Marksmanship Hunter 8.3

[Valks 2,375]

This thread is for comments about our Marksmanship Hunter guide.

[Guest kittybleu]

Thanks for the updates ?

slight typo, is the argus life relic supposed to be ROOT of the lifebinder?

also, no hunter's mark or serpent sting in the opener? (is hunter's mark like its previous iteration, just marks target but doesn't aggro? so we should put it up before pull?) those were the recommended talents unless i read too fast....

thx again for getting us ready right off the bat ?

[Valks 2,375]

8 hours ago, Guest kittybleu said:

Thanks for the updates ?

slight typo, is the argus life relic supposed to be ROOT of the lifebinder?

also, no hunter's mark or serpent sting in the opener? (is hunter's mark like its previous iteration, just marks target but doesn't aggro? so we should put it up before pull?) those were the recommended talents unless i read too fast....

thx again for getting us ready right off the bat ?

Azor is fixing the BiS issue, it's just a problem with the item ID, yeah.

You can put up Hunter's Mark before fight and Serpent Sting will be somewhere after Rapid Fire - Azor is adding it.

[Kaedys 5]

I don't understand why you're recommending Serpent Sting over Murder.  In an ideal situation, the two are nearly equal on DPS.  But Serpent Sting also reduces the number of Arcane Shot casts per minute below the maximum number of Precise Shot charges you can get in the same window, causing waste of some of them on average.  In addition, at least last I checked (last week), Serpent Sting still did not scale with haste and was incapable of critting, reducing its DPS value even further.  For single-target sustained, Murder is the clear winner, at least until Serpent Sting is fixed (but even then, the Precise Shots clipping issue was reducing the net value of the talent by ~30% in my model).

I also don't understand why this guide, and hunter commentary in general, still insists that MM has "limited mobility".  MM mobility in 8.0 is definitively superior to 7.x, and is far more mobile than any caster specialization.  Aimed Shot is the only thing we can't use while moving, and we spend a maximum of 21% of our time casting it.  So 80% of the rotation can be done while moving, and the charge system on AiS gives us a ton of flexibility about when we cast it.  If we anticipate movement properly, we can generally handle movement without any DPS loss.  The only risk to DPS from movement is if the hunter in question isn't planning ahead for timers, or due to mechanics that both have unpredictable timing and require a quick reaction time to avoid (generally it's one or the other, though).

The recommendation to not cast RF above 70 focus (or 75, the guide is inconsistent) is a bit weird.  It generates 10 focus naturally, plus another 9 from passive regen over the duration, which should mean not casting it above 80.  Streamline increases the focus regen by 6, not by 3, because extending the duration also extends the period of passive regen during the channel, so that being recommended at 72 or lower is reasonable.

Lastly, I don't understand where your stat priorities are coming from.  My model is showing Mastery > Crit > Haste >= Vers, with Mastery ahead by a solid margin.  The issue with Haste is that, aside from AiS CD, the only thing it really does is increase our filler GCDs (Steady and unbonused Arcane), which are quite weak.  The fact that our highest DPCT ability (RF) doesn't scale with haste strongly limits the benefit of the stat.  Haste just doesn't scale as well as the guide asserts (hell, until they added allowed AiS's CD to scale with it, haste was worth roughly half the value of Mastery or Crit), nor does Vers, and the guide is severely underselling Mastery and Crit.  I'm assuming these weights are coming from SimC, which says to me that SimC isn't quite on the ball yet, since my formulaic model has been vetted and verified by several other theorycrafters over the course of Beta in the big ass thread I made on the main forums (and my model was fundamentally based on Adreaver's).

Edit: actually, could just be the inflated secondary stats from the prepatch.  You'd only need about 20% crit chance for versa to exceed crit.  I did my model at 120 using a pool of 3400 secondary stats (my guesstimate of available secondary stats at the Heroic ilevel in the first raid tier), which provides a significantly lower available crit chance.  I still don't buy the haste recommendation, though.

Edited July 17, 2018 by Kaedis

[Valks 2,375]

1 hour ago, Kaedis said:

I don't understand why you're recommending Serpent Sting over Murder. 

