Rivkah

It turns out the problem Cranch was reporting was related to the target armor change I made a few days ago. Turns out the target armor is in two different places in the Excel spreadsheet and the original setting I had was the correct one (as opposed to the change I made on 10/23). I've fixed the target armor back to 10643, if you have a saved setting with the other armor value just change it under settings and resave it with the new value.

 

Cranch

delete

 

Rivkah

I added a new feature today to the site which allows you to revert your settings back to your last "Update DPS". The button to revert will appear if you've made any changes to your settings since the last time you hit Update DPS and will revert all your form settings to their values at that time.

 

Lahiri

minor thing first:
Currently the "Deadly Ametrine" is listed as "Deadly Flawless Ametrine"

second

Wouldn't it be possible to instead get a result by using something like:
(estimated uptime*dps during proc) + (estimated downtime*dps without proc)

it should be more accurate than just averaging out the ArP gain.

 

Rivkah

I've fixed the ametrines, thanks for letting me know.

With regards to your other suggestion, the way the dps is calculated wouldn't allow for doing that without doing some pretty severe reworking of how everything is calculated. Right now all dps calculations are based on the stats that are calculated. In order to do what you're suggesting, there'd need to be two different ArPen calculations and they'd need to be applied to every shot twice. It's possible to do but it's a lot of reworking so it's not something I'm likely to do soon. If Shandara's spreadsheet changes the way it's handled I'll go back and make changes to match, but it's not a small endeavor.

 

boli

A possibly better approximation of the real thing, but still reasonably easy to do, might be: removing the trinkets altogether, adding the passive trinket stats as custom values and calculating four sets of dps

• No proc (only passive stats): dps0
• Trinket A up (passive and proc A stats): dpsA
• Trinket B up (passive and proc B stats): dpsB
• Both trinkets up (passive and proc A and B stats): dpsAB

These four dps are summed up weighted with the appropriate uptimes.

The weights I use are as follows, where pA is the uptime of trinket A, pB is the uptime of trinket B, which I usually take from Shandara's spreadsheet (Calculation Tab):

• "Uptime" of no trinket: 100% + pA*pB - pA - pB
• Uptime of trinket A only: pA - pA*pB
• Uptime of trinket B only: pB - pA*pB
• Uptime of trinket A and B simultaneously: pA*pB

So the final weighted dps is:

dps0*(100% + pA*pB - pA - pB) + dpsA*(pA - pA*pB) + dpsB*(pB - pA*pB) + dpsAB*pA*pB

Doing this instead of using the spreadsheet numbers I found that certain trinkets seem to be undervalued in the spreadsheet as well as the dps analyzer. As an example, based on my current gear, [Dark Matter] is shown as a 3 dps upgrade to [Darkmoon Card: Greatness] (agility version of course) in the gear planner, whereas doing the calculations above results in a 39 dps upgrade. Unsurprisingly for armor penetration trinkets like [Mjolnir Runestone] the difference is even more pronounced due to the way dps scales with pen.

As I've come up with this late at night when I wasn't totally sober, please let me know if you think I'm way off.

 

bronnum

@boli's post #156

At first look this might seem plausible and perhaps useful to determine choice of trinkets. However, I believe the calculations are prone to over-simplification:

Passive contributions are fairly easy to model and (IMO) holds in your equations. But the simultaneous contribution from i.e. non-linear stats (armPen) will be inaccurate calculated with straight multiplication. Further, as we know, every shot do not benefit from armPen, while others will hit with huge values - say crit/armPen proc and using Aimed Shot during these procs.

I guess these issues boils down to Shandara's "problems" with accurate modeling of procs, it's simply very RNG-related and requires dynamic update of stats to provide the right numbers. I still acknowledge your way of modeling as "second opinion" to determine trinkets, but I wouldn't recommend this to Shandara nor Rivkah.

Looking forward to a sober reply

 

Shandara

The main problem is that the spreadsheet and the online version are not 'true' simulations. To calculate the effect like proposed would mean calculating the dps 4 times, remembering the results and presenting the eventual result.

To make the sheet work predictably, there are built in a lot of assumptions and we don't use RNG either in the rotation simulation (otherwise the numbers would change each time).

For procs that are irrespective of other procs this is no problem and leads to a fair approximation. But for the ArmPen procs this unfortunately does not hold.

Only a true simulation (like SimCraft) can determine the real worth of these items (either alone or in combination) by running many 1000s of trials and computing the average DPS. Doing the same in Excel (or online) would be practically impossible, neither platform really lends itself to it.

