The current state of Affliction theorycrafting and guidelines for gearing

[Deleted]

Wah, SB:SS can miss, iirc it's calculated on a per dot basis, so you'll sometimes find agony and corruption apply, but your UA miss. I've never been "refunded" my shards for missing a dot with SB:SS.

When you cast SB:swap the spell with id 119678 needs to hit first and do some direct damage. It has a miss chance of 0.15 - hit_percentage on a boss. If it does not hit, then an aura with id 74434 will not be removed from you, and that aura can be used for any purpose that those auras can serve. Since the choice after missing with SB:swap is often to SB:swap again, there is not a Shard lost in this miss event, or in other words the number of Haunts remains the same for that fight.

If 119678 lands then dots are casted on the targets for you. Dots do not need hit rating, by which I mean that they scale poorly with hit, unlike the other spells that were listed to us. It does not matter if you cast 1 or 10 dots with one click, when they have separate rolls for missing. Therefore threre is no need for hit with SB:swap.

With all the rolls, there is by the way only a 0.9025 chance to hit with a specific dot in SB:swap, if a single dot has a 0.95 chance of hitting from gear. But when you have three of them dots then "hitting with any two or three" has a chance of 0.94311.

Missing with SB:swap comic:
http://imageshack.us/a/img195/8308/sbswap2.png
http://imageshack.us/a/img17/5016/sbswap3.png
http://imageshack.us/a/img837/6356/sbswap4.png
http://imageshack.us/a/img27/5644/sbswap5.png
Number of shards, as Shards, becomes 2 before the last picture because there is a Nightfall.

In soviet Europe, you can click mouse6, alt-mouse6 in one strip.

I understand you're European, and so English might not be your first language, but it is very difficult to understand what you are writing.

In soviet Europe, you do not prove wrong, those people, who you do not understand.

This is the boss with only single target dps. Log is short so I didn't waste too much time. If boss dissappears during proc, then the refresh can be expected to be 18 seconds after the last seen refresh.

Okay? You did no such analysis in your post.

In soviet Europe, explaining the data or analysing other things, does not stop after first results.

I told I am mostly talking about defaultee cases, so choosing something with single target is not a huge surprise. If you click the link I made you can click 3 times to find how many targets he hit and maybe 15 clicks each to find what bosses have more targets.

Length of the log is presented in seconds, if you press the link.

Soviet Europens learn it as a child that short logs take less time to read than long ones.

"Any number higher than 18 would mean one more doable cast."

My post meant that in every case when the events are over 18 seconds you expect another cast. Makes no difference what the cause is. And there was an example where a procedure was made.

When I wrote that the boss is dissappeared, I did not analyze anything but made a guess to explain the data. My explanation can not be proven to be free from my friendly affections towards Marilina, so in the name of objectivity, I am ready to accept it is false, and the stupid rat just didn't see his dot drop.

No? There is a refresh that comes .45s before Dark Soul fades.

I don't understand your ..|. notation.

It is French. And the "." after "|" stands for last gcd, or tick duration, or your .45s. If each mark was one tick apart from each other, then it is a way to draw my approximate distribution: ...... and so forth, but you clearly made a "|" pop out when you looked at DS. Also in ..|. the time axis is not growing, if you give this distribution the same name I did, but now it shows time in the combat and you can draw distribution on that axis if you want.

Finally, you're the one who brought Marilina's log into the discussion. You expected to find a result in your favor, and now when it's been shown that Marilina plays to maximize procs you're going to discount the log. Poor form.

This did not happen by chance:
First it was someone else who said some procs are possibly missed without cap. I asked what is the rate of missing them, or average amount of this kind of misses, or even the difference caused for dps. I get no answers so I tell you how to get part of the answer. And present you with approximation 1/total_ticks as one term in calculating the probability. I knew no one could say anything against my approximation. But you begged for data on something not so specific and now it is here.

How am I discounting the log?
In soviet Europe, lab rats are the number one test subjects.

