Going from T15 BiS to T16 BiS (PTR build 17337)

[rogue7777]

shadowcraft shows that haste>mastery and this is definitely wrong

[Sicktight]

I tweaked dextar's assassination a little bit and got 414,506.5 switch 6 crit/agi gems to 6 haste/agi gems. I would post a link but I do not have enough posts to do so.

[Dextar]

shadowcraft shows that haste>mastery and this is definitely wrong

Shadowcraft has said Haste>Mastery for a while, I've completely disagreed with it and quit using it awhile ago for gem/reforging. I'll still be rolling with Mastery>Haste>Crit like I have been all 5.2.

[fierydemise]

shadowcraft shows that haste>mastery and this is definitely wrong

Comments like this bother me. I'll be the first to admit the limitations of the ShadowCraft assassination model, its cooldown modeling could be better and there are some long standing limitations to how ShadowCraft handles envenom uptime but in general it works pretty well. Perhaps more importantly the results ShadowCraft spits out are very similar to the results from SimCraft, there can be slight differences but overall ShadowCraft and SimulationCraft basically agree. Thats a big deal, that we have two independently developed tools saying the same thing gives me a lot of confidence in the results and should give you a lot of confidence in the results too because the chance of that happening without both tools being reasonably accurate is incredibly low. Bugs certainly happen, see for example the negative dps bug with the combat model and I won't claim any ShadowCraft model is perfect or bug free but they are probably pretty accurate.

More broadly how do you know haste>mastery is incorrect? Why is mastery the best assassination stat a priori? If you have a way for us to determine stat weights a priori I'd love to know because that would make things so much easier than having to hack new features into ShadowCraft.

[Fluorescent0]

To further expand on fierydemise's reply, I'm going to quote a couple of posts I did a while back on the same argument.

Do keep in mind that meta's interaction with haste won't be nerfed and that we'll have a new powerful RPPM effect in our cloak. Reports have the cloak at about 8-9% of our damage, with the meta probably between 5 and 7%. Considering the meta was the primary reason we switched to Haste this tier I could definitely see it as a viable stat, especially now that the buffs to Mutilate and Dispatch make having more energy even more of a bonus.

I was asked for a backup to my claim that the meta is the more prominent effect in our decision of switching to Haste:

The effect of the meta was by far the prominent effect. Simple example given using my character profile on shadowcraft.

Current:
Haste 1.431
Mastery 1.375

No meta:
Haste 1.434
Mastery 1.531

Meta but valor trinket instead of juju (complete reforging and regemming included, both of which still favor haste)
Haste 1.364
Mastery 1.343

See how much more values fluctuate when you swap out the meta compared to when you swap out one of the trinkets?

So I wouldn't be -that- surprised to see Haste still a viable option.

[Dextar]

The whole reason people started pushing haste was to increase procs from RPPM trinkets, mastery scaling, and the meta gem attacks IIRC. Haste is completely unreliable and doesn't increase the procs enough to justify running full haste. Simcraft shows combat as being on top of assassination even just by a bit and that's not the case, these tools don't account enough for real world performance.

Just grabbing the numbers quickly off Simcraft, DP dot and instant account for 32.5% of our total damage, mastery not only increases that damage done but also the damage from VW and Envenom (although quite a bit less). Haste doesn't give them a direct increase and the indirect increase it gives is not enough to push it over. Furthermore I dislike that ShadowCraft keeps saying to throw full secondary stat gems in sockets, that directly reduces your AP which lowers your damage further. I find it interesting that it continues to push those full secondary stat gems so hard, which is completely different from what simcraft says as well.

The closest I could find between myself and Chiefwiggum (yes I know I have a little bit better gear):
http://www.worldoflogs.com/reports/r...8/?s=408&e=775 - Chief
http://www.worldoflogs.com/reports/r...?s=6068&e=6436 - Dex
It's the only week I have where I wasn't helping ranged but our active times and kill times are only off by a second.

Edit to add reply to Fluorescent0:
The cloak was almost 10% of my damage but it got nerfed quite a bit, I don't remember the exact damage off the top of my head but the last testing we did last weekend it wasn't nearly as high. If you look at the logs I just posted above Chief had 10 more procs but my damage was higher (though not much) and while their kill was on the 8th it looks like the only piece of gear difference was the featherflight belt. I tried to keep it as close as possible on as close to patchwerk as possible.

[Fluorescent0]

I... don't really get what you're trying to show with two logs of one single fight? That's not enough to draw conclusions. It's anecdotal and doesn't really show anything at all. If anything, you had 50% renataki uptime, which just screams pure luck to me.

Keep in mind that not only Haste increases procrates, but it also increases autoattack damage, poison procs, envenom uptime and direct abilities damage. It's a lot more complicated than you're trying to make it.

