Java App Modelling Rogue DPS (Updated for 3.3.3) - Page 2

Ottosen · 01/04/10, 3:34 PM

Well made program, really liking it so far. Haven't tried the latest version yet, but the previous version didn't quite match up to my in-game stats, even though I considered the lack of professions etc. I'll double check.

Also, when I'm using it, I can't get tooltips over items at all. Any idea? *It started working after I updated my java version..not sure that's what made it work, but yeah.

Now the real question is, which is the better guideline now - rawr, Aldriana's sheets or this

Can't wait to hear the debate on this later on.

trunicated · 01/04/10, 4:17 PM

Originally Posted by Ottosen

Well made program, really liking it so far. Haven't tried the latest version yet, but the previous version didn't quite match up to my in-game stats, even though I considered the lack of professions etc. I'll double check.

Also, when I'm using it, I can't get tooltips over items at all. Any idea? *It started working after I updated my java version..not sure that's what made it work, but yeah.

Now the real question is, which is the better guideline now - rawr, Aldriana's sheets or this

Can't wait to hear the debate on this later on.

Realize that this is just a model. It's mathematical equations that are being used to determine the EP numbers and DPS (which is why I'm so curious about the source, so I can comb through it a little.) As such, it's going to be a great way to get a quick idea about what you should be able to pull, what gear might be a good upgrade, and a rough estimate of what your EP values are going to look like.

However, it will never be as accurate as a spreadsheet that runs actual simulations with random numbers. Especially if you're trying your hardest to minmax on every stat. I just know that Aldriana's sheet, with it's in-depth rotation setup, specific talent setup, and buff control (not to mention tons of time behind it) will be better for those that need to squeeze every last point of damage out of their character. However, not all of us like running spreadsheets for hours on end for a few gem slots or one piece of gear. :-)

I expect to see this catch on as a good way to get a good idea of what you want, and Aldriana's to be the absolute solution.

robfang · 01/04/10, 5:59 PM

To be exact, Aldriana's spreadsheet does not do simulations. If modelled correctly, this program will be the same as Aldriana's spreadsheet with better user interface perhaps.

You should be referring to Mavanas's simulation spreadsheet if you want some combat simulation.

By the way, Chack, the 2nd version of the tool calculates lower dps values than the first version. Also, I failed to see any effect of the professions like engineering or leatherworking.

Chack · 01/04/10, 8:00 PM

Aldriana's spreadsheet is indeed modelling as well. Generally modelling has the advantage that it's faster and easier to see if an item is "better". Simulation will give you more realistic dps end results, but it will take alot of data to figure out if 10 haste is better then 10 agi. Also there is no guarantee that the simulated results will actually tell you the truth.

@robfang
- The first version's starting attributes were 10 ap too high. In addition in the current version there is a bug where 10 stats on chest only gives 10 agi, so you will probably have 20 ap less then compared to v0.5.
- To make use of professions like engineering and leatherworking, you also need to equip those enchants

• Aldriana · 01/04/10, 8:08 PM

Realistically, something like this would be a direct replacement for my spreadsheet if done well. As would Rawr. Some of you may recall I've experimented with such things in the past - I simply haven't had time to reach any reasonable level of sophistication with it, and I have to commend Chack for the work he's done so far. On the other hand, I feel like there's a fair number of features missing so far, and the fact that there seem to be some EP discrepancies makes me suspicious of the underlying model - I can't 100% swear that I'm wrong and this tool is wrong, but given the relative age - and thus the amount of use - of the tools in question, I'm inclined to guess that the more established tools are more likely to be correct when there's agreement - at least until I see evidence or an explanation as to why the new tool has "better" modeling that is causing the difference.

So yeah. Looks cool. Hopefully you can do some good things with it and it will become a valuable part of the rogue's theorycrafting arsenal. I'm just not sure it's quite there yet.

Shoujahitsumetsu · 01/05/10, 6:41 AM

Originally Posted by trunicated

Realize that this is just a model. It's mathematical equations that are being used to determine the EP numbers and DPS (which is why I'm so curious about the source, so I can comb through it a little.) As such, it's going to be a great way to get a quick idea about what you should be able to pull, what gear might be a good upgrade, and a rough estimate of what your EP values are going to look like.

The source code is available in the release itself. You can use an archiver like winrar to break down the .jar file into various .class files. Those class files have all of the source code, however they're arranged into bytecode making them much harder to read. Just use a java decompiler, plenty of freeware ones on google, and you can convert the .class files into their original .java files. From there, reading them is quite simple; just use a text editor like textpad, notepad, etc.

Tomate · 01/05/10, 7:22 AM

Originally Posted by Chack

I am not sure what you mean with faster tooltips. If you want them to pop up faster when you mouser-over an item, I'm afraid I can't change that.

Yes, it was what I mean.
Concerning the other points in my first post can them be done?

I have noticed that stats of the baelgun heavy crossbow which drop in the chest after anubarak in 10 player are wrong.
It should be the same as the horde version.