As for ruling out weaker buff stuff - the point is not that we're ignoring weaker buffs - the point is that what you looked at with Inner Brilliance is a flawed, tiny piece of the picture. IF you're going to look at ONE BUFF ONLY you would look at Lust/DS. If you're going to look at all buffs then consider them all. But to consider one buff only then look at Inner Brilliance, if you never considered that your analysis might be garbage you are being disingenuous.

You were ignoring every buff, in everything, until I made a post about one, for something.

My approximation can be used for "looking" at many things and to cover several buffs. When it means that "looking" is finally comparing the profiles, and no logs are used anymore. That kind of looking takes different kind of work, when so far I was contempt with only telling you there is a real approximation for the possibility of procs and dots being poorly synced. There is hardly any point to check if the approximation is correct by going through more buffs. Or if you must, then it should occur to you at some point that the data could be more useful when gathered by looking at the other dots, against the same buff.

There are more procs like Inner Brilliance, than there are cds like DS. You would need to know if DS happens often enough. But having a random proc was one of the important points. My whole purpose is gone however, if the buff is too strong. Can we agree that every person is using a SB:Swap when their DS/BL is running out, regardles of what their dot durations are at that point?

You don't need any logs to know what is going to happen in that case: Hit capped profile always delivers. Non capped has a miss_chance to miss once and miss_chance^2 to miss twice etc.

If you make a distribution in my footsteps, then in your ..|. you are only looking at opinions that people have about "when it would be nice to click a button?", or you can maybe measure the fear vs. greed on persons who don't cap. If you insist calling it anything else, like it was somehow helping you in the hit cap discussion, then you are not supposed to make such distributions by looking at one kind of profile alone. Non capped locks could have made entirely different choices, when choosing their minimum time.

You could not be expecting to find proof for the premise (is it yours, how could we tell?) "There is a possbility to fail at refreshing a dot before DS runs out", by using hit capper logs. My samples however are not nearly as narrow, in my purposes, because Mirilian did not precast the Corr, so much that his distribution would start to differ from flat approximation, and non cap lock would probably not do that either.

Finally, you're the one who brought Marilina's log into the discussion. You expected to find a result in your favor, and now when it's been shown that Marilina plays to maximize procs you're going to discount the log. Poor form.

It has not been shown that Marilina maximizes procs. First, my samples do not have anything to do with it, after you see that there is no precasting, ie. results with only small values. I remind you of this:

If we don't hit cap we will sometimes not be able refresh the dots with procs when we want to. Not only will we waste globals on misses but also will we get a lower uptime on the said buffed dots and have to start channeling Malefic Grasp with these procs up later. It just puts us behind a lot.

Yes sometimes the lock can have a proc running out, and a dot that has little over 0.5 duration remaining, where 0.5 is the legit max after which he can refresh. If he has one miss event, there after 0.5, he does not get a buffed dot. This will happen in default action list.

(And everyone, except the pre-caster and a fel flamer, is like a default profile rat.)

Thing that proc abuse profiles actually have different, is taking place after dot duration is under 0.5, or when remaining ticks is total_ticks/2. Therefore if you are interested in how Marilina plays, but can not ask him, you could make a distribution of the time periods between Inner Brilliance coming up, and the next Corruption he casts. Then the low values would indicate that he was witing for a proc to appear. This would have more inaccuracy than my work, because he might not prioritaize one proc at this.

Then secondly, those DS refreshes are just ruled out as something any profile could do. But what if Marilina was a proc user then? Then Marilina would become a person who has more potential without hit cap? Maximizing DS benefit is a thing that makes non capping better than already when not doing it. Ask Bonkura about theese if you don't belive me.


You did not say anything about what you thought was going to happen with your DS. What do you think that looks like, in soviet Europe? Basicly you did not even state the purpose of your work. Or do you have in mind now how to utilize the ..|. ?