[Pathal]

Quite a lot of people have issues separating themselves from confirmation bias. It's one thing to trust SimC over Shadowcraft, I really don't have any problems with that, but some arguments like "Assassination does a lot of poison damage so obviously mastery should be worth a lot" are pretty terrible.

[fierydemise]

For obvious reasons trying to compare two logs to draw broad conclusions about stat priority is ridiculous. With RPPM trinkets variences of up to +/- 7% can be expected so any kind of single log comparison is going to be far more likely to see difference based on RNG than actual theoretical dps differences.

Further your conclusion about haste being unreliable is incorrect. This was a topic of much discussion several months ago but the long and the short of it is haste mitigates RNG it doesn't increase it. Additionally when we're talking about the legendary meta gem and cloak procs which do continue to scale with haste in 5.4 the RPPM values on those trinkets are high enough that considering them as scaling linearly rather than discretely as we do trinkets is perfectly valid.

[Warstar]

RngPPM trickets are 100% luck based. A full standard mastery build has a higher potential dps. Haste has a high weight to it due to RPPM trinkets but even a haste stacked class like combat who only gems combat vs. a full mastery assassination rogue has a chance to have a lower RPPM proc rate than the full mastery build and vice versa a full mastery build has a chance to have more procs than a full haste build. While haste directly influences a lot of stuff there's still a large luck value. Stacking haste is like stacking magic find in Diablo. It doesn't guarantee you a higher drop (proc) chance but it raises those odds by a lot where as mastery is a 100% guaranteed damage increase. Over 10000 fights if you look at the numbers haste will 100% give you that damage increase but in real time it's still luck based more than anything.

Hell even on shadowcraft with my gear doing a full mastery build and only reforging to haste if I have a mastery/crit piece but other wise keeping mastery, reforging to mastery and gemming agi/mastery in red and mastery in yellow shadow craft gives me a whopping 560 higher dps and I have 2 RPPM trinkets and the meta. That's not a huge dps increase especially when you consider that a lot of that "increase" is just a higher luck chance.

Now also combine the fact that shadowcraft is a patchwerk style fight sim. A fight like Tortos that has a high aoe component completely skews that into mastery favor. A fight with a high movement with less boss up time also favors mastery for the higher poison damage done.

So personally where I have a direct damage increase with no arguments there. Mastery increases damage. Or haste which is a minimal damage buff but a higher luck increaser I'll choose mastery. Also because mastery + random dumb luck will give potential higher damage.

[Dextar]

For obvious reasons trying to compare two logs to draw broad conclusions about stat priority is ridiculous. With RPPM trinkets variences of up to +/- 7% can be expected so any kind of single log comparison is going to be far more likely to see difference based on RNG than actual theoretical dps differences.

Further your conclusion about haste being unreliable is incorrect. This was a topic of much discussion several months ago but the long and the short of it is haste mitigates RNG it doesn't increase it. Additionally when we're talking about the legendary meta gem and cloak procs which do continue to scale with haste in 5.4 the RPPM values on those trinkets are high enough that considering them as scaling linearly rather than discretely as we do trinkets is perfectly valid.

The same could be said for almost any logs, Chief was haste but I got lucky with the proc. The legendary gem and auto attacks IMO don't provide enough damage to change to haste and as I said, from what I've seen personally I felt haste was too unreliable. Chief has full haste yet I got the luck of the draw and got another proc yet the fight was the same length, no matter how you want to spin it, haste didn't help him with that RNG just screwed him.

You look at logs from Jommys and Kidneyspears, #1 and #3 on Ra-den (Mastery and Haste) while Kidney had a longer fight they only ended up with 8 Juju procs where Jommys got 15, that's AMAZING (another reason why I absolutely hate RPPM ) but there shouldn't be that big of a difference between the two. 20 second difference and almost a 50% more proc rate to Jommy (who is Mastery mind you).

[Fluorescent0]

The same could be said for almost any logs, Chief was haste but I got lucky with the proc. The legendary gem and auto attacks IMO don't provide enough damage to change to haste and as I said, from what I've seen personally I felt haste was too unreliable. Chief has full haste yet I got the luck of the draw and got another proc yet the fight was the same length, no matter how you want to spin it, haste didn't help him with that RNG just screwed him.

You look at logs from Jommys and Kidneyspears, #1 and #3 on Ra-den (Mastery and Haste) while Kidney had a longer fight they only ended up with 8 Juju procs where Jommys got 15, that's AMAZING (another reason why I absolutely hate RPPM ) but there shouldn't be that big of a difference between the two. 20 second difference and almost a 50% more proc rate to Jommy (who is Mastery mind you).