Guradia · 01/05/10, 8:53 AM

*raisesHand* as someone extremly interessted in contributing.
I'm in no way professional with Java, but at least a professional programmer ^^
Would love to see you providing the sources.

Rawr ist a fine and versatile program, but it lacks cross-platform capabilities (eauqal to VBA extensive speadsheets).

Chack · 01/05/10, 9:56 AM

Originally I did not plan to release the source, but I will give it another thought.

@Aldriana
When I have more time I can compare your formulas with mine and compile the differences between the underlying models. This is going to be a very time consuming process tho. One thing I noticed at least, is that the EP for hit and exp stay above 0 even when going past the caps. Maybe we have just different methods of calculating the EP values.

@Tomate
They can certainly be done, but not all of them are top priority at this time. I just added colored highlighting for hit/crit/arp caps tho.

I put up a new Version with a couple of fixes and missing items.

iDPSv0.7-b12.jar

redpilul · 01/05/10, 10:29 AM

Thanks for this very useful app.
It is true that it may not be as precise as the spreadsheet atm. But I do think that it has the potential to replace it in the future.
My java skills are far from those of a pro so I won't offer you my help with that part, but the idea of releasing the source code could be interesting for me to translate it (in french for example since it is my first language).
Keep up the good work

Lightshadow · 01/05/10, 11:07 AM

Neat tool. I was curious about the EP values I was getting back though, so I tried plugging my gear into this v0.7, Aldriana's sheet, and Manavas' sheet as a test. Everywhere I could I tried to keep settings exactly the same, though I'm not entirely sure what the 3rd glyph you're using here is. (I also swapped my belt for vengeful noose since I'll be picking that up today and mine wasn't supported here, and yes I let Manavas' sheet run overnight).

 
        Here  Aldriana  Manavas  
Agi     2.05    2.00      2.08
Crit    1.88    1.83      1.84
Haste   1.84    1.94      2.03
Hit     1.83    1.99      2.05
Exp     1.83    2.29      2.43
ArP     2.21    2.03      2.12

Given the circumstances, I'd expect a fair bit of variation in the results, and some things seem to match up pretty nicely. Without knowing anything about the methodology behind this it's hard to say whether the low values for haste and hit are expected or not, but the value of expertise is rather shockingly low and armor pen seems a bit high.

Keep up the good work.

Chack · 01/05/10, 12:25 PM

Hmm that seems to suggest that my armor formula is off. Are you using ARP proc trinkets? Or in the other spradsheets you are getting alot more damage from white mele and poison.

private float calcArmorMod(float arp) {
		float armorC = (467.5F * 83 - 22167.5F);
		int armorDefault = 10643;
		float armor = (armorDefault * 0.8F * 0.95F);
		float armorRemovable = Math.min(((armorC+armorDefault)/3F),armor);
		armor = armor - (armorRemovable * Math.min(arp/1400, 1));
		float dr = armor / (armor + (467.5F * 80 - 22167.5F));
		return (1-dr);
}

About the huge expertise difference, did you spec full weapon expertise in the other spreadsheets?

PS: Combat is using SS, KS and ToTT Glyphs.

robfang · 01/05/10, 12:32 PM

For the muti spec, the biggest difference I have observed with respect to the other sheets is the value of 4 set bonus. Maybe it is because I am experimenting with the BiS items but removing 4 set bonus causes a huge dps loss (270 dps) when using 51/13/7 spec.

Edit: To be exact, after meddling with the other items, I could reduce the dps loss to 224 dps when using 2 set bonus instead of 4.

Neckface · 01/05/10, 12:34 PM

Originally Posted by Drakulina

Well, good luck with your application.
But I think it would be much easier for you to fix rogue module in Rawr.
Rawr is an open source project - so you can download it's source code and make it work correct. By the way the latest version of Rawr working rather well, it's dps for my rogue is only 700dps lower than in simulator, but there are some strange things, like using bloodlust\heroism doesn't affect my damage at all о_О
Rawr developers are looking for people who want and can help

It's written on C# it's very close to Java, you will easily understand it.
There are lots of features in Rawr that are already implemented and working so you will save yourself a lots of time.

I tried to edit rogue module in Rawr, and well it's not hard, but the problem is that i don't actually know ALL formulas
for calculating rogue dps... and it takes too much time to find right formula.

I'm pretty happy to have another option for the Mac, actually.

Good work.

Has anybody made comparisons to Aldriana's spreadsheet(s) to see how close the numbers (overall dps, not just EP values) are?

Chack · 01/05/10, 2:36 PM

4 Set Bonus is getting a high score because I haven't properly modelled envenom clipping. Basically my model runs with "perfect" pooling so you can get up to 90% uptime on envenom in 4pT10. I could go the other extreme and cut off envenoms as soon as the energy cost for that cycle is regenerated, which is basically zero pooling. That way you get very little additional envenom uptime from 4pT10.