This is exactly the kind of thing I was going to do, but haven't had the time.
I'd like to see some more fine-tuned results above 1600 AP -- stepping AP up by 100s and crit and hit by 1%. I'm not that well geared, but am above 2000 with battleshout/might/kings, and trinkets.
Also, on quick glance, I'm not sure the "dps from crit" approach (rightmost columns) is accurate. Wouldn't it be more straightforward to start with a baseline, such as say [1600AP, 20% hit, 20% crit], then increment each stat separately from there and measure the difference for each?
Also, I understand the results best when back calculated to rating. Your results would be 34 AP = 22 crit = 32 hit, or 1.5 ap ~ 1 crit ~ 1.5 hit.
I'd also convert that into stat-point-efficiency of 0.75 str/apvalue ~ 1 critvalue ~ 1.5 hitvalue. I'll be surprised if this holds true much above 1800 AP, where I think AP efficiency will fall behind.
I'm really hoping that WF in 2.1 acts like dual WF5/WF5 does now - 36% chance on hit after the 3 second rule. With 2.6/2.5 weapons, i got about an 18% "overall" proc rate on live that way.
Isn't it great how even that hack fix doesn't reach the 20% on the tooltip at existing weapon speeds? There are no decent weapons slower than 2.6, after all. I certainly don't "hope" it works that way, although I suppose it's better than 20% with a linked cooldown. Such lazy design.
This is exactly the kind of thing I was going to do, but haven't had the time.
as an aside, i've called my dad Pater since i first started taking latin in 8th grade. 18 years later, still call him that.
I'd like to see some more fine-tuned results above 1600 AP -- stepping AP up by 100s and crit and hit by 1%. I'm not that well geared, but am above 2000 with battleshout/might/kings, and trinkets.
while this is doable, it's probably best done with a specific goal in mind - is it to see the relative value of different attributes? What in particular would you be interested in seeing, so i can provide the relevant info?
Also, on quick glance, I'm not sure the "dps from crit" approach (rightmost columns) is accurate. Wouldn't it be more straightforward to start with a baseline, such as say [1600AP, 20% hit, 20% crit], then increment each stat separately from there and measure the difference for each?
what i was looking for were scaling trends. So, that dps from crit column is looking at "everything remaining constant, only crit changing".
So, one of the rows might be 800 ap, 16% hit, 14% crit. That would be compared to 800 ap, 16% hit, 18% crit. It really is holding all comparisons as parameters where the only changing variable i crit.
Is it possible to - rather than seeing how AP, Crit and Hit affect DPS - fix those values, and then try testing the effect of different weapon speeds. Say, from 1.6s to 2.7s, in 0.1s increments, and see which combination(s) provide the best return. Given the shared cooldown coming in, slow MH and equal or slower OH seems to be the best option, but it would be interesting to see just how dps varies with the different speeds.
Pater - because the results will largely be the same, i'm using what you suggested, but running in 200 AP increments and 2% crit/hit increments. we're still going to have a boatload of data points.
Vaeys - Ok, i just redid the script so I can loop through weapon speeds in addition to ap/hit/crit. When pater's simulation is done running, I'll run one for you using the weapon "dirge" as a baseline (1.4 speed, 83 dps), and then scaling up the weapon speed (and dmg per swing), leaving the others alone. Im planning on using 2000 AP, 20% hit, 26% crit.
From what I can see, your test runs caps out at 1600 AP. That's much lower than I'm hitting with full raid buffs and UR proc'd, and I'm not very well geared. I believe that crit and hit really start to take off above 1800 or so.
AP from 1500 to 2500, 200 ap increments
Hit from 18 to 26, 2% increments
Crit from 18 to 36, 2% increments
As expected, by doing such small steps, the vagaries of combat sims make the standard deviations a larger part of the average, but the averages themselves are pretty good, and we have a pretty interesting result, which is relatively consistent with expectations and the last run.
according to these sims, the following have equivalent dps impacts:
1% crit = 37AP = 2.3% hit
(to compare with the lower-end gear, where 1% crit = 34AP = 2% hit)
For each 10% you increase OH weapon speed, expect a 2% increase in your dps.
For each 10% you increase MH weapon speed, expect a 3% increase in your dps.
