+- SWTOR Mechanics Forums (http://mmo-mechanics.com/swtor/forums)
+-- Forum: General (/forum-1.html)
+--- Forum: Game Mechanics (/forum-5.html)
+--- Thread: Primer on calculating DPS stat weights (/thread-689.html)
RE: Primer on calculating DPS stat weights - RuQu - 03-03-2012 06:08 AM
(03-03-2012 05:08 AM)kray Wrote: That's where you're wrong. It is my task as someone who writes simulation to generate stat weights based off of DPS. I run thousands of sims with changes in various stats, and calculate their value off of that. In that case, the calculated DPS needs to absolutely be accurate and precise.
The issue with your graphic (awesome by the way!) is that the top end player -- the one who really cares about stat weights the most -- is going to be doing everything in his/her power to reach that bullseye. You also don't include (nor does the link I sent you) the fact that as you run more and more simulations and average the results, those dots for simulations get closer and closer to the bullseye.
But, simulation vs formulation is an argument as old as time. Fact is, I'm doing it this way and want to find out a way to improve this method!
RuQu, I like the idea. Dunno if I'm going to include it in mine though.
The question is, though, how close can you come to the "true" answer. And is there even such a thing? How close is close enough, considering the spread of ability even among good players?
I can't view your work, as I don't have the necessary software package (Maple) to run it.
I spent most of last night and today getting Gunnery added into my calculator project over at Google Code. (Link is here) I was hoping you'd give it some testing, see if there are any obvious bugs or major discrepancies with your results. Source code is up there as well.
*edit*Caught a small bug with FA, new version is uploading now. 1.0.9b1. The DPS number I get out is lower than I'd expect and than I see others reporting getting, but I'm not sure why. HS damage checks out right with in-game, as do the tech and ranged damage and crit values. I don't play Gunnery, though, so the rotation logic is just what was in the guide here and I can't tell if it is doing something unexpected.
RE: Primer on calculating DPS stat weights - Kaedis - 03-03-2012 06:29 AM
Ok, the Simulation versus Formulations debate has occurred ad nauseum at nearly every level of academia at some point. We don't need a reprise here.
Back to the original topic guys.
RE: Primer on calculating DPS stat weights - LagunaD - 03-03-2012 08:00 AM
(03-03-2012 03:14 AM)kray Wrote: Wanted some input really quick from you guys. Recently did a coding overhaul on my simulator, increased the speed per sim by 10-20x, and am now able to run stat weights properly.
Heh.
(03-03-2012 03:14 AM)kray Wrote: I know it's a bit hard to read, but essentially what I'm seeing in DEP (what I really care about) is a variation of +/- .03 or so, which is exacerbated when looking at the easier-to-read AEP format. In practice, this variation is absolutely minimal, but I'd like to do what I can to get rid of it at this point. Any recommendations, or is the answer "run more sims"?
A couple ideas:
1) Use the average damage for every attack that actually lands, rather than distributing the damage values between minimum and maximum. That is easy to do, should actually speed up the simulation (slightly) and will remove a *lot* of pure RNG noise which is not affected by any stat.
2) Use quasi-random instead of pseudo-random numbers. The Sobol' sequence is used sometimes to reduce the fluctuations in simulation. I am not sure whether it will degrade performance, but here is a graph of a Sobol' (quasi-random) sequence generator:
![[Image: Low_discrepancy_10000.png]](http://upload.wikimedia.org/wikipedia/commons/f/fc/Low_discrepancy_10000.png)
and here is the same graph for a pseudo-random sequence generator for the same number of points:
![[Image: Random_10000.png]](http://upload.wikimedia.org/wikipedia/commons/7/79/Random_10000.png)
Notice how the Sobol' sequence very uniformly samples the space, while the pseudo-random sequence has clumps and gaps.
Since #2 is more work, I'd definitely try #1 first.
RE: Primer on calculating DPS stat weights - kray - 03-04-2012 08:08 AM
Quote:I can't view your work, as I don't have the necessary software package (Maple) to run it.My sim runs in Excel, coded in VBA. Linky
The edits I refer to below haven't been posted online yet, still have a bit more work to do on other features.
Heh indeed. At about 1/100 of a second per sim now. I'm OK with that for now. Been cleaning up the code a lot as I go, picking up fractions of fractions of seconds here and there, it's always fun.
I appreciate the suggestions and the shove in the direction towards Sobol... got a lot of stuff to read through in more detail later on.
As far as option one goes, I'm not the biggest fan of it. Damage variation is a part of the game and is to be expected, and though I'm sure little to all variance would be removed, it rubs me the wrong way. I'm trying to mimic the in-game environment as closely as possible, call it my own paranoia.
Option two kind of follows along the same path I was thinking of -- leaning away from psuedo-random generators. VBA has an inherent issue with a lot of quasi-random algorithms though, from what I've read, due to it's handling of the Long datatype.