Having close to a flat distribution with Inner Brilliance in the above log at the moment means this: when I said that I would use such a thing to predict a chance for damage penalty, I am correct. But also you should get the idea that the 1/total_ticks I am talking about is the worst possible term that can be used in those situations. So if I am not correct, then the more stupid you make your hit capping to look like. It is still in the favor of my stand, to have an upper limit for the bad events, if for you they are otherwise only irrational fears.

The opener is also misleading because there are shorter lasting procs than the Dark Soul. In Marilina's case, it is:

You said this earlier. And you were mislead as far as to china atleast, because length of the buff or proc cd was irrelevant for flat distribution approximation itself, or the possible tests (two of a kind) on its accuracy.

[crysilaero]

So.. let me get this straight. Analyzing your post I came to the conclusion that you are from soviet Europe, is this right?

Anyways, I think ppl under-analyze hit rating. Among other reasons, there are three huge reasons:

1- Hit rating affects soulshatter. Missing a soulshatter on a critic time with dots rolling can lead to a big wipe.
2- Playing affliction on this content it's all about refreshing dots on a very specific time. On pull you have a ~150k burst dps with 500 ilvl during around 10-15 secs. After that, all your procs are falling off at about the same time, and that's the time you refresh with a instant soulswap. Doing it at the exact time you can easily extend your burst for another 15 to 20 secs, getting it to 200k with some good shard regen rng and haunt uptime.

Based on that, if you miss a dot refresh on that soulswap, you can lose a lot of dps, because there is no fix for it - the buffs are already gone.

Haunt is casted almost always on burst phases or during good procs. So you lose 2 secs to put a good debuff on a target. You miss a haunt, you missed 2 secs while under good procs, you miss a shard and you miss a 20% buff bonus on boss during a critic time. Also, since haunt has a big travel time, you will only know if he missed or not when it lands.


The only reason to not get hit rating is due to impossibity. Since it's not the case because we could get 600hit+600exp on trinkets since heroics, I don't see any reason to not get it.


Yes, it can give more dps on sometimes, but you are subject to RNG. And RNG is a bitch. At the right time, on the right pull, the boss will die, you only need to burst the add, you RNG bitches you, blam wipe.

[nightfalls]

Symer, you are just rambling at this point.

---------- Post added 2013-02-07 at 12:39 AM ----------

2- Playing affliction on this content it's all about refreshing dots on a very specific time. On pull you have a ~150k burst dps with 500 ilvl during around 10-15 secs. After that, all your procs are falling off at about the same time, and that's the time you refresh with a instant soulswap. Doing it at the exact time you can easily extend your burst for another 15 to 20 secs, getting it to 200k with some good shard regen rng and haunt uptime.

Based on that, if you miss a dot refresh on that soulswap, you can lose a lot of dps, because there is no fix for it - the buffs are already gone.

Haunt is casted almost always on burst phases or during good procs. So you lose 2 secs to put a good debuff on a target. You miss a haunt, you missed 2 secs while under good procs, you miss a shard and you miss a 20% buff bonus on boss during a critic time. Also, since haunt has a big travel time, you will only know if he missed or not when it lands.

This is a good argument. You're trading a very small amount of theoretical DPS for possibly fucking up these proc refreshes, and possibly having to refresh early to avoid bad RNG. In theory (Simcraft) it may be an average DPS gain, but in practice, I can't see how it possibly leads to significantly better performance even with best play.

[Deleted]

That's some very fancy math that is frankly beyond me, from what I understand, I was wrong about the SB:SS not missing, it does miss - but it can a) miss and refund a SS, or b) hit, then miss a dot application since it rolls for the 3. But if you're trying to argue that SB:SS never (or incredibly rarely) applies 2/3 dots then that strikes me as odd, happens to me semi-regularly at 14% hit.

If you're not, then I apologise.

[Deleted]

After that SimCraft also tells us that if we get Mastery or Haste too far from eachother we will start losing DPS.

