The Frostfire Bolt thread (updated 3.2.2) - Page 39

Hotan · 03/24/09, 1:39 PM

To add to the list of mod's from Hot Streak detection: Power Auras

Power Auras is not only good for Hot Streak and mage specific things, but it is very good at informing you of ANY buff/debuff on you or your target.

Omnia · 03/24/09, 3:00 PM

I use Power Auras to keep track of:

- MBAM procs
- Hotstreak procs
- Imp Scorch timer
- LB timer
- Innervates
- IV timer
- Any missing raid buff on me
- Blizzard effect (Sapphiron)

threep* · 03/24/09, 3:54 PM

Whoa, every time someone mentions a specific addon 50 people post their favorite HotStreak-BrainFreeze-whatever helper... Can we please stop that.
Thank you very much.

Frah · 03/25/09, 12:25 AM

Originally Posted by ash2ash

That's wierd. I think you might be raiding with a prot warrior that doesn't keep tclap up, resulting in a 12% drop in dmg of your main nukes for fireball/arcane specs.

Was a dk tank so that thing DKs do was up all the time anyway.

I have a funny feeling it might be to do with CDs. At what duration of fight does fire overtake frostfire. I assume frostfire would be higher for a set amount of time due to IV+combustion at start of the fight. Then at another point frostfire will again catch up/go ahead/increase lead when i had to full evo as fire. Would be intresting to figure out some rough numbers of when these events might happen.

akimsko · 03/25/09, 1:26 PM

Originally Posted by Frah

Was a dk tank so that thing DKs do was up all the time anyway.

I have a funny feeling it might be to do with CDs. At what duration of fight does fire overtake frostfire. I assume frostfire would be higher for a set amount of time due to IV+combustion at start of the fight. Then at another point frostfire will again catch up/go ahead/increase lead when i had to full evo as fire. Would be intresting to figure out some rough numbers of when these events might happen.

Both specs are fairly RNG sensitive. So I suspect they may fairly frequently end up miles from one another. But you should pull ahead on the grander scale, given equal gear and circumstances.

Are you sure your ttw/fire gear, is on par with your FFB gear?

Lilianavess · 03/25/09, 8:53 PM

I've looked around in other threads for this, but I was wondering about the LB glyph. When the dot crits, does it proc Ignite? If so, that seems like the worst synergy with critting nukes.

And as a related thought, when the dot crits, what exactly is it critting for?

sparktoria · 03/26/09, 9:36 AM

If you are just looking for some numbers, with the Living Bomb glyph on the PTR, my DoT ticks average around 700 damage, and crits are for 1400. Your mileage may vary, of course.

Zaldinar · 03/26/09, 10:10 AM

Originally Posted by Lilianavess

I've looked around in other threads for this, but I was wondering about the LB glyph. When the dot crits, does it proc Ignite? If so, that seems like the worst synergy with critting nukes.

And as a related thought, when the dot crits, what exactly is it critting for?

At present, it generates ignite debuffs, and benefits from CSD and Burnout. It has the same crit modifier as the kaboom. When it was first implemented it was only getting benefit from CSD, but that has since changed.

It does produce a significant increase in Munching/Hicuup bugs, but they seem to average out to the same basic resulting numbers as far as overall balance of damage goes.

More disturbing is that Ignite is getting partial resists now on boss dummies, which is stalling mass data collection on the glyph for me.

ThrusterNYC · 03/26/09, 11:51 AM

I have a question about mage end game specs....

Am I required to spec Arcane for max raid dps?

I am Frostfire Bolt (FFB) specced. As my gear has improved I've seen both my dps and my threat increase. I am now always number one in threat and must watch it carefully. But my dps seems to have plateaued. I'm nowhere near the top of the dps chart and can't touch Arcane mages.

I'm confident my gear is good and spell rotation and spec are correct. (12/16 gear items are 213, hit capped, enchanted, gemmed, blah, blah, blah.)

It seems that arcane is "the bomb" and nothing else compares. I've resisted switching to arcane for two reasons.

