way. Should be unhastedMHSpeed / hastedMHSpeed.

I changed haste(30%) and armorpen(25%), are they correct? What about expertise, I read somewhere you would get more expertise too?

I implemented that copy from clipboard functionality. Just use the same format as with config files. EnhSimGui checks does the text start with "race" and is the character after "race" whitespace character. I also did the item ID thingy.

Have you thought about doing the whole GUI with Rawr? I guess you can't get the EnhSimGUI implemented in the official Rawr, but Rawr is open source right?

Haste is +30% effect of haste rating, armorpen +25% of armorpen rating, expertise is only 3/6/9 from Unleashed Rage, the other effect was a bug, with UR being 4/7/10%.

Other changes flurry is 5/10/15/20/25% down from 10/15/20/25/30% haste, the two new set bonuses ofc, Stormstrike is now 8 sec base and 4 charges, Imp.SS is reworked to return mana and has no affect on CD or charges anymore. Glyph of Feral Spirirt gives spirits +30% of your AP as a bonus.

The sticky at MMO has the full list so you can check everything off it as you implement it. *3.1 PTR Shaman Patch Notes (Updated 3/19/09)*

On the clipboard copy : both exports have a 4 line comment header lots of *'s with 4th line always blank 5th line starts race. I'll verify its exactly the 5th line in both cases and amend if not.

The whole GUI in Rawr is probably tricky for a number of reasons but the biggest one I think is that we would have no control over when a new version was released. I suspect it's better to deal with exporting as much as possible that's basic to Rawr and the EnhSim specific stuff stay in EnhSimGUI.

The other issue is that the project has limited authors with svn access and I'm not sure exactly what the licencing terms are, its a colabarative effort but I feel morally it would be wrong to try to hijak the project. Personally I feel it would be better that people who only want to use the sim can and don't have to use Rawr. People who like to use Rawr for its excellent kit exchange features can use a copy to/from clipboard model and run the sim in an app totally under your control.

One issue I did want to raise is buffs. People have asked if Rawr could export the buffs ticked. This would be a simple matter to export extra lines to the clipboard config data. If you would be happy to have buffs specified in the clipboard config I can add them. Would putting them at the end be acceptable? ie: anything after talents that Rawr exported would just replace anything in the current config? Or would you prefer we agree a specific ordering. If the latter simply tell me what works simplest for you and I can implement that.

tukez · 03/25/09, 5:51 AM

Originally Posted by Levva

Haste is +30% effect of haste rating, armorpen +25% of armorpen rating, expertise is only 3/6/9 from Unleashed Rage, the other effect was a bug, with UR being 4/7/10%.

Other changes flurry is 5/10/15/20/25% down from 10/15/20/25/30% haste, the two new set bonuses ofc, Stormstrike is now 8 sec base and 4 charges, Imp.SS is reworked to return mana and has no affect on CD or charges anymore. Glyph of Feral Spirirt gives spirits +30% of your AP as a bonus.

The sticky at MMO has the full list so you can check everything off it as you implement it. *3.1 PTR Shaman Patch Notes (Updated 3/19/09)*

On the clipboard copy : both exports have a 4 line comment header lots of *'s with 4th line always blank 5th line starts race. I'll verify its exactly the 5th line in both cases and amend if not.

The whole GUI in Rawr is probably tricky for a number of reasons but the biggest one I think is that we would have no control over when a new version was released. I suspect it's better to deal with exporting as much as possible that's basic to Rawr and the EnhSim specific stuff stay in EnhSimGUI.

The other issue is that the project has limited authors with svn access and I'm not sure exactly what the licencing terms are, its a colabarative effort but I feel morally it would be wrong to try to hijak the project. Personally I feel it would be better that people who only want to use the sim can and don't have to use Rawr. People who like to use Rawr for its excellent kit exchange features can use a copy to/from clipboard model and run the sim in an app totally under your control.

One issue I did want to raise is buffs. People have asked if Rawr could export the buffs ticked. This would be a simple matter to export extra lines to the clipboard config data. If you would be happy to have buffs specified in the clipboard config I can add them. Would putting them at the end be acceptable? ie: anything after talents that Rawr exported would just replace anything in the current config? Or would you prefer we agree a specific ordering. If the latter simply tell me what works simplest for you and I can implement that.

Alright, clipboard import will read config file format, I won't add any specific checks. I made some error checking, so Gui shouldn't be so easy to crash when bad data is inputted. I'll add that buff import too, the order of options doesn't matter for the import (or config files). New config options in the next version are "melee_haste", "spell_haste", "set_bonus1", "set_bonus2" and "set_bonus3". The old "haste" will be removed.

About Rawr, I meant you would take the Rawr source, add EnhSimGui functionality, possibly strip it out of unnecessary stuff and build/release it on your own. Rawr is licensed under Apache license, so you are definately allowed to do that. IMO, if reusing the code is a problem or creates some hard feelings, it shouldn't be licensed as open source.