This works as expected - going from a 1.4 to a 2.0 OH is a 42% increase in speed. That will cut down on eaten MH windfuries more than going from 2.0 to 2.6, a 30% increase in speed. If i presented the numbers in that .htm file as absolute weapon speed, it wouldn't really tell the story that moving out from 1.4 to 1.6 is much more valuable than moving from 2.4 to 2.6. If that is more meaningful to you, you can certainly copy/paste the file into excel and do the formulas accordingly.
So, how much do weapon speeds affect this? Lets compare the rule of thumb formula to the actual sim results. going from 1.4 MH/OH to 2.8 MH/OH should =
base 1.4/1.4 dps * (120%) * (130%) = 1.56.
My 1.4/1.4 dps was 621.
so, 2.8/2.8 should be 621*1.56 = 968
the tested dps result from the sim, as you can see in the htm sheet, is 930 (which is a 1.50% increase)
In my mind, for being such a huge swing of weapon speeds, and using "rules of thumb" based on iterative data, that's pretty darn close.
next up... raiding dps numbers, tied together: how much of each stat = 1% increase in dps.
For raiding numbers you may want to remove parry/block. I don't know if it would make much of a difference but they can be avoided the vast majority of the time.
yeah, i was thinking about removing them, keeping them in, or putting them at 1/2 their normal rate. If we get wide acceptance that with them in, these numbers seem accurate and real life raiding dps increases approximate the ones offered here (in terms of stat equivalencies), then i can re-run the whole lot of them, probably doing like parry/block/dodge at a combined 10% or so (since dodge will always exist).
Ok, so on to the equivalencies part...
Looking back at the raiding sheet, we see that:
1.90% hit =
0.81% crit =
30.3 AP =
1% DPS increase
Note: on the sheet it was normalized to 1 AP equivalency for the pawn-strings, this is using the raw %dps increase per AP or % crit.
From the speed test we know that:
5% increase in OH Weapon Speed =
3% increase in MH Weapon Speed =
1% DPS increase.
My only concern at the present time ties in to crit.
a less than 1% increase in Crit chance gives a greater than 1% increase in DPS. This, while definitely possible, seems a bit weird. Anyone have thoughts to either completely discredit this result, or support it?
Major factor supporting it:
- More crit = more flurry and more UR. For UR, going from 18% crit to 30% crit brings a 10% increase in UR uptime.
- Since flurry gets "eaten" on non-crits, this scales pretty well with crit, though of course there will be overlap when a crit happens when it would already be flurried.
I apologize for the slight de-rail, but I have a quick enhancement shaman question which I don't think is worthy of its own thread, and which I'm guessing the majority of the people posting here can answer very quickly.
We currently have an enhancement Shaman who uses a 2-handed weapon, Singing Crystal Axe (287-431 dmg, 3.5 speed, Chance on hit for 400 haste rating for 10 Seconds), and hoping for the 2H axe off Prince. His DPS is respectable, enough that I don't feel gimped bringing him to raids, but its not anything special, he usually ends up below the top cluster of DPS on each raid. He is no more or less geared than the rest of the raid, in general.
The conventional wisdom I'm seeing is that DW blows 2H out of the water when it comes to PvE DPS. To the point that its just assumed when discussing a Enh Shaman that they are using DW.
I'd like to talk to him about this 2H choice, and get him to working w/ DW, but I frankly have little to no experience or understanding of Shaman Melee dps, and whether I'm right or not. Am I correct in understanding that DW is pretty much always the right choice for PvE DPS (don't care about PvP or anything).
For reference, this is a guild w/ Gruul and Kara on farm, working on Magtheridon.
Two handed weapons would be a lot more viable if the Windfury cooldown wasn't around. It's quite literally impossible to find a 2H weapon that will not have a dead swing when flurried, and Unleashed Rage and Shamanistic Rage both suck without a high volume of attacks. It's possible to have a 0/31/30 build with respectable DPS that offers a good deal more healing power than a pure enhancement build, but it's a pretty poor raid spec.
commentary - hit becomes less important as you stack it, compared to AP, whereas crit becomes more important.
Hit is useful until you cap out hit rating somewhere north of 250, because if you don't hit you don't proc SR and you don't proc WF. I know the latter is taken in to account in the simulation, but it's more important than raw DPS numbers might indicate.
