5.2 Trinkets Discussion

[Coldkil]

So the Rune isn't great for ALL Rogues or just mut/combat (which I agree, is nearly 99.9% of all Rogues)?

The problem of Rune is that the proc doesn't just add X to a stat (which is definitely a dps increase, no matter how little). The problem is that the difference between stats and their sinergy makes the stacking of one non convenient enough - the more you stack of one, the better the others become.

While we talk about gemming/reforging, the delta of stats is wy smaller - the effect of this "stacking" is small to the point it's negligible (to a degree). RoRO on the contrary give a freaking huge boost to one while bringing to 0 the other two. This means that the EP of the stat granted will go down a lot, and the EP of the other two will raise a lot (possibly becoming istantly better).

Sub actually is the spec which has the "worst situation" - all three stats are very near in terms of EP (which basically makes "viable" to focus on each one of the three) but their effects make also better each other. RoRO wuold break the balance hence making the stat boosted worse than it was.

Beware: RoRO IS a dps increase. Just is way smaller than it looks by the trinket description - rogue mechanics don't favor at all it.

[Maelstrom51]

Beware: RoRO IS a dps increase. Just is way smaller than it looks by the trinket description - rogue mechanics don't favor at all it.

Its a dps increase over really shitty trinkets or nothing at all. But compare it to any decent trinket and no chance.

[wimp]

Note for theorycrafters: I'm starting to lose faith in my ability to model the talisman. The Poisson modelling isn't holding up - neither Hamlet's nor Conjor's maths are modelling this trinket accurately. And Ryme's and my own efforts haven't succeeded.

I doubt there will be many forum readers here who are into Markov processes, Step functions and Poisson distributions... That's roughly where I've got to.

If you think you've got the maths I'd be ever so keen to hear from you. A good test would be to see if you can model this much simpler trinket:

Talisman of Egregious Maths: it very rarely procs (say, after taking haste into account 0.01 times per minute), but when it does, it gives you SO much haste that it will be proccing like crazy (after taking haste int account, say 100 times a minute). Question: what is the total uptime?

Here is where I'm at:

• Our current Poisson approximations break down.
• the answer needs to be dependent on the fight length (for short fights the uptime on the 'Talisman of Egregious Maths' would be close to 0%, for infinitely long fights, it would be close to 100%.
• I'm starting to suspect there is no exact closed form solution, and that simulations (and testing) might be the way here...

While on the subject of simulations: there's something up with Simulationcraft: the sum of the uptimes of the different stacks don't quite add up to the overall trinket uptime (the two are out by around 1%, which is a margin of error of around a 2%). That would't be a drastic change if the error is on the overall uptime, but is a significant change if the error is on the 5th stack. While I haven't looked at the simulationcraft code, for now that discrepancy (which I've shared with the simulationcraft people, who agree something's up in there) is making me a bit worried about simulationcraft results for this trinket.

I don't have a Talisman myself - but even when I do eventually get it, I'm not sure how to get WoL to show me uptime for each stack size. I could do this manually parsing the logs in excel - but then I'd run into not having a sufficiently long sample (it would need to be a long test, probably around 2 hours, since we're interested in the uptimes per stack size, and the uptimes of the 4th and 5th stack are pretty low, so you need a long test duration to get to a point where your statistical deviations (errors) are small enough...

The maths isn't very far off (we're talking differences of around 1%). But just wanted to give a general heads up that it's not perfect (nor is Shadowcraft, which uses Hamlet's maths, nor is Simulationcraft - not sure how the simulation works).

The end of the thought process? Until we get new and better maths/modelling: Heroic Thunderforged Renataki's, Bad Juju and Talisman are all comparably good.

[Deleted]

While on the subject of simulations: there's something up with Simulationcraft: the sum of the uptimes of the different stacks don't quite add up to the overall trinket uptime (the two are out by around 1%, which is a margin of error of around a 2%). That would't be a drastic change if the error is on the overall uptime, but is a significant change if the error is on the 5th stack. While I haven't looked at the simulationcraft code, for now that discrepancy (which I've shared with the simulationcraft people, who agree something's up in there) is making me a bit worried about simulationcraft results for this trinket.

The Simulationcraft issue with the last stack uptime was a data collection/display issue. The modeling of the RPPM mechanism w/ regards to last stack was not bugged. It's fixed in the current simc release (520-8), where the uptime sum for individual stacks should now be within 0.01% or so.

[wimp]

Good to hear Navv. Thanks for your continuing work on simulationcraft too! Downloading 520-8 as I type...

[Ryme]

Seriously, screw that trinket I went through an entire marker pen trying to grasp a complete model of it. I even started reading up on things like Cox processes; it was not a fun read but ultimately I believe could yield the solution, albeit the work to get there would be enormous.