Few bits of feedback from Azor concerning what you said:

• SS vs Crows is very close and with recent in-game and SimC changes, Crows is very slightly ahead due to SS having a negative impact on the rotation. He'll update this shortly to reflect this change.
• In terms of RF, it's mainly just a safety-margin choice, in order to allow for players that will not always be 100% perfect in their casts and timings. He'll look into fixing the inconsistency issue, but the Focus point is simply a preference of being slightly lower to allow for more lenience, rather than being slightly too high.
• On the stats point, things are changing currently and Azor has an update pending to change the priority somewhat. Crit seems to not be showing much value in his testing so far, but I'll wait before he puts up the update before saying exactly what he is putting the priority as (I don't actually know it).
• The guide is specifically aimed at the pre-patch, so using an approximate pool of stats for BfA tiers isn't going to help much for the information here, since it will be updated in the future when the BfA tiers are being raided. The inflated secondaries here are definitely not to be ignored in the pre-patch, since that's what we're aiming to give a guide for.

We definitely appreciate the feedback, it's great to see people putting so much effort into it, so thanks for that. Hope this answers all questions/worries.

[Azortharion 60]

Just wanted to note that Marksmanship's Serpent Sting DOES crit, not just the initial hit but the DoT as well.

It is possible that it crits less often, you never know with Blizzard and flagging AP and "Spells" differently for many things, but it does crit.

[Kaedys 5]

Fair enough on the pre-patch part.  Nearly all of my theorycrafting has been aimed (lol) at 120 and the stats available in the first raid tier.  The massively inflated secondary stats available during prepatch are very likely to muck with that.  I do wonder, however, how much of the stat weights suggestions are based upon the simmed weights at a particular disparate distribution of stats.  Example, if your sim profile has, say, 14k mastery (or whatever that translates to after the stat and ilevel squish) and only 4k haste, then ya, haste is going to gain a lot of ground on that due to the fundamental nature of how marginal benefits work in an additive system (ie. 100->101 is a larger percentage increase than 190->191). 

I feel like we should be, if we're not already, generating these stat recommendations based on a equal split of stats (which is typically how I've done it in my model).  Since these stat weights are generally only used by players that aren't super serious about optimizing (serious players use Raidbots, or SimC directly, and regularly re-sim their weights), they should reflect the general priority of those stats, and the only way to generate values like that are if the sim profile in question has an equal split of stats to avoid the marginal benefits issue.

Did Azortharion mention or acknowledge anything with regards to the Serpent Sting secondary stat scaling issue?  I'll retest it tonight when I get on, but if it still isn't scaling with crit or haste, it's going to be way behind Murder.

Edit: ninja'd.  Ok, good that they fixed that.  It was a complete and total lack of ability to crit (beyond the initial hit) previously, so if we have confirmed that the DoT can crit, that's at least one of the two stat issues fixed.  I'll see if I can get some hard numbers on haste for it when I get home tonight.

Edited July 17, 2018 by Kaedis

[Azortharion 60]

Whenever I run stat weights, I do so with all the stats normalized. In this case, I add up the total secondary pool, divide by 4, and then run the stat weights with an equal amount of each stat rating.

This is what I was referring to when I mentioned relatively high Vers. Generally speaking, even when Vers is a good stat for you, it'll be a pretty low one as it is rare on gear. So in a more realistic/organic stat distribution, it is likely to be just as good or even better.

I have done this method for many years to try and get around the fact that the lowest stat in a distribution will often appear to be the better one. It's not perfect, and I have toyed with other approaches like a 1:1:1:0.5 ratio (with Vers being the lowest) instead. At that point, the differences are minute.

I acknowledge the lack of Haste scaling, but not the Crit one. As mentioned in my post (we probably posted simultaneously), it does Crit. It does not scale with Haste in SimCraft either, so there is not a discrepancy as such between SimC and ingame, though I will be the first to acknowledge that SimC isn't end-all-be-all. It does take into account that SS takes Focus out of your rotation, this should be said. With that out of the way, the people working on SimCraft's Hunter Modules (NuoHep and Jayeasy) are very intelligent people and dilligent about making sure things line up with how it is in-game. This does not make SimC infallible, there are still some question marks regarding certain DoT's (including but not limited, to Serpent Sting).

Crows is the better talent. While there is little between them in raw DPS, Crows is easier to use, and easier to maximize when circumstances permit it (add spawns..), and it has less of a rotational impact generally. This makes it a desirable choice, ease of use is not a factor to be ignored even when we try to maximize every little bit.

[Kaedys 5]

But isn't the point of these stat priority lists to list out the stats in roughly descending order of how much of each of them you should have on your gear?  Like, listing Vers over Crit, I at least read as the guide saying that we should, ideally, have more Vers rating than Crit rating, when that's not actually the case.

As for the SS disparity, it could be pre-patch.  I only tested my model at 120 for 0 and 3400 total secondary stats (equal distribution), which gives a peak of 12.5% haste.  Significantly increased haste due to pre-patch inflation could easily raise the number of filler Arcane Shots above the maximum number of Precise Shots charges, negating the clipping issue.  I am, however, curious if SimC is seeing the same benefit reduction due to clipping that my model show, when simmed at 120 in first-raid-tier gear.