Could you point me where in the thread you present arguments for this assumption? If you check my reforges: http://eu.battle.net/wow/en/characte...nowin/advanced
You'll see I'm going for way more mastery than haste. That is because every time I try to simulate getting haste and mastery even, the simulation gives me worse results. Though not very far from each other, but still a DPS loss.

[Bonkura]

Could you point me where in the thread you present arguments for this assumption? If you check my reforges: http://eu.battle.net/wow/en/characte...nowin/advanced
You'll see I'm going for way more mastery than haste. That is because every time I try to simulate getting haste and mastery even, the simulation gives me worse results. Though not very far from each other, but still a DPS loss.

I'm not sure if data have been posted in the thread, but the argument is based on the stats scaling with each other. We know that SimCraft sometimes shows either haste/mastery pulling ahead of the other by quite bit, but it has not yet been proven why. With this in mind it's a tricky topic, but if you've tried different values in SimCraft and found the best results noone is going to tell you that you are doing it wrong.

What results do you get if you grab the 6637 haste level compared to what you have now?

[Deleted]

I'm not sure if data have been posted in the thread, but the argument is based on the stats scaling with each other. We know that SimCraft sometimes shows either haste/mastery pulling ahead of the other by quite bit, but it has not yet been proven why. With this in mind it's a tricky topic, but if you've tried different values in SimCraft and found the best results noone is going to tell you that you are doing it wrong.

What results do you get if you grab the 6637 haste level compared to what you have now?

Can't reach it with my gear. Although, I can understand that the stats have a synergy between themselves, that seems quite intuitive, but I was wondering why wouldn't numbers back that up, in my case specifically.

[rezoacken]

In soviet Europe, you do not prove wrong, those people, who you do not understand.

It doesn't help them to make a point though.

I mean you're probably an intelligent person and we could have a good debate but I have no idea what you're saying (language, not the material) every other paragraph and therefore kind of don't bother reading big wall of texts written as such :/

Not mentioning I don't remember what your point is in the end.

[Kael]

Can't reach it with my gear. Although, I can understand that the stats have a synergy between themselves, that seems quite intuitive, but I was wondering why wouldn't numbers back that up, in my case specifically.

You can reach the next haste cap with gems alone. Switch the word "mastery" for "haste" on your sim profile for 2000 haste worth of gems and do an analysis from that haste reading. Aside - the haste proc trinket should be a small boost to the relative value of mastery, as more haste doesn't make a haste proc better, but more mastery does. I doubt this is "game changing" but the two should be damned close together anyway.

[Deleted]

You can reach the next haste cap with gems alone. Switch the word "mastery" for "haste" on your sim profile for 2000 haste worth of gems and do an analysis from that haste reading. Aside - the haste proc trinket should be a small boost to the relative value of mastery, as more haste doesn't make a haste proc better, but more mastery does. I doubt this is "game changing" but the two should be damned close together anyway.

Yes, of course, forgot about gems.
Anyway, re-gemming towards haste, being hit capped, with 6653 haste and 5464 mastery, opposed to 4722 haste and 7400 mastery, I simmed 20 less DPS, which falls under the margin of error.

[Deleted]

Ideas that are far fetched or of very low damage changes are not changing the std. dev. so much. Way it has changed so far from cap profile is that you suddenly have runs with +/-20 MG ticks around average, and a +/-1 million or smaller size Haunt events. Theese are based on miss chance % which is 3-8. Next there are some new (read previous stuff though) happenings with their own frequency, like a bad proc/dot sync has chance of 1/total_ticks, if the proc can come at any time. You multiply this with miss chance to have an event with the damage penalty. Also MG is clicked more often per minute than dots.

Actually one should start with the number describing how often the proc is popping in the fight. I will use my above sample from the logs.

Inner Brilliance procs: 9

Now at each proc the chance to have a bad situation, which first requires refreshing without a single miss is 1/tick_count. To have both the situation, and a miss, at one proc, has a probability of miss_chance/total_ticks, where I input numbers 0.05 / 11 = p, for Corruption.