Highlighted the massive flaw in the first sentence. Can we get away with opinions and try and model stuff instead? Everyone believed the earth was flat, yet it wasn't.

Again, top parses are outliners of massive luck, especially on an easy and forgiving spec such as Assassination. If you can't understand that, I don't really think you should be discussing stat priorities. If you can, your example has just been invalidated because everyone agrees that Mastery has higher potential DPS. (keyword being potential)

[Warstar]

From what I've experienced mastery isn't just higher potential it's a more even damage output.

Stacking has is trying to game the system into giving you more procs so you have a lot wider swing of dps from a high proc fight to a low proc fight where mastery is more evened out over all but spikes to a higher thresh hold.

[Dextar]

Highlighted the massive flaw in the first sentence. Can we get away with opinions and try and model stuff instead? Everyone believed the earth was flat, yet it wasn't.

Again, top parses are outliners of massive luck, especially on an easy and forgiving spec such as Assassination. If you can't understand that, I don't really think you should be discussing stat priorities. If you can, your example has just been invalidated because everyone agrees that Mastery has higher potential DPS. (keyword being potential)

You have logs which are real world results or you have models which are unreliable. There are plenty of logs where haste beats mastery and plenty of logs where mastery beats haste. The models? Haste. They aren't designed to handle human error, input lag, any kind of lag properly, multiple target switching and they don't handle RNG very well. Haste is supposed to negate the RNG but it's NOT, the damage you gain from the meta is negligible at best so if you agree that it has higher potential DPS why are the sims pushing for haste so much? They are incorrect, that's why. All it takes is one thing to be modeled incorrectly for the whole thing to be off.

So again, if everyone agrees that Mastery has higher potential DPS then why does SC force people into haste? Incorrect modeling?

You agree and you're running haste, why?

[fierydemise]

The same could be said for almost any logs, Chief was haste but I got lucky with the proc. The legendary gem and auto attacks IMO don't provide enough damage to change to haste and as I said, from what I've seen personally I felt haste was too unreliable. Chief has full haste yet I got the luck of the draw and got another proc yet the fight was the same length, no matter how you want to spin it, haste didn't help him with that RNG just screwed him.

Yes chief had bad RNG on that fight but that doesn't prove that haste doesn't protect you from RNG anymore than if I looked at a much lower than expected trinket uptime in one of your logs and said "this trinket uptime is low because you have low haste." The only way to get an accurate view of dps from logs is by using large scale log aggregation such that your data set is large enough to hide the RNG effects.

You have logs which are real world results or you have models which are unreliable. There are plenty of logs where haste beats mastery and plenty of logs where mastery beats haste. The models? Haste. They aren't designed to handle human error, input lag, any kind of lag properly, multiple target switching and they don't handle RNG very well. Haste is supposed to negate the RNG but it's NOT, the damage you gain from the meta is negligible at best so if you agree that it has higher potential DPS why are the sims pushing for haste so much? They are incorrect, that's why. All it takes is one thing to be modeled incorrectly for the whole thing to be off.

So again, if everyone agrees that Mastery has higher potential DPS then why does SC force people into haste? Incorrect modeling?

You agree and you're running haste, why?

You missed the important part of Fluorescent's point. Top ranks generally represent RNG outliers, fights with RNG that you don't expect to see very often. That doesn't mean haste is bad, haste should in the general case make RPPM mechanics more consistent rather than less however for ranking given that ranks are highly RNG dependant mastery will end up better because if you benefit from exceptional RNG in a mastery build you spent less of your itemization budget to get those trinket uptimes. Just because mastery may be better for ranking when relying on crazy RNG doesn't mean its better for average case dps which is what ShadowCraft is based off of.

Of course this entire conversation becomes irrelevant in 10 days when haste will no longer increase trinket proc rates. The value of haste will still be increased by the meta gem and cloak but since those high higher RPPM values the impact of RNG swings will be much harder to notice. That is the difference between 5 procs and 6 procs is much greater than the difference between 20 procs and 21 procs.

[Fluorescent0]

You have logs which are real world results or you have models which are unreliable. There are plenty of logs where haste beats mastery and plenty of logs where mastery beats haste. The models? Haste. They aren't designed to handle human error, input lag, any kind of lag properly, multiple target switching and they don't handle RNG very well. Haste is supposed to negate the RNG but it's NOT, the damage you gain from the meta is negligible at best so if you agree that it has higher potential DPS why are the sims pushing for haste so much? They are incorrect, that's why. All it takes is one thing to be modeled incorrectly for the whole thing to be off.

So again, if everyone agrees that Mastery has higher potential DPS then why does SC force people into haste? Incorrect modeling?

You agree and you're running haste, why?

Higher DPS potential, as in it can (when stars align) pull higher numbers. Not higher DPS potential as in it averagely does more damage. Those are two completely different things, and you want reliability during progress.