[Azortharion 60]

If you set Crit and Versatility to the same amount in a sim and run your stat weights (in my BiS Prepatch Gearset as a baseline), then Vers will simply give you more bang for your buck. This indicates to me that if you could, you would indeed have more Versatility than Crit on your gear. That does not mean that we are necessarily able to. 

Generally, realistic item decisions are almost never made based on what is the 2nd, 3rd or 4th best stat. Either way you slice it, it is very close.

I have not personally toyed much with level 120 things so I cannot offer you a good answer on that one. If the math is correct, then I see no reason SimC wouldn't be modeling this, however.

I just tested Serpent Sting MM for Haste scaling ingame, and I am seeing an increased tickrate (therefore, more ticks per cast) from more Haste.

[Kaedys 5]

20 minutes ago, Azortharion said:

If you set Crit and Versatility to the same amount in a sim and run your stat weights (in my BiS Prepatch Gearset as a baseline), then Vers will simply give you more bang for your buck. This indicates to me that if you could, you would indeed have more Versatility than Crit on your gear. That does not mean that we are necessarily able to. 

I don't get how this is even remotely possible.  If all of your DPS scales with both crit and versatility, you'd need to have ~20% more (multiplicative) crit than versatility, as a percent, in order for them to break even.  My testing on beta placed the stat rating conversions at 17.625 crit per 1% and 21.15 vers per 1% at 110 (72 and 85, respectively, for 120).  If you had, for example, 500 of each, you'd have 28.37% crit from rating, plus 10% baseline, for 38.37% crit, and 23.64% vers.  Adding another 100 to either would raise crit by 5.67% to a new total of 44.04%, or versa by 4.73% to a new total of 28.37%.  The DPS gain from the addition of those stats ( (1+new) / (1+old) ) are 4.10% for crit, and 3.83% for Vers.  That still clearly has crit in the lead.

Even if we scale the numbers to absurd levels, this remains.  Example, if we had 1000 of each, we'd have a 66.74% chance to crit and 47.28%.  Bumping either by 100 again raises that to 72.41% crit or 52.01% versa, for a gain of 3.40% for crit and 3.21% versa.

Am I missing something here?  If we're setting them to equal rating, it should be mathematically impossible for versa to overtake crit simply due to the different conversion ratios, at least at any reasonable secondary stat pool value.  And if Serpent Sting can crit, all of our abilities scale identically with both crit and versa, so they can both be treated effectively as simply multiplicative damage bonuses.

Edited July 17, 2018 by Kaedis

[Azortharion 60]

I am getting 20.82 Vers per 1%, and about 17.65 Crit per 1%. Maybe I am missing some decimals that you are not, but this appears to be the case for me (not that it really makes a difference). RaidBots (what I use as the basis for most of my SimC work) is currently down for maintenance, otherwise I'd link directly to a sim showcasing what I describe.

Do you not think that Careful Aim providing a significant Critical Strike bonus to Rapid Fire and Aimed Shot is the culprit here? On the other hand, Careful Aim essentially buffs Crit Damage too, which would turn in Crit's favour.

Here is an older sim I did which is imperfect but probably good enough for now: https://www.raidbots.com/simbot/report/6363XKvL2ziV6fBmL8Wb9f

It has 338 Crit and 338 Vers, or 29.18% and 16.25% respectively.

[Azortharion 60]

Jayeasy reminded me that Lethal Shots also has this detrimental effect on Crit's value. Have you taken this and CA into account in your model? CA might not make a huge difference, but Lethal Shots will.

[Kaedys 5]

Careful Aim doesn't increase crit chance, though.  It provides a random damage increase, which is just a raw multiplier.  It shouldn't have any impact on the relative value of crit or versatility.

That said, Lethal Shots definitely could.  I haven't mucked with stat scaling on the context of talents much, because talent tuning and mechanics have been in flux so much.  A Lethal Shots build I could definitely see working that way.

Mastery still baffles me as well.  It has the same base conversion ratio as crit (1 Mastery per 17.6xx rating), but 1 Mastery for MM grants a 0.625% increase in shot range and a 1.4% increase to shot damage, and we start with 5% and 11.2%, respectively.  By the exact same logic as crit, mastery should be way ahead of versatility at the same rating level.  And unlike crit, talents don't affect mastery scaling at all, though Lone Wolf vs pet can, as pets don't scale with mastery.  Still, it takes 65.3% more versatility rating to provide the same nominal increase to DPS, and the base amount of mastery and the pet aren't nearly sufficient to negate that.