Probability of having the complete damage loss event once with 9 procs is 9p(1-p)^8 = 0.0394. If you are not capping you have already accepted a similar thing to happen, because your total dps was not meant to be increased for so often as a number like this.

In comparison when casting a Corr 9 times, there is a probability of 0.314, to miss one or two. Theese average misses have the strength of a damage done by MG gcd. While proc misses also reduce damage of the next ticks of the DoT and MG.

You can also consider that the situation where an addon shows green light for refreshing, is like one proc. "Green light" has a duration and some proc rate. Number of "Green light" procs has some average number like other procs do in a single fight, only with more variance. Also when comparing hit cap and non cap, we would assume that the duration of "Green light" is more than one gcd.

There is a probability to have less than 1 gcd durations on green light. For example there can be two 20 s buffs that proc once per minute. Assume that we are in the middle of the fight already and that there is a procular chaos between theese two. Then if time is a discrete variable with chunks of 1 gcd, I would say one minute takes 48 of theese chunks, and both procs have same probability % to pop at any moment. Number of possibilities is 48^2 whereas the 1 gcd "Green light" means that if first proc is at 1, the other has to be at 16. I will also call it so that first proc at 48 means second proc has to be at 3, so that next minute of fight is actually considered. Basicly the proc has 1 min icd and, if one proc was actually at 3, and one at 48, then it is just as the first case. Makes no difference which proc plays the part of 1 and which 16, so the number of possibilities is 96. Probability to get one of those is 96/48^2 = 0.04166.

How many green lights could there be for three dots? Whenever one pops one should calculate all the 1/total_ticks rolls. If a "Green light" pops again without actual cooldown, it is more than one dot can handle, but all such impossible procs fit into [1/total_ticks, 1-1/total_ticks] interval, for sync purposes. Again one should just approximate the number of "Green light" procs and just sum the possibilities that arise with three separate dots. Error might again rise from requiring actually longer than 2 gcd lights, for overlapping situations. Since the hit cap profile also needs longer than 1 gcd periods, the extra damage losses (with one miss) only happen in cases, where duration of "Green Light" is "not less than 1 gcd and not more than 2 gcd's". In the above example this has the same probability as having less than 1 gcd "Green light".

But for the number of "Green lights", consider there are 4 procs with 1 min cooldown. One could approximate that after pull they are overlapped, and that at the end of the fight all are separate. Each buff proccing 9 times could produce 1+2+2+3+3+3+4+4+4 = 30 "Green lights".

Using 1/total_ticks with values (15, 11, 9), gives probabilities like "having one damage loss event". At 95 % hit, this is


30p(1-p)^29 * (1-q)^30 * (1-r)^30 + (1-p)^30 * 30q(1-q)^29*(1-r)^30 + (1-p)^30 * (1-q)^30*30r(1-r)^29 = 0.2703

when

p=0.05/15
q=0.05/11
r=0.05/9

Probability of "coming clear from one green light" is 1-p, 1-q and 1-r, for each dot. Then having this happen 30 times for all dots has probability

(1-p)^30 * (1-q) ^30 * (1-r) ^30 = 0.6677

Loosing damage more than once therefore has 6.204 % probability.


With 10 "Green lights" same numbers are only P("loose damage once") = 0.1180

P("no loosing") = 0.8740

and more than one event has a probability less than 1 %.
........

2- Playing affliction on this content it's all about refreshing dots on a very specific time. On pull you have a ~150k burst dps with 500 ilvl during around 10-15 secs. After that, all your procs are falling off at about the same time, and that's the time you refresh with a instant soulswap. Doing it at the exact time you can easily extend your burst for another 15 to 20 secs, getting it to 200k with some good shard regen rng and haunt uptime.

