Confusing Windsong results. Theorycrafters please advise.

[Nestar]

So, was reviewing some logs this evening, and noticing some incredibly strange numbers on uptime, specifically for windsong, in comparion from myself and a fellow guildie shadow priest.
My logs/sims are showing I should be seeing approximately 16.91% uptime on all 3 windsong enchants over the course of a fight, which, my logs are fairly consistent with. (19.9, 17.5, 13.6) on the log in question for myself. However the other shadow priest, I'm seeing his logs for that fight and he had a whopping 36.0, 30.3, 28.4 across the exact same fight.
He was lower for uptimes and ticks across the board aside from mind flay (it's pretty dang close though between the two of us) Please keep in mind this was last week and our gearing has changed since then.

At the time of the logs I was in about 478, he was 476. Sims put us at about 68.5k for myself and 70.6k for him in the gear we were using at the time (I ran them after the fight to see how accurate it was). I was using a 463 haste/crit staff at the time which seemed to account for the lower sim in my gear.

Me: http://us.battle.net/wow/en/characte...estar/advanced (havn't pulled a meta for my helm yet)
Him: http://us.battle.net/wow/en/characte...aydax/advanced

The log in question. http://www.worldoflogs.com/reports/f...=13900&e=14380

So basically, ya what's going on with Windsong? Is there something I'm missing in uptimes/casts? Or did he just get incredibly lucky beyond belief.

Found another log with basically the same results

http://www.worldoflogs.com/reports/r...?s=3872&e=4337

Me at 19.9, 10.8, 10.3 him at 41.4 36.6 and 17.0

---------- Post added 2012-11-10 at 06:45 AM ----------

Only thing I can think of at this point, was the fact he was running Divine Star on both of those fights, if it indeed does proc off healing at the same rate as Blizzard's post mentioned, this could literally mean in a 25man fight (as above) losing out on a bit of damage by choosing divine star (which hits Everything in its path) over Halo (which only hits 6?) will actually be benficial as it will increase your uptime by that much. I"ll have to try it out in lfr later on to test it.

[Rfx]

Windsong has had a LOT of issues, I'm not sure if they've been fixed or not but you should go ahead and buy Jade Spirit =]

[Nestar]

Windsong has had a LOT of issues, I'm not sure if they've been fixed or not but you should go ahead and buy Jade Spirit =]

http://blue.mmo-champion.com/topic/2...ation/#post110
Jade Spirit will likely be following behind shortly with the same "issues"
Honestly I don't see a problem with Windsong, but from what I'm seeing here, there's something that can likely be "exploited" to vastly improve the value of the enchant (possibly only in 25 man). If this works, we will likely be able to do the same with Jade Spirit come 5.1.
edit - looks like there may still be a 50 second cd on Jade Spirit when it migrates over, will have to test when that happens.

Also, if these numbers are correct. It puts windsong ahead of jade spirit in terms of value (specifically in 25man where you have the ability to guarantee divine star hits 20+ targets each pass)

If I remember correctly, Jade Spirit was simming at about 25-30% more than Windsong assuming the 17% uptime on each buff. If you can consistently get the uptimes shown in these logs, it would put Windsong WAY ahead of that. I'll have to see if I can run some numbers on it later.

[Danner]

Ask Mr Robot has a nice trinket quantifier, which also includes enchants.

http://www.askmrrobot.com/wow/trinkets/priestshadow

Assuming the Ask Mr Robot theorycrafting is correct, then yes, Windsong is ahead, at least unless you are very weirdly geared.

That said... enchant procs procs off different things for different specs, meaning some specs can muster way more procs - and thus value - than other specs. This is the reason the "real ppm" system is coming. Nestar's link should give you all the info available.

[Nestar]

You may be slightly mistaken on what you're reading.
AskMrRobot uses values based on average procs for single target fights (basically what a standard Sim will provide) it should (and does) show Jade Spirit as being ahead, on single target fight.

What I'm proposing will likely only apply to 25man guilds and is basically using the Divine Star talent as a filler to gain large amounts of procs (at the expense of a small amount of damage that using Halo would provide).

Divine Star effectively gives 40ish (assuming 20 targets on both passes, leaving room for outliers such as a healer or tank not stacking with the group) chances to proc windsong every time it is cast. With blizzard's system. If windsong hasn't procced for 1.4 seconds (the example they used on the blue post) and with 22% haste (most of us will be approaching 25% by now) you should have a 5.83% chance to proc, so a single pass from Divine Star which hits 20 targets, should theoretically guarantee a proc. Two passes would have a high probability of doing so twice (as the 2nd pass will be almost a full second after the first).

Using this, You should be able to guarantee at least one (possibly 2) windsong proc's on every Divine Star CD, with a 15 second cd, that's a lot of extra procs.

[Aica]

If windsong hasn't procced for 1.4 seconds (the example they used on the blue post) and with 22% haste (most of us will be approaching 25% by now) you should have a 5.83% chance to proc, so a single pass from Divine Star which hits 20 targets, should theoretically guarantee a proc. Two passes would have a high probability of doing so twice (as the 2nd pass will be almost a full second after the first).