WotLK Hunter DPS Spreadsheet

 

boli

I'm aware that a 'true' simulation should give the most accurate results. What I posted above was meant to be a simple way to get a better approximation for trinket dps.

I could of course be wrong in the assumption that it would in fact be better than the currently used averaging, which I understand works as follows (please correct if I'm wrong):

I'm using Mjonir Runestone as an example, which adds 665 arp on proc and has (roughly) 20% uptime. What I think the averaging does, is adding 133 arp (=20% * 665 arp) to the arp from gear/food; then the dps is calculated like that.

What Lahiri suggested (for a single single trinket) would be calculating with no additional arp (no proc) and weigh this number with 80%, then calculate the dps again with 665 additional arp and weigh this with 20%. Due to the arp scaling this should give a considerably higher number, assuming one is close enough to the arp cap with the proc.

I simply extended this idea to two trinkets, since some trinket combinations have nice synergies.

 

Blazeflack

First of all, thanks for this great tool.

I have a few minor problems though.
1.: It seems like it uses rare gems when calculating gem values for pawn instead of epic gems.
2.: For some reason some of the values in the pawn string differs a lot from the pawn string in the xls spreadsheet.
From spreadsheet:
From your site:
As you can see it's intellect, mp5 and hit values that are way off (although I don't really care about hit since I know I am hitcapped already).

I have tried looking through all settings in the spreadsheet and in your online version and I am pretty sure I have everything matched between the two.

Do you have any suggestions as to why I see this big difference?

Thanks in advance.

 

Rivkah

Thanks for bringing the problem to my attention, I've updated the Pawn string to use epic level gems. I'd forgotten to change that when I implemented them.

With regards to the differences in mp5, int and hit, the way those values are calculated may be slightly different. Both the spreadsheet and my site calculate the dps per stat by increasing by an increment and recalculating the dps. I'm not sure if we're using the same increment which would have some effect, I'll have to doublecheck what the spreadsheet is using, at one point I reduced my increment to increase accuracy a bit on those stats which don't scale in a consistent fashion.

Given that the int and mp5 values are off I would expect that you're getting a different viper penalty between the spreadsheet and the site, which may indicate small differences but I can't tell without seeing a copy of your settings. You can send me the file in email if you want, my email is listed in the feedback button on the site. I'm not sure about the hit gap, it may be because the spreadsheet splits out pet and player hit, where I just calculate them all together.


With regards to the gearstrings that Bluesfear posted about awhile ago, I looked into them and I'm not sure it's worth implementing an import for. The format they're in is XML-like but the tags are invalid for XML so I can't use a standard parser to read them. In order to setup an import for them I'd need to parse them manually. Additionally I noticed the file generated by Excel creates a bunch of different profiles in it, so in order for someone to import just one they'd have to find the one they want and paste it in, and there also seem to be some control characters in there that are causing problems with reading pasted data. It just seems like the time investment to make it work would be pretty considerable and I'm not sure how easy it would be for people to use, so for now I'm going to shelve that and maybe take another look at it later.

 

Plasmox

I noticed that when I load in a hunter with enchanting, and their rings are enchanted, then load up a hunter that doesn't have enchanting, the rings remain enchanted.

It's not a big deal, just annoying because i have to remember to remove the enchants on the rings...

 

alienangel

Rivkah, is there a reason the site won't let me specify a wait time of 0 seconds for the shot rotation test? I do see the option to disable waiting for Arcane/Aimed/Multi, but on the spreadsheet at least, my highest DPS is (and has been for a while) with no waiting for anything (i.e not chimera/killshot either).

Thinking about it, not waiting for aimed but waiting for chimera may indeed be better than not waiting for either, but right now with no option to disable waiting in your site, and no option to only wait on shots other than Arcane/Aimed/Multi in the spreadsheet, it's a bit hard to compare.

 

Rivkah

Thanks for bringing my attention to the problem. There may be some technical issues with having a zero wait time which is why it's not allowed for now (although I may fix that later). But what I have done is change the settings so that you can now enable/disable wait time for kill shot, aimed/multi, explosive/arcane and chimera separately. This should allow for more customized testing (and if you don't want to wait for any shots you can just turn off all of them).

I tried to adjust things so that older saved settings should default to logical values, but it's possible that some may be affected by the change so if you're getting weird results on older saved settings be sure to check under the settings to make sure the wait shots are setup properly.