I'm not really saying that the sims aren't saying what you are asserting they are saying.  What I'm saying is, if the sims are saying that, there's something wrong with the sims, because they are asserting something that directly contradicts a mathematical analysis of the same system.  It's also worth noting that I'm pretty sure the sim you linked was before they changed the mastery, and it appears that Serpent Sting was not scaling with haste at that point.

Also, I dug into the SimC source code to resolve the conversion ratios.  They have them assigned out to 8 decimal places, and I assume that are the result of data dumps from the game files.  For 110, those conversion ratios are:

Crit: 17.62282985 per 1%
Haste: 16.64378375 per 1%
Mastery: 17.62282985 per 1 (ie. per 0.625% range and 1.4% damage)
Versatility: 20.80472968 per 1% damage / 0.5% DR

Reference: https://github.com/simulationcraft/simc/blob/78b35c0bdcef84a2050dd6ef8d210f9be8ecab52/engine/dbc/generated/sc_scale_data.inc#L1117

The conversion ratios previously referenced for 120 (72, 68, 72, and 85, respectively) are accurate and are actually integers.

Lastly, I forgot to include this in my original commentary, but it might be worth noting in the Steady Focus talent description that it's currently basically worthless because it doesn't affect the GCD on Steady Shot.  The first stack drops the cast time to 1.4s, and the second to 1.05s, but the GCD for both is still 1.5s (with haste reducing the cast time and GCD at the same rate).  Until that's fixed, Steady Focus literally has zero chance of being viable.

Edited July 17, 2018 by Kaedis

[Putro 1]

Steady Shot GCD is affected by Steady Focus. 

Do you by happenstance have any link to your formulaic model and it being verified and vetted? I'm interested in seeing this. 

[Kaedys 5]

Just now, Putro said:

Steady Shot GCD is affected by Steady Focus. 

Do you by happenstance have any link to your formulaic model and it being verified and vetted? I'm interested in seeing this. 

That's new, it definitely was not affected by Steady Focus as of ~10 days ago.

As for the model, the comparison to other models is buried in >100 pages between two threads (this one and this one, the latter being the one I started).  I'll see if I can get the spreadsheet cleaned up and uploaded to google docs tonight for y'all.

[Putro 1]

Scrolling through those threads, I see no theorycrafters that I can recognize verifying or vetting anything. 

On topic of stats:
Have you remembered to account for enchants, trinkets and pet being affected by vers and not mastery in your comparisons? These things combined deal a significant amount of damage, in azors sim they are dealing 20.6% of the total damage. 

[Kaedys 5]

2 hours ago, Putro said:

On topic of stats:
Have you remembered to account for enchants, trinkets and pet being affected by vers and not mastery in your comparisons? These things combined deal a significant amount of damage, in azors sim they are dealing 20.6% of the total damage. 

Trinkets and enchants, no, I hadn't included.  But even that still doesn't help.  Say, for example, 500 of each stat.  That gives us 24.03% versatility or 50.92% mastery (including the base 11.2%).  Adding another 100 of either bumps that to 28.84% vers or 58.87% mastery.  If vers affects 100% of damage and mastery affects 79.4% of damage, then adding 100 vers increased damage by 3.88%.  Adding 100 mastery increased it by 4.49%.  So even in that situation, mastery is still clearly in the lead.  At least 35% of your damage would have to be coming from sources that scale with Vers but not Mastery for that gap to close.

Models and vetting and everything aside, can anyone explain to me how it's even mathematically possible for versatility to be ahead of mastery in any achievable gearset, if we take as a caveat the stat normalization Azor uses (ie. pool all secondary stats, set each equal to 1/4th of total pool)?

Edited July 18, 2018 by Kaedys

[Guest Squizzlemadizzle]

just wanna point out u have 2 completely different stat priorities mentioned in ure guide. haste, vers, crit, mastery under "easy mode" and mastery, vers, haste, crit under "stat priority".

 

since aimed shot recharge isnt affected by haste i guess that the mastery, vers version is the more likely one?

[Azortharion 60]

1 hour ago, Guest Squizzlemadizzle said:

just wanna point out u have 2 completely different stat priorities mentioned in ure guide. haste, vers, crit, mastery under "easy mode" and mastery, vers, haste, crit under "stat priority".

 

since aimed shot recharge isnt affected by haste i guess that the mastery, vers version is the more likely one?

Aimed Shot's recharge IS affected by Haste. That said, the one found in the "Stat Priority" is the correct one and the Easy Mode discrepancy will be fixed!