[Deleted]

I would like your help with my current trinkets becasue im rly comfused - Atm i have the following trinkets Renataki hc thunderforged, vicious (valor point one), talisman normal, bad juju normal.. According to some posts i have read seems renataki and talisman are the best to use BUT according to shadowcraft the best choices are renataki and vicious plus with renataki im at 10% expertise which is useless.. So i dont know wat to do and would love some help from u.. My armory page is that http://eu.battle.net/wow/en/characte...Nerya/advanced

[wimp]

I have edited the trinket list to reflect the latest maths (not reposting as to not clog the forums up with multiple tables): you can find it here (this is a link)
I have also changed the order of stuff in that post, to have the info people want to see first, and the maths second. And I'm being a bit less ego-centric this time and posted the EP values as per the BiS gear (rather than my own).

Main change: talisman is a bit less good than we previously thought.

[Ryme]

Comes out pretty damn close to the calculus model, with far less headaches

[Kael]

http://eu.battle.net/wow/en/characte...Nerya/advanced

I'm not entirely sure how ShadowCraft models ToB right now - FieryDemise might show up to let us know, likely - but the logical response would be ToB/RSC (522/541). ShadowCraft seems to think Bad Juju/RSC are your top options. I'd recommend checking against SimulationCraft for ToB/RSC and Bad Juju/RSC as combination options, given that between 2 RPPM trinkets and the legendary meta, hastes value has already risen - to the point where it's head to head with mastery, but doesn't seem to be receiving that value in trinket comparisons.

[rayanne]

Random question that I hope hasn't been answered (scanned a few pages but didn't see it)

With old ICD trinkets, like the shado-pan one, you can re-equip it prior to a pull to activate the ICD and try to adjust when it procs during the fight--which can be helpful for fights like Jin'Rokh or Primordious. I was wondering--does re-equipping have any effect on RPPM trinkets? I know there is no ICD, but would it reset the "time since last proc"?

[theherecy]

I know there is no ICD, but would it reset the "time since last proc"?

Believe it does. Which considering that BL is popped on first pool (usually) doesn't make it attractive to do so.

[rayanne]

Believe it does. Which considering that BL is popped on first pool (usually) doesn't make it attractive to do so.

Yeah I was thinking more for H-Primordious, trying to time it for when I am fully mutated.

[theherecy]

Yeah I was thinking more for H-Primordious, trying to time it for when I am fully mutated.

Honestly I wouldn't bother, you usually get a couple of stacks of haste from there. And it's gg... many procs, left side.

[Deleted]

I'm not entirely sure how ShadowCraft models ToB right now - FieryDemise might show up to let us know, likely - but the logical response would be ToB/RSC (522/541). ShadowCraft seems to think Bad Juju/RSC are your top options. I'd recommend checking against SimulationCraft for ToB/RSC and Bad Juju/RSC as combination options, given that between 2 RPPM trinkets and the legendary meta, hastes value has already risen - to the point where it's head to head with mastery, but doesn't seem to be receiving that value in trinket comparisons.

I did simulacraft for all my trinkets and those are the results: 176191 (vicious, renataki), 173440 (renataki, talisman), 177354 (renataki, bad juju).. It seems simulacraft also agree with shadowcraft and the best option is rsc-bad juju and if reforge both of them to haste the dps going higher.. So i should follow that?
I was just thinking with talisman uptime which is high renataki would proc more as more haste = more chances to proc so dno why shadowcraft / simulacraft seems to have that combo so low for me

[Kael]

Assuming you're reforging each profile around the gear changes, of course?

I'm not sure why ToB is showing so low, but unless someone has an explanation for why it should be ahead in your gear it's probably worth following the advice of both platforms.

[Deleted]

Ye ofc im reforging around the gear

[Ahdehl]

For you math dudes that are way smarter than I am:

There was a post earlier about a graph showing the value of each trinket at 0:01 of a fight to like 8:00 of the fight. Could wimp/Ryme remake that graph using the new trinket valuations, and extend the graph out to around a 15:00 total fight time? Donations to your favorite charity (or a pizza/etc) would be made

[wimp]

Sure thing.

The first five minutes of the fight:



And from minute 5 to the end of the 15th minute:



Same assumptions as for the current trinket ranking: 16.09% haste, 5% spell haste buff, heroism/bloodlust were used. Stat EP values for BiS gear as per the EJ Assassination OP (Agility: 2.72; Crit: 1.15; Mastery: 1.17; Haste: 1.39).

And the same notes as for the previous graphs:

1) The behaviour of non RPPM trinkets is pretty accurately modelled from this graph: for a given fight, they *will* proc soon after they go off cooldown.
2) The behaviour of the RPPM trinket is much less predictable for any given fight (Ghostcrawler would argue more interesting): the curves just show their 'average' EP value after x amount of time - this curve is the statistical value, and what you'd see if you compared many different fights. Their actual value will vary from fight to fight - with probability to vary decreasing with the size of the variation.
3) This graph assumes you've waited long enough to guarantee an RPPM trinket proc on pull.
4) This calculation spreads out the value of bloodlust/heroism over the entire fight (thought for fights under 40 seconds it assumes bloodlust happens at the start, hence the minor oddity on the talisman graph at the 40 second mark. Didn't notice that? Good, nothing to see here... Move along now

[Ahdehl]

Ya, those were the graphs I was referring to. What I was wondering is if it was possible to get the graphs with the 541 versions of the trinkets? And would the graph values change depending on what combination of trinkets I'm using?