What kind of a pull is it that does 150k when procs are up, and 200k when procs are gone? And by "it" do you still mean burst dps that is local? Or is that your dps since the beginning, so that you are performing more than twice your "burst dps", without there being procs on filler/Haunt damage when compared?

Also procs are not such that they come to you at the same time, not even on a pull. You need rare rng to have them lined at the end of 15s. In worst case some procs can be dormant until this point where early procs start to disappear. Do you know any good pull rules that are reactive to proc situation, so you could find out next what kind of stats maximize the statisctical dps on first 30s?

Not that I care. With the default simc profile we know that hit cap profile has higher peak in its local dps. It does not necessarily mean that it has more average damage on first 30-40s, because you would have to integrate the local dps function over time, to find out. I studied instead what kind of effect the Bloodlust has on the dps curve.

http://img688.imageshack.us/img688/9596/bloodlustw.png

Default BiS 14 H with

gear_hit_rating=5100
gear_haste_rating=6637
gear_mastery_rating=8152

vs.

gear_hit_rating=3100
gear_haste_rating=6637
gear_mastery_rating=10152

In these profiles the highest dps still remains at the pull, and Drain Soul highest local value can be ~90 % of it. Difference between the max dps peeks at pull becomes shorter, between gear profiles. Total dps drops, and non cap gets futher ahead. This can be due to normaly having less casts performed under 20 %. Drain Soul channels can last 5-6 seconds (regardless of haste buffs, since interrupt depends on Haunt debuff), or more than twice as long a MGs. At the same time Drain Soul does not necessarily have a DPET that is twice as strong as MG has, so the miss events are not becoming stronger as fast as their number is reduced. Summing up, hit is worse under 20 % than above. if bosses try to wipe you under 20 %, instead of the pull, then your BL and gear choices might have priority in this phase.

About the pulls:

If non-capper plays so that his final SB:swap tries to catch all procs, but tries to do it one gcd in advance, before the capper would. Then the results are that he either misses first with miss_chance probability, and clicks again ending up with same strength on dots as the capper. He has one less gcd to use aswell (see below). Missing twice has a probability < 1 %. If he hits on the first try, he will have one gcd (1 sec with DS) less time, before his buffed dots run out. At lower than 50 % haste, gcd =< tick_time, so that if capper gets say n Corr ticks, which are fast ticking with haste buffs, then the non cap gets n-1, next the time to his n-th tick is tick_time_1 - tick_time_2 or around half seconds. After that both would get ticks at the same rate. His ticks are less than 1 tick behind away so far, and if they catch up with same tick count, non capper has 1 unbuffed tick less, but same tick count, per dot, when damage between spell power buffed and non buffed ticks is alot less than a tick with only constant stats.

I don't care, I think statistically the starting early and trusting for good rng are both giving same dps.

"Missing with anything" and loosing MG uptime is far more important than anything people write about fancier things. During first 30s there could be 3 SB:swaps/dots, 5 Haunts and 9 MG casts (for 22 s uptime with incomplete channels). If you miss with anything, the MG uptime goes down, and for most of that time MG can tick every ~ 0.6 second so one miss (gcd=1 s) is like two less Corruption ticks, for shards. MG does some damage too when it ticks though. And MG ticks even faster if there is both BL and DS. Well the number of misses out of 17 casts has a neat distribution with average of 0.88, when hit probability is 95 %. "Missing more than one spell" has a probability of ~20 %.

Getting full Haunt uptime, with for example five Haunts, on a pull has a probability of hit_chance^5. At 0.95 hit that is 77.37 %, and getting "four or five Haunts" has probability 97,74 %. However Haunt miss has the biggest damage penalty which I think was dealt with. Almost 1/5 of the pulls done in simc are loosing that kind of big damage on pull and yet only less than 1/20 pulls become worse when removing hit (in my correct material explanation).

Haunt is casted almost always on burst phases or during good procs. So you lose 2 secs to put a good debuff on a target. You miss a haunt, you missed 2 secs while under good procs, you miss a shard and you miss a 20% buff bonus on boss during a critic time. Also, since haunt has a big travel time, you will only know if he missed or not when it lands.