Yah, I hadnt thought about the SR aspect of it nite_moogle, that's a good point.
as far as the 2H vs dual wield, I ran the sim comparing: singing crystal axe:
2000 AP, 0% miss, 26% crit. DPS = 588
compared to: Planar Edge and Runic Hammer:
1750 AP, 5% miss, 22% crit. DPS = 663
I didnt have it model the haste proc. But, we can do the following to fudge it... white damage was 304 dps of the axe. White damage will be the primary reason the 40% haste will be better. Yes, there's some room for UR to be better (only up 78% of the time with the 2H compared to 95% of the time with dual wield), but for simplicity's sake, let's approximate how much time the haste proc would have to be active in order for the white damage to make up for that gap...
304 * .4 * x = (663-588)
121x = 75
x = .62.
So... If the haste proc is up 62% of the time, then the axe could be seen as a better weapon.
Oh, and notice that in the sims I gave the axe a 250AP and 4% crit advantage (essentially saying the dual-wield build has to spend those points on hit rating).
SR is such a huge problem w/2h's, it cannot be underestimated, the points past 31 in enhance are just completely wasted in a 2h build... SR will do virtually nothing and UR's uptime will be pretty poor. SR is really important on any boss fight, it prevents you from being totally out of mana 3 minutes in.
Great work in general BTW Disquette, thanks very much for your contribution.
Overall, 2h shaman dps is slightly worse in gimpitude than 2h fury warriors; unlike warriors, we have flat chance-to-proc-on-hit abilities which are amplified by frequent attacks.
I would say that attack power's greater contribution to dualwielding is what really makes it win flat out. UR/SR uptime is worth noting but pales in comparison. With two weapons you can get two procs, two sets of different stats, two sets of sockets, and two enchants. Dualwielding gives you one more stat to improve your character in melee hit, which isn't automatically 100% when dualwielding.
Also as already noted, in order to waste 0 swings in windfury cooldown with flurry active, you need a 4.0s 2hander, which don't exist. One fine point there is that you waste more time unable to windfury with a weapon that's slower than 2.0s but faster than 4.0s without flurry, because you have to wait for the weapon's speed to expire before the next attack as well. And with only a single weapon there's no chance your offhand will proc WF during weapon cooldown; it's flat out wasted.
I'd hope that any review we eventually get would pay attention to 2H weapons. My enhance talent mockup from a couple of weeks ago is one direction.
I'd hope that any review we eventually get would pay attention to 2H weapons. My enhance talent mockup from a couple of weeks ago is one direction.
I'd seen that talent build that after someone mentioned your name on the bliz forums, and I looked it up to see if "you were you". When i went through your post history I saw that. I was rather impressed with the way that you made 2H an attractive option.
by the way, whereas Dual Wield wins out for both raw white and windfury dps, it's interesting (and makes sense, i suppose) that 2H out dps's dual wielding if you look strictly at stormstrike damage. I think this is in large part because of missing/parry/block/dodge on the first attack cancels the second.
And yes, the enchant situation is a large part of what makes dual wielding attractive as well. Having an additional mongoose or crusader seems pretty nice.
Check my math. I focused on a value range near where I am:
2100 AP, 22% hit, 20% crit = 5.13e6 dmg
+200 AP = 5.28e6 dmg = +2.88% dps
+2% crit = 5.27e6 dmg = +2.76% dps
Efficiency comparison for itemization purposes: at my level, to boost dps by 1%:
- it costs 200/2/2.88 or 34.7 item points invested in AP
- it costs 2*22.1/2.76 or 16.0 item points invested in crit.
This appears to say that at this level, crit is almost twice as efficient (item point for item point) at increasing my dps than ap. (This matches what I expected the results to be.)
Turning to hit, I find that increasing hit rating [2100,22h,20c] --> [2100,24h,20c] actually reduced damage. I don't see any way that this can be accurate. It's either a hit table bug or a result of wide distributions. If the results are really that varied that you can have two samples where more hit results in less dps, I would want to run several simulations at each data point and take the mean result (though I don't see why this would have any different result than just increasing simulation length).
Yeah, +hit acts kinda weird, and because it has such a small effect (based on what I've seen), it's easy to get upside-down answers. One thing I found amusing when doing testing on blasted lands mobs, where variables were kept to a minimum, was how much dps could fluctuate just watching recap. It would go up or down 1.5% regularly. If I could have glances and misses in that same scenario, I'm guessing lucky and unlucky strings would have swung it even wider in local time periods, though it would even out.
One of the reasons that I did so many row comparisons is to minimize the overall impact of wide distributions. It effectively averaged together a bunch of widely scattered points. You'll notice on those tables (the htm's) that there are some negatives interspersed with the positives you'd expect by increasing your hit value.