[Deleted]

I'd like to refer you to Wimp's post (last on the page) way back in May on the haste / mastery topic. It's getting pretty tedious with these threads still popping up after all this time, and all the debates and spats about "which is better".

http://www.mmo-champion.com/threads/...-gritty/page11

Strictly in my opinion, i prefer Mastery as a priority due to the current reliability it offers, yet i am fully aware of the potential gain from running Haste. Understandably this gain is reflected in ShadowCraft yet it still relies on 'luck'. Either way, it doesn't have such a big impact on your gameplay in 5.3 that one is a clear option. If you choose one or the other, you won't have a significant gain / loss, and you can still produce top ranks if you play correctly.

[fierydemise]

Saying haste relies on luck is playing around with the meaning of the word luck somewhat. If I flip a coin 100 times and get 50 heads and 50 tails is that a "lucky" result? If I flip it 100 times and this time get 55 heads and 45 tails is that a lucky result? Haste increases the average number of procs you expect over a fight, that isn't really luck.

What does complicate things is that since you tend to not have that many trinket procs over the course of a fight you have to consider trinket procs as discrete rather than linear. That is 5.7 procs isn't any better than 5.3 procs persay because there is no such thing as .4 procs. However those extra .4 procs can in practice protect you from some amount of bad RNG. This point about discretized trinket procs is something that ShadowCraft doesn't handle well at all and is best evaluated using SimCraft. However going back to my point earlier about ShadowCraft and SimCraft basically agreeing that balanced haste/mastery is optimal suggests that this issue isn't that relevant to accurate dps modeling.

Broadly speaking the goal of modeling isn't to exactly represent a system, that would be hard and for most systems a simulation would be far better than a model for that purpose. The goal of a model is to handle the important things and the 2nd and 3rd order effects that don't really impact output can be safely ignored to simplify things. This is what ShadowCraft does, it doesn't need to handle everything exactly to be an accurate tool and that allows you to evaluate gearsets in ShadowCraft much faster than you can do with SimCraft.

I don't really care too much about the overall conclusion of haste vs. mastery, I think that is basically a settled issue. In 5.2/5.3 given legendary meta gem and 2 RPPM trinkets the optimal gearing appears to be balanced haste/mastery. That is what wimpy's numbers show, that is what ShadowCraft shows. My objection is to statements that mastery>haste as some a priori principle. If stats could be determined a priori then we wouldn't need tools like ShadowCraft or SimCraft. If you want to make the case that mastery is better for a fight like lei shen because of AoE or mastery is better for ranking (both statements I think are pretty well supported) then say that don't make blanket statements about stat weights unless you have some evidence that there is a bug with our tools causing them to over/under value certain stats.

[Dextar]

I don't really care too much about the overall conclusion of haste vs. mastery, I think that is basically a settled issue. In 5.2/5.3 given legendary meta gem and 2 RPPM trinkets the optimal gearing appears to be balanced haste/mastery. That is what wimpy's numbers show, that is what ShadowCraft shows. My objection is to statements that mastery>haste as some a priori principle. If stats could be determined a priori then we wouldn't need tools like ShadowCraft or SimCraft. If you want to make the case that mastery is better for a fight like lei shen because of AoE or mastery is better for ranking (both statements I think are pretty well supported) then say that don't make blanket statements about stat weights unless you have some evidence that there is a bug with our tools causing them to over/under value certain stats.

The problem is ShadowCraft doesn't show balanced haste/mastery, it shows purely haste. In my current gear with 8527 haste and 13267 mastery, it's giving an EP value of 1.535 to haste and 1.264 to mastery, reforged and regemmed via auto and optimize it gives haste an EP value of 1.506 and 1.358 to mastery (11984 Haste/9810 Mastery).

[shadowboy]

Strictly in my opinion, i prefer Mastery as a priority due to the current reliability it offers, yet i am fully aware of the potential gain from running Haste. Understandably this gain is reflected in ShadowCraft yet it still relies on 'luck'. Either way, it doesn't have such a big impact on your gameplay in 5.3 that one is a clear option. If you choose one or the other, you won't have a significant gain / loss, and you can still produce top ranks if you play correctly.

Yup, sticking with preference is why I am focusing on haste on my assassination rogue. I prefer the feel and I consider combat my main spec but I often feel it's a big liability to be combat on some fights (others I will play either spec based on my whim). Going with a haste assassination build is more compatible with combat so I go that route (it leaves my combat gear somewhat suboptimal though since I use agi gemming for combat).

I'm indecisive about 5.4. The gearing is different enough that I think I will have to choose a spec. I will likely choose combat since it is my preferred spec and switch to deft/quick/lightning gemming.