Inconsistency between gear chooser & simulation

[Kered]

I'm finding inconsistencies between the gear chooser's numbers and what I get in the simulator. 

To explain...the gear chooser shows Nighttime as being 87.5 dps over Singed Vis'kag and 90.2 dps over Empowered Deathbringer as an offhand for DW Frost.

However, when running the Sim 10 times and taking an average, I'm getting Deathbringer as the top performer and Vis'kag being 14 dps below that and Nighttime being 51.5 dps below.  Essentially a reversal of what the gear chooser would have me expect.

Using my own logic, I would have put Deathbringer at the top as well, but I put a lot of weight on weapon damage, of which Deathbringer wins out. 

I'm not sure if the gear chooser isn't taking procs into effect but the sim does or what, but it's definitely not helping me very much. Easy enough to check in game, but frustrating at the moment.

Any reason I'm not thinking of for this to be taking place?

[Devloc]

The DPS estimates are kinda static and are calculated beforehand using predetermined gear sets.  They're more or less guesses, but I've found them to be pretty accurate.  The thing that really messes them up is non-stat values, i.e. procs.  Those are tough correctly evaluate.  But you had the right idea, just run the sim and get some actual results.

Oh, and rather than run the sim 10 times and calculate the average yourself, hit the Settings button, and just crank the number of times to run by 10.  Save yourself some work =)  And on that topic, I've found 5k runs (at 600s each) has been a pretty sweet spot as far as stability in numbers goes, less than 7 +/- difference between runs.  I run that on a e8400 (dual core, oc'd to 3.8GHz) and it takes about 4 minutes (using both cores of course).

I wanted to bring that up, because if you're running 100 or 200 simulations, it would be expected to see a large variation in the resulting DPS, one run could bee 150dps or more different than the same exact run at those numbers.  But - you're running it 10 times and averaging them out... so... you probably already realized this.  GG.

[yellowfive]

Yeah items with procs can be difficult to estimate.  Each class or spec might have a significantly different up-time or number of procs... so the estimate code just takes a guess.

I plan to put some very obvious visual cues on the UI for items that have procs that may be estimated less accurately than other items, and suggest to the user that they run a simulation.


I usually leave my fight durations around 5 minutes (300 seconds), since most dps-burn style fights are in that range or even shorter, and then crank up the number of iterations to get more accuracy.  One of these days I'll get around to optimizing the simulator so that it runs faster... so many things to do!

[diablas]

codes do not guess all weapons pocks or ability ppm or other are all based off a specific algorithm theres no guess or randomness in mathematics

[Revulva]

what?

[jokeyrhyme]

I'd just like to say that I really like the gear rankings being static. Anyone who's ever used Rawr will remember the wait times after every single little change. Very frustrating. If MrRobot does end up having dynamic gear rankings, I would love it if there was a "refresh rankings" button, or a setting to keep it static. I really don't care for 5-10 seconds of interface lag after every little modification.

[yellowfive]

The gear rankings actually are slightly dynamic... they will change with your character's gear/stats, if your overall spec changes (e.g. go from unholy to blood), or if you change a gem/enchant.

Currently it gear rankings not update when you change individual talents or glyphs.  This was the trade-off that we made to keep the tool more responsive.  I'd like to add glyphs to the gear rankings sometime soon, since I think those could be handled without any noticeable delay.

We figured that while changing your talents/glyphs will most definitely affect your DPS, it is not all that common that it will affect the ranking of gear.  That's why we only update the rankings when a significant change is made, e.g. from one style of play to another.