You've misread the example. The enchants' proc percentage is supposedly based off of chance to proc, that is, every time you deal damage/heal, the "time since last chance to proc" variable is reset to zero. Using their examples, Divine Star has no more of a chance to proc than Mind Blast. In fact, they said that flat out when they stated

Whether you’re an Affliction Warlock dealing very frequent DoT ticks and Malefic Grasp ticks, or a Holy Paladin casting purely Holy Lights, or a Combat Rogue quickly attacking and using specials, or an Enhancement Shaman attacking with slow melee attacks and spells, or a Shadow Priest channeling Mind Sear on fifty Onyxian Whelps,

the proc chance is the same for all of them.


However, Blizzard's explanation is incorrect. They give two formulas to calculate the proc percentage.

The first formula:

you’ll get 2*Haste Windsong procs or 10*Haste Elemental Force procs per enchant per minute

is quickly debunked by the fact that getting 4-5 PPM per stat is easily possible with Windsong right now, but this would theoretically require 250% haste.

The second formula:

For example, if you have 22% Haste, it was 1.4sec since the last chance to proc, and you’ve got Windsong, then the chance to proc is 2(ppm) * 1.22(haste) * 1.4(time since last chance) / 60 (sec per min) = 5.693%.

This results in haste not affecting the overall PPM at all:
Percent chance to proc
= (X PPM) * (% haste) * (average time between damage events with 0 haste) / (% haste) / 60
= (X PPM) * (average time between damage events with 0 haste) / 60

Yet we definitely see a PPM increase with haste.


Something's fishy here.

[Nestar]

It almost seems to me like they didn't account for attack/heal increasing without haste.
Apparently resto shaman's are seeing the same kind of results with multiple riptide's rolling.

I think the calculations are a bit off. The "2 PPM" they're getting is merely the multiplier they are sticking in the equation. The whole equation still is affected by massive amounts of proc'able casts. If you can spam attacks fast enough, you can get proc's way faster than the standard. Which is why this appears to be working so well, I mean, 40 attacks/heals on 1gcd every 15 seconds, that's huge.

---------- Post added 2012-11-10 at 08:47 PM ----------

Using their examples, Divine Star has no more of a chance to proc than Mind Blast. In fact, they said that flat out when they stated the proc chance is the same for all of them.

That's not quite true however, It is if you look at one specific event, with both at the same % (same time since last) however, when you need to account for 40 instances, vs 1. The chance is clearly in favor of the one with 40. And the one without, just gets an increased chance when continuing.

Honestly I think the issue lies with the cap and how fast it counts up. In order to fix this. They would more than likely need to switch to a non-linear curve on the proc chance, so have it with (time since last^2) and then remove the cap. Possibly put a mention of since combat started into the calculations somewhere. So that it's not a guaranteed proc off the bat.

Another option would having the calculation divide the chance to proc, by the number of hits landed since last proc. That way if you happen to get say 40 hits before it reaches one second, it will adjust your chance. That's likely the only way to accurately adjust for every class.

Divine Star with 40 hits merely ensures it doesn't spend much time beyond the 10 second cap. When you're at the 10 second cap, haste doesn't affect it anyways, only thing that can help at that point is volume of hits.

The formula they would likely have to use to properly balance it for all classes regardless of attack speed would be.

PPM x Haste x Time Since Last / (events since last +1) / 60 x Pre-determined coefficient = Chance to proc

The coefficient would be required to properly balance the PPM value to actually result in how many they want to fit in per minute.

[Aica]

You're not getting 40 instances all at the same time, though. You're getting 40 instances all milliseconds apart. That is, only the first hit gets a decent chance to proc, while the other 39 have almost zero. This falls under the "Shadow Priest channeling Mind Sear on fifty Onyxian Whelps" example being the same as "Enhancement Shaman attacking with slow melee attacks".

[Nestar]

They do all get the same chance though, that's the thing. They actually get a marginally increased chance.
It's since last proc, not since last hit. So if your first hit has a 6% chance, and it misses, the next one has a 6.01% based on the time, if you're hitting 40 times, odds are at least one of them will proc. Once one hits, yes the chances goes to next to 0. But seriously, casting something which will roll 40 times in a second, you're basically gonna proc at least one, and reset it within that one cast. On a 15 second cd, that's minimum 4 per minute, ignoring all possible proc's from dots.

[Aica]

You're misunderstanding. It IS since last hit. Every hit has a chance to proc, which is what the formula is supposed to be based off of. As soon as you do damage, you have a chance of proccing Windsong, and the timer resets to 0, even if you didn't actually proc Windsong.

[Nestar]

The logs I linked don't support that in the slightest. If you look when the windsong proc's are happening, All but 3 of the windsong procs during the course of the fight are started while divine star is active. Out of 48.

That's 1345 of the proc chances (all being clustered very close together which by your description should be an almost impossible chance to proc), causing up to a potential 45 of the procs (allowing for dot/filler proc's during the divine star uptime), while the other 2766 only causing the remainder. Clearly something is off with the formula

The other option is however, that it's using a seperate calculation for heals and damage procs. The numbers there are surprisingly close to double the expected. If divine star is being considered seperately that would explain it.

Just went and tested just divine star at an AH with about 8 targets for a few minutes every cd. I had what seemed a 70ish percent chance of proc'ing on each cast, with about 10% proc'ing two different types of windsong. Would have to test this longer to be sure.

[Vashi]

According to my logs. For instance garalon 10 normal kill.

Windsong mastery - 24,7% uptime ( 88 seconds )
Windsong critical - 17,3% uptime ( 61,6 seconds )
Windsong haste - 10,1% uptime ( 36 seconds )

the fight took us 5 minutes 54 seconds.