If you think Haunt can take 2 s to cast, and that casting it, is a waste, (as if, incase if you could get the debuff some other way, it would become a waste to get it with Haunt), then you are not aware of things in this game. (Otherwise you would talk about loosing seconds and not putting anything on a target, because it was a miss). Time, such as react_time + cast_time + travel_time - overlap_time, can take alot of values, also negative ones with 1 s gcd and 0 s travel time.

(If damage event is of size direct_damage + damage_from_debuff + filler_damage + shard, then shard=0. You can not list a "shard" as a source of damage as you listed also everything that a shard can do. To "miss debuff during critical time" is true if shards run out. If they do not run out then only "n secs is lost during good procs" is happening. If you meant MG would fill that time, it could not be n=2. If it is damage_from_debuff, then here n is anything)

You can compare different Haunts though and say that their debuffs are not necessarily overlapping completely with "good time". But either "Haunt is casted always during good times", and then there are probably more "good times" than there are Haunts. In this case every damage loss event is big and hit rating is relatively good. You can insist that there are too many Haunts for this and damage losses are more varied, but then the hit may not be so good. Player decisons effect the value of hit in the first case, because even if the "good time" is shorter than debuff_duration + n seconds you actually don't have to cast Haunts that are partially good. Which means that after missing a good Haunt there is no reason to fail more and cast a bad Haunt when it is too late, but to wait until better times.

If a haunt lands during good procs and you are holding 2-3k mastery because you are non capped, then your damage is larger than anyone elses. You will land that single Haunt 95 % of the time, if your hit % is 95. If it is a critical time on a boss then you brought more than one shard, to make the debuff down times small. If it is so critical that it is going to wipe you, then chain shoot two Haunts in the beginning, and don't wipe.

[Deleted]

When you can't integrate, you can always simulate the first half a minute, without Drain Soul, or have a target with target_fixed_health_percentage=100, and read the total damage from values of average dps.

Using the default pull you see how the high peak was misleading. I'm using again default BiS 14 H with

gear_hit_rating=5100
gear_haste_rating=6637
gear_mastery_rating=8152

vs.

gear_hit_rating=3100
gear_haste_rating=6637
gear_mastery_rating=10152

Then if you use 25s combat duration with "10 %" variation, you get theese results with different BL settings

no BL:

cap: 188094
no cap: 186736

yes BL:

cap: 231606
no cap: 231750

To explain that, I would understand that if you simulate to 40 seconds, then for default action list there would not be dots remaining from the DS+BL phase anyway, as the list would overwrite them and make the gear profiles always have similar dots. So we would not be looking at results of missing dots with any fight duration. Default action list still has use in 25 second fight, when this kind dot missing/hitting should not be affecting the results, and we only see what stats are good when DS (+BL) is up for the first dots, casted at the beginning of the fight, and then for the fillers. Therefore if non capping is better, when BL is turned on, it means that mastery is scaling very well with huge haste %. This has exceeded the dps loss from miss events, while having several Haunt casts and even gcd losses that hold max amout of MG ticks.

...

Simulating another pull for 35 s then. This action list casts one unbuffed SB:swap. Then it MGs atleast for 4 ticks and reacts when 2/3 procs are up to activate DS and other voluntary buffs (default BiS is an orc engineer). Another SB:swap comes then, and final SB:Swap is cast always when DS has less than 3 seconds remaining. This last one is not as dynamic as the second Swap, but atleast the trinket procs at the start of DS are up at that time, if they were to last 20 s. Delaying DS has little effect when fight only lasts 35 s. This way the simc mimics the dream pulls people think they are getting, instead of bad rng with the procs.