[Putro 1]

Versatility then baseline ‘buffing’ the three other stats could be pushing it further ahead. This is also accounted in stat weights. What you’re doing is comparing stats against one another in a vacuum, when in fact they feed into one another - this is significantly harder to simply show in simple calculations in a spreadsheet. One of the reasons SimC is a great tool is this fact. 

I, and others, will continously take further looks into SimC over the course of pre-patch and BFA. But it seems to be accurate when comparing in-game numbers to what the sims are getting in terms of damage, casts and so on. This leads me to believe the fault is in not accounting for trinkets, enchants and interaction with other stats in your calculations. 

 

As a spreadsheet geek, I am still interested in seeing the spreadsheet even if it is messy! If you’re not comfortable with posting it here, you can always PM it to me on Discord Putro#6093. 

[Kaedys 5]

14 minutes ago, Putro said:

Versatility then baseline ‘buffing’ the three other stats could be pushing it further ahead. This is also accounted in stat weights.

Versatility doesn't buff other stats.  Versatility is a straight multiplicative increase to all damage done.  While it will increase the absolute DPS gained from other stats, it does not increase the percentage gained from other stats, and thus does not affect relative stat weights, because its increasing base damage and the absolute damage gain from each stat by exactly the same multiplier.

ie. if you're doing 1000 DPS, and X crit increases that by 10, then adding 10% versatility will increase your base DPS to 1100 and your gain from crit to 11.  The percentage gain from that crit is still the same, 1%.  It's a completely independent multiplier.

I'll shoot you it in a PM, but be warned, I've made no attempt yet to clean up the sheet style.

[Putro 1]

But stat weights aren’t relative... They show absolute DPS gain per 1 stat.

Also JayEasy reminded me that auto-shots are also affected by vers but not by mastery pushing us to a higher number than 20.6%, without access to SimC right now, probably 30+%. 

 

At the end of the day, I think you’re trying to do some very simple math in a setting where said math isn’t enough, because it’s disregarding tons of factors that SimC accounts for. (This isn’t meant to imply SimC is perfect, and as I said several people are hard at work improving it in multiple aspects to ensure it is as accurate as possible and prepatch may have introduced weird things. With that said it does seem pretty accurate still.) 

[Kaedys 5]

5 minutes ago, Putro said:

But stat weights aren’t relative... They show absolute DPS gain per 1 stat.

Also JayEasy reminded me that auto-shots are also affected by vers but not by mastery pushing us to a higher number than 20.6%, without access to SimC right now, probably 30+%. 

Auto-shots are a fair point, and that pushes the total damage not affected by mastery to right at the point where it might equal vers, but doubtful below.

And yes, stat weights are, in general, annotated as absolutes.  But adding more of both stats causes they effect on both (from the other), and thus cancels out, or mostly does.  Stats with a non-zero base value see a very small reduction in relative weight compared to the other stat due to increased additive effects.  This effect, however, is very small.  The only way stats really overtake each other in relative value is if there are breakpoints or softcaps, or if you're adding a really large amount of the stats, or if you're not at equal levels of the stats.  ie. if you keep adding crit, then yes, eventually versa will overtake it due to the diminishing marginal gains from ratings being additive.  Adding versa can do nothing but reduce the value of further versa, though.

Basically, while versa increases the absolute DPS per point of, say, mastery, mastery also increases the absolute DPS per point of versa.  And by in large, those two effects negate each other unless you're dealing with extremely large (ie. irrelevant magnitudes, given gear limitations) amounts of stats.

I mean, ya, I get it.  A formulaic approach can never quite equal SimC's sheer probabilistic weight.  But that doesn't mean that a formulaic approach is necessarily worthless or irrelevant, and that formulaic approach can provide a good check against SimC.  I mean, SimC is still fairly early in its overall development and vetting cycle for the expansion, so the probability of there being a bug sufficiently large to cause erroneous stat weights is rather substantially larger than it would be in, say, a year, or even six months.

[Kaedys 5]

Unrelated critique of something else in the guide:

Quote

Soul of the Huntmaster is a Legendary ring that gives you the Lock and Load talent without having to choose it, meaning the value of its effects is the same as the value of this talent.

Since Lock and Load is considered the best talent on that tier, so you'd be taking it anyway, wouldn't the value of this legendary affix be the value of whichever of the other two talents it let you pick up?  And since neither are terribly thrilling, that would make this ring not terribly thrilling either.

The ring is appropriately very low in the legendary ranking lists that follow this statement, but the statement itself is a bit misleading.