First, I understand arcane spec is very mana intensive and relies heavily on reduced cooldown Evocation. But there are so many things in boss fights that interrupt Evocation I'm worried about being left without mana. Does this happen to arcane mages?

And second, at a deep emotional level, I just like fire. With all the balancing Blizzard has been trying to do, it seems strange that arcane would so outclass the other specs. Is there really no way to make FFB work for end game?

I was planning on dual speccing FFB and PvP. But now I wonder whether arcane for raiding is mandatory....

Your thoughts?

Pinsor · 03/26/09, 12:17 PM

Originally Posted by Zaldinar

More disturbing is that Ignite is getting partial resists now on boss dummies, which is stalling mass data collection on the glyph for me.

It is rather ironic that they tried to fix double-dipping for ignite but then cause this bug with partial resists. I will bump the bug thread as I just reproduced it on the PTRs. I got some heafty partials too, not exciting at all.

BeeLz · 03/26/09, 12:23 PM

Well, if you have threat problems you might want to consider taking threat reducing talents.

Arcane does better dps than FFB, but TTW/fire is slightly better than arcane if your gear allows it ( no hit talents ) and it has exactly the same playstyle as FFB ( except no Icy veins and shooting fireballs )

Anyways, your gear isn't bad but it's far from optimal and so is your gemming.

A WWS would help though, what kind of dps are you doing right now? what's your typical group setup? etc

Zaldinar · 03/26/09, 12:24 PM

Originally Posted by Pinsor

It is rather ironic that they tried to fix double-dipping for ignite but then cause this bug with partial resists. I will bump the bug thread as I just reproduced it on the PTRs. I got some heafty partials too, not exciting at all.

I find it extra ironic that ignite wasn't double dipping anything important anyway, so they're fixing something that isn't broken, and broke it.

Edit: On the bright side, as far as I can tell the LB Glyph is incapable of self-munching the Kaboom ignite. Certainly screws with your other spells, but it won't munch itself.

◊ manly · 03/26/09, 7:24 PM

So, I was writing a PM about how ignite works, and as I was explaining it, it dawned upon me that I may have found the missing link as to why now ignite is far more reliable than before.

My ignite theory has always been that buff/debuff application time had a role in the ignite mechanics, but it was unclear why it wasn't travel time, and why logically it should be travel time that causes it for the most part. Well my new revised theory on the matter is that, as you finish casting a spell, they check first against a list of every bolt spell that you have cast to determine the base_ignite damage, rather than only the ignite debuff on the target. I think this is why a few months ago we noticed far less ignite munching (following my thread on rolling ignites). They changed their code from checking only the mob ignite debuff to check also all the spells that are currently cast but haven't landed yet. This would cover most cases, and explains why we don't have a major amount of munching going on while doing fireball/hs pyroblast.

Let me re-iterate in a more readable format

old code

protected virtual double ComputeIgnite(Target target, Spell spell){
   if(!spell.IsCrit)
      return -1;
   var igniteDebuff = target.Debuffs.FirstOrDefault(o => o.ID == Debuffs.Ignite);
   var base_ignite = igniteDebuff==null ? 0 : igniteDebuff.DamageRemaining;
   return Math.Ceiling(spell.FinalDamage * 0.40d) + base_ignite;
}

new code

protected virtual double ComputeIgnite(Target target, Spell spell){
   if(!spell.IsCrit)
      return -1;
   double base_ignite = 0;
   var travelingIgnites = _player.SpellsTraveling.Where(o => o.IsCrit && (o.School & Schools.Fire) == Schools.Fire);

   if(travelingIgnites.Count > 0){
      base_ignite = travelingIgnites.Max(o => o.IgniteDamage).IgniteDamage;
   }else{
      var igniteDebuff = target.Debuffs.FirstOrDefault(o => o.ID == Debuffs.Ignite);
      base_ignite = igniteDebuff==null ? 0 : igniteDebuff.DamageRemaining;
   }
   return Math.Ceiling(spell.FinalDamage * 0.40d) + base_ignite;
}