List is not optimized for non-capping in other ways than the last SB:Swap. In theory there can be no shards for last SB:Swap if all Haunts are missing, and at that point single dots are not actually used in a best possible way. Also the hit capped profile would not have to cast at the same time, 3 s before DS runs out, in game, but if this is changed to - 2 s then the profile is loosing dps. Some procs can remain there at -3 s, or the MG casts are preventing the spell completely because they are not interrupted. Results always revealed some single dot casts (0.0 average casts), among 50k runs, for both profiles, but the maximum fight length ended up with is ~48s which also requires UAs and Corrs.

Typical actions can be seen here:

FIKEWBCGJKWEWPKWEWW

where

K is SB:Swap
E is Haunt
G is DS
W is MG

And unluckly the non cap does more damage in both cases, while BL has reverse effect to that above:

no BL

cap: 181476
no cap: 183235

yes BL

cap: 223161
no cap: 223580

Looking at the BL runs there is still a higher peak, in local dps, for the cap profile: 285815 vs. 283339 (this tells you the pull is perhaps better than in the pictures above while I have not compared the total dpses with same fight durations )

Also Std. Devs for "no BL" are 7760 and 10664.

About theese dps differences, if a profile did 2k dps more in 40 s, then it only leads in damage by 80k, for that fight, which is like one good MG tick.

[Shodan30]

I'm not sure I get this. Are you saying achieve even level of stat ratings? Like paladin tanks during Cataclysm tried to keep dodge/parry ratings even?

Id like an answer to this to. When the guide says keep haste and mastery close, do they mean % or rating numbers?

[Bonkura]

Unbuffed stat rating.

[xskarma]

SNIP LOTS OF MATH STUFF

Symer, that post I quoted, what were you trying to prove with it? Not trolling, just not good enough in math and theory to follow what's going on very well. I'd love to thank you for the work you put into this, but I don't know what exactly you proved >.<

From the results at least I get the idea that hit capping or not hit capping, even at a crucial time at the start of the fight, doesn't actually make a whole lot of DPS difference. Was that what you were getting at? Cause if so, it means most of us were pretty much bending backwards agonizing over the hitcap at the start of the tier when it in fact doesn't look like it mattered all that much. Or is it possible that it mattered more at lower gear levels vs this BiS profile?

[Delvaz]

I am sitting at ~4300 haste and ~8600 mastery and could reforge mastery -> haste but would result in a dps loss according to simulation craft. Should I just ignore this and at least go for the second haste tick of 4717?

note, I was trying to link some images but says I need more posts..

[Bonkura]

Well, that's what the OP suggests, but if you're somehow able to understand any of Symer's posts you might find a different answer there. What most people tend to do at least is to grab that Corruption tick while maintaining hit cap, then they might stack mastery until they can keep it a fair bit above haste after have aquired the 6637 rating (which you won't be able to do until around 505 ilvl).

Seriously just do whatever you feel like but at least get 4717 haste. This topic is getting outdated with the release of 5.2 where we'll be looking at completely different stat levels, so it's not that interesting to keep discussing atm.

[Shodan30]

I am sitting at ~4300 haste and ~8600 mastery and could reforge mastery -> haste but would result in a dps loss according to simulation craft. Should I just ignore this and at least go for the second haste tick of 4717?

note, I was trying to link some images but says I need more posts..

I was doing the same thing, then rebalanced to hit the haste threshold (The 6630 or whatever it is), my dps did increase even with mastery going down about 18% buffed. Most of the reforge sites keep wanting you to go all mastery, but hitting the threshold really helped me. Of course this does make rebalancing new equipment a pain in the ass...

[Marilina]

am I only a lab rat now ;(?
im more curious about affliction p2 sha of fear performance http://www.worldoflogs.com/reports/x...=10656&e=11407

[nightfalls]

am I only a lab rat now ;(?
im more curious about affliction p2 sha of fear performance http://www.worldoflogs.com/reports/x...=10656&e=11407

Damage done Misses
147924818 0 (0.0 %)


You hit capped, according to Symer's explanations above that's a bad thing. Bad!