For now, since the EX-TA-REME speed increase, I'm going to continue to improve the simulator as is, and see how the community as a whole reacts to the long sim times. I've found that 1500-2000 sims per stat is the breakpoint to ensure accuracy when it comes to weeding out variance in stat weight calcs, which (altogether) takes around 2 minutes on my laptop, pointedly faster on my desktop.
I really appreciate the discussion so far everyone, a ton of really great information here.
RE: Primer on calculating DPS stat weights - LagunaD - 03-04-2012 09:58 AM
(03-04-2012 08:08 AM)kray Wrote: As far as option one goes, I'm not the biggest fan of it. Damage variation is a part of the game and is to be expected, and though I'm sure little to all variance would be removed, it rubs me the wrong way. I'm trying to mimic the in-game environment as closely as possible, call it my own paranoia.
With all due respect, that makes no sense at all.
If you were trying to model the variation of damage output on a fight-to-fight basis, then it would make sense. But you aren't (you are running many simulations for the express purpose of removing that variation), so it doesn't.
The average of a sum is equal to a sum of the averages. For a sufficiently large number of runs, the only way the two methods can give different results is if you have a bug in your model. You will just get the same answer much faster by using average damage.
RE: Primer on calculating DPS stat weights - kray - 03-04-2012 01:08 PM
(03-04-2012 09:58 AM)LagunaD Wrote:(03-04-2012 08:08 AM)kray Wrote: As far as option one goes, I'm not the biggest fan of it. Damage variation is a part of the game and is to be expected, and though I'm sure little to all variance would be removed, it rubs me the wrong way. I'm trying to mimic the in-game environment as closely as possible, call it my own paranoia.
With all due respect, that makes no sense at all.
If you were trying to model the variation of damage output on a fight-to-fight basis, then it would make sense. But you aren't (you are running many simulations for the express purpose of removing that variation), so it doesn't.
The average of a sum is equal to a sum of the averages. For a sufficiently large number of runs, the only way the two methods can give different results is if you have a bug in your model. You will just get the same answer much faster by using average damage.
No offense taken. One of the benefits of doing the theorycrafting this way, as you mentioned, is being able to determine things like variance... an incredibly important metric in my opinion. For instance, Pyrotech has a typically higher DPS at the cost of a relatively large variance, whereas Advanced Prototype has a lower DPS with a much smaller variance due to a much smaller reliance on procs. True, using averages would only eliminate some of the variance, but having a simulator that can reproduce in game variance to determine standard deviation, etc, is something I probably won't give up anytime soon. Running large sets of sims isn't the only function of my particular sim -- I don't want to pigeonhole myself. Better to build an all encompassing tool then chop it up from there than the other way around.
Realistically, it would make the most sense to have a switch be flipped while running large sets of simulations. An option the user can turn on/off would be ideal, for sure. I'll have to work on adding it in a future build =]
At the end of the day, I want to incorporate all of the things we discussed here, with the option given to the user to pick and choose how the sim runs.
RE: Primer on calculating DPS stat weights - West - 04-05-2012 02:47 PM
(02-03-2012 01:25 PM)Kor Wrote:Quote:My understanding is that bosses have 10% avoidance, necessitating 110% accuracy. Is this correct? Do we have any information on their damage reduction?
We're still unsure as to actual numbers, as calculating the precise accuracy cap requires statistical analysis of a large dataset, which is nearly impossible without an exportable combat log. However, some less rigorous testing has suggested the operations boss cap may be 8% rather than 10%.
Any news on this from the PTR? Just from general observation of my own flytext damage, with 100% accuracy I get 0 misses/dodges/deflects on my DPS Guardian/Jugg, and I'd love to see the actual numbers on this.
RE: Primer on calculating DPS stat weights - Kaedis - 04-08-2012 12:57 PM
Quote:Any news on this from the PTR? Just from general observation of my own flytext damage, with 100% accuracy I get 0 misses/dodges/deflects on my DPS Guardian/Jugg, and I'd love to see the actual numbers on this.
Recent evidence both from live and from the PTS is suggesting 10%. The flytext is apparently very unreliable for actually reporting misses (particularly in channeled abilities that can miss on a per-tick basis) when your miss chance is very low.
RE: Primer on calculating DPS stat weights - Titan - 04-30-2012 12:30 AM
Holy cow. I wish I could begin to understand the math of all thats been going on in this thread. Has anyone touched Guardian / Juggernaut DPS/stat weights?
RE: Primer on calculating DPS stat weights - Raganok - 05-16-2012 02:56 PM
Yeah haha I'm not going to try to understand this math T_T. But I was wondering and I've been searching the internet...I still don't know the numbers for a DPS Sith Assassin. All I got from asking people is I want to be at 75% surge rating, 35% crit, and 100% accuracy? Can anyone give me the facts on this because I don't know how important accuracy is for an assassin. I'm dpsing as Deception spec btw.