Enthorn · 03/26/09, 7:59 PM

Would that also then explain why the vast majority of ignite munching seen is from Frostfire + Pyroblast combinations? When I went through my log, nearly every event of an ignite being lost wasn't the tick of ignite damage. The game very rarely, if ever, loses a stack of ignite damage. What it does lose is entire debuffs from spells. For instance:

19:56'56.614 Frostfire Bolt [9452]
19:56'56.682 Pyroblast [9778]
19:56'56.803 Ignite [4775]

In that example, the 3781 ignite damage from the Frostfire Bolt was lost. The tick never got calculated. The Pyroblast did, and the ignite damage from the Pyroblast was included in the ignite tick at 19:56'59.236.

It also appears that anytime a spell's ignite debuff is lost, it is because a subsequent spell that has landed at roughly the same time (down to hundreds of miliseconds) takes priority. I haven't found an example, and maybe one is out there, of say, a frostfire bolt and pyroblast both landing, both critting, and in that order, and the pyroblast's ignite damage is the one that is lost. That never seems to happen. It's always the first spells.

So why doesn't the game recognize that two spells landed, two spells crit, and calculate both of them? Why is anything getting overwritten, is a better question.

◊ manly · 03/26/09, 8:36 PM

Its a racing condition. The base problem hasn't changed. The final damage dealt by a spell is calculated once it leaves your hand, not when it lands. While travel time doesn't seem to be the main factor in play, if their code is multi-threaded (very likely in order to allow the code to scale to multi-cpu / multi-computer servers), then it is likely that buff/debuff application is scheduled to be ran by different threads/CPUs. The end result is that the original ignite damage calculation is done at a time where possibly another thread is currently playing with the value that was just read before another thread/CPU made the change to that current value.

In other words, until they change ignite to be calculated upon landing (unlikely change), the calculation will be done at the spell cast time. This leads to possible racing conditions. I understand that under a simplistic view of things it makes no sense that it would 'lose track' of damage, but you need to think about a case where there are multiple CPUs running at the same time it isn't that simple to do. I mean, it could easily be done, but synchronizing even the simpliest of operations (dealing damage would fit that category), then you would tremendously slow down the code execution.

Silanre · 03/27/09, 4:07 AM

Some great guestimating on how the programming works!

I have been running arcane and fire/arc back and forth recently, but my questions for 3.1 at this point are:

-Will the Living Bomb glyph in 3.1 increase ignite munching?
-If so to what extent is this a loss from the raw calculated value of LB?

-Is there a practical solution to improving ignite munching? (without taxing response time in programming)

-Will haste become more valuable due to high movement fights where the longer cast time (vs frostbolt, arcane blast) ends up rewarding haste over crit even though it is looking like their relative values will be somewhat close according to RAWR etc.?
(example: fights require similar or more moving and adjusting then say OS3d, having done it both arc, fire/arc & FFB)

Thanks to everyone who is/has been submitting responsible feedback on the various bugs on the PTR and in the various forums.

ash2ash · 03/27/09, 4:37 AM

Originally Posted by Silanre

-Will haste become more valuable due to high movement fights where the longer cast time (vs frostbolt, arcane blast) ends up rewarding haste over crit even though it is looking like their relative values will be somewhat close according to RAWR etc.?
(example: fights require similar or more moving and adjusting then say OS3d, having done it both arc, fire/arc & FFB)

I don't think comparing crit to haste in the context of movement-intensive fights is going to give you any tangible results.

Not speaking from experience on the PTR, but unless you're talking about an ungodly amount of haste, movement fights are going to limit you to instants and sub-1.5sec casttime spells (re: scorch) No way I'm going to risk casting a 3s casttime spell during heigan dance with the haste available on current gear. If anything, crit might actually help you in the case of fireball/frostfire specs by generating more hotstreaks which you could cast on the run.

Point being, comparing haste vs crit for single target dps spam is going to be a better indication of what is better for you in any given situation.