Anyway, the image below has results from a pair of 20 hour simulated tests, with the 2100,22h,20 and 2100,24h,20 figures you used. Notice how even in a long test like this, the %s can be so far from the expected values, as with this miss portion on the 22% hit test. I dont know what to say other than maybe the number randomizer sucks. In general, the hit/crit/etc stuff looks about right. I've included the command-prompt output from my current version of the app so you can see where I got my numbers...
again, this is why i did so many row comparisons, because the variability is pretty high even in "long" tests. Maybe I'll set it to run for 100 hours for kicks, to see how that turns out...
As a side note, I thought the dodge rate of bosses (from behind) was 15%, not 10%. Do you have a good link for the 10% value? (I don't have a link--just some vague memory of reading that although mobs can't parry or block when you attack from behind, their dodge rate is 15%.)
Hey, I'm the enhancement shaman Orestus was referring to. I ran with a 31/30 spec (2 hander with stormstrike and 5/5 purification) for a while, all the way up through Gruul (which you can see in action here). I liked that spec a LOT and had fun with it, but it wasn't cutting it for dps in a 25 man raid. As the flexibility of benig able to swap to healing gear was no longer a desirable aspect to have.
After that fight, I respecced to my current build of 20/41 ele/enhance with 2 hander, UR and SR, with elemental talents purely for shock efficiency. It increased my average DPS by about 150-200 just from being able to consistantly shock throughout a fight(base damage ES is still about 700/5 secs non crit), more if you count having UR up.
The main things I wanted to try with this build were being able to ignore most +hit (although I missed a bit too much on Gruul for comfort, so am pushing for the +5% cap now) in favor of other stats like stam and attack power. I'm fairly happy with it, and still want to see how getting a BoP two hander will help(slowly working towards arena two hander as it blows the current dropped ones out of the water). As my current axe has no stats, and is probably worth about 115 dps baseline with the proc frequency.
I am however going to build up a +hit set so I can respec DW for a few weeks and see how it stacks up. I haven't ran into another decently geared DW shaman on my server to compare numbers with yet to see objectively. And nobody that I know would even give using a two hander a shot with a spec/gear setup for using it properly to see the comparison.
One point I did want to make though was that SR scales fine with a two hander as it's proc per minute. It procs on almost every swing of the two hadner while it's up, and I've seen it regain anywhere form 25-50% of my mana. I can typically sustain my mana with constantly shocking/stormstriking with just using the small combat mana pots.
As a side note, I thought the dodge rate of bosses (from behind) was 15%, not 10%. Do you have a good link for the 10% value? (I don't have a link--just some vague memory of reading that although mobs can't parry or block when you attack from behind, their dodge rate is 15%.)
hey pater, i put that number in as an implicit suggestion from someone who said that a lot of the time you can avoid the parry/block portion of avoidance by being behind a mob. the way it breaks down is (roughly):
5% dodge
5% block
5% parry.
dodge you can never get rid of, regardless of where you stand. So, i took the other two, totalling 10%, and said to the sim "pretend you're able to be behind the mob half of the time". So, it's a compromize between MC style tank & spank with optimal positioning, and more mobile encounters where you're forced to spend a good amount of time in front of the mob you're dps'ing.
I just use the term "dodge" when I actually mean "all avoidance, whether it's dodge block or parry" for simplicity's sake.
As a side note, I thought the dodge rate of bosses (from behind) was 15%, not 10%. Do you have a good link for the 10% value? (I don't have a link--just some vague memory of reading that although mobs can't parry or block when you attack from behind, their dodge rate is 15%.)
Conventional wisdom has a +3 mob at 5.6% to dodge(5% base + 0.04 for each point of their defense skill over your weapon skill). Block/parry can only happen when attacking from the front.
If you're looking to make the raid simulation as accurate as possible I would lower the block/parry to no more than 1%. There aren't any current raid encounters where you can't be behind the boss basically all the time. Avoiding cave-ins on Gruul is probably the worst case right now and even there you're forced to spend a very small amount of time in front of him.
Last edited by Morelis : 04/30/07 at 7:08 PM.
Reason: elaboration
TY, that's what I thought originally, from the combat mechanics 3.0. Guess I let someone lead me astray that led to the 15% belief. (Fortunately haven't made any decisions or coding based on that number.)
04/29/07, 8:48 PM
Similar Threads