Silanre · 03/27/09, 8:36 AM

It may not make any significant difference, but on the other hand 1-2 tenths of a second adds up when you are getting short turns to chain a spell. 1/2 a second might mean the difference of an additional cast. I would tend to be wary of stacking too much haste for arc for example(for mana reasons) but the loss of the third arc blast or missile baraage casts are also the largest dps loss although some of that is returned by not using as much mana.

I don't mean to go off topic, I am just curious more then anything else. Sometimes a shorter cast time is an advantage and might skew gearing choices, especially for fireball/FFB which have longer cast times to begin with compared to arc and frost which also have better/more options for casting while moving.

If anyone has done more then a few of the PTR fights, a comparison to current content like OS3d might be a fair comparison. Thanks in advance!

Jonny_Monroe · 03/27/09, 8:45 AM

Originally Posted by ash2ash

I don't think comparing crit to haste in the context of movement-intensive fights is going to give you any tangible results.

Not speaking from experience on the PTR, but unless you're talking about an ungodly amount of haste, movement fights are going to limit you to instants and sub-1.5sec casttime spells (re: scorch) No way I'm going to risk casting a 3s casttime spell during heigan dance with the haste available on current gear. If anything, crit might actually help you in the case of fireball/frostfire specs by generating more hotstreaks which you could cast on the run.

Point being, comparing haste vs crit for single target dps spam is going to be a better indication of what is better for you in any given situation.

Crit will also buff your glyphed living bomb, which means even more 'mobile damage'.

Crit is rapidly becoming the superior stat in all situations.

Althea · 03/27/09, 9:45 AM

Originally Posted by manly

In other words, until they change ignite to be calculated upon landing (unlikely change), the calculation will be done at the spell cast time. This leads to possible racing conditions. I understand that under a simplistic view of things it makes no sense that it would 'lose track' of damage, but you need to think about a case where there are multiple CPUs running at the same time it isn't that simple to do. I mean, it could easily be done, but synchronizing even the simpliest of operations (dealing damage would fit that category), then you would tremendously slow down the code execution.

If they are having problems calculating ignite upon landing, they could just put a condition in the ignite mechanic. If the ignite debuff is already on the target, then do as usual, else create a new debuff without having the limit of 1 ignite per mage

I dont think this will screw the debuff pool, having 2 crits at the same time isn't something that occour so many times, in the 4 seconds needed by the debuff to fade, i guess we will have max 2 ignite at the same time from the same mage

Zaldinar · 03/27/09, 9:56 AM

Manly, what are you using to measure how much munching you're getting?

From what I've done with ZIT in recent months and this PTR, I'm observing a completely un-quantifiable and anecdotal no change. I noticed the partial resist bug trying to see how nasty the hiccup/munching would be with the LB glyph, and the numbers are significant still...

This was five mana bars worth of normal FFB rotation (Scorch > LB > HS Pyro > FFB)

Actual Ignite Damage: 252431
Expected Ignite Damage: 269712.4
Damage Gained through Bugs: 4659.8
Lost due to Bug: 21944
Lost due to Death: 0
Unaccounted for Damage (rounding errors) 2.8

Overall vs Expected (ignoring death): -17281.4

Bugs observed: 48

The bugs observed is inflated due to the resist issue, as is lost, but I'm still seeing a fairly normal amount of clipping going on.

Brodrik · 03/27/09, 11:43 AM

There's enough going on in the game that a fine-grained synchronization lock shouldn't negatively impact code performance (and, for that matter, they must be using some degree of synchronization already). On the other hand, we already know race conditions are reasonably common in the game -- you can get killed by Gluth if Commanding Shout happens to wear off in between when they calculate Decimate damage and when they apply that damage.

galzohar · 03/27/09, 11:53 AM

A long time ago someone showed some math for the effects a faster cast would have on "interrupts" of the kind that makes you cancel your cast (void zone style). While obviously faster casts result in fewer spells canceled, the increase in # of spells not canceled caused by 1% haste increase is much much much smaller than 1% (for various interrupt rates), and therefore can be safely neglected. If anyone wants to do more accurate math again to show it I'd love to see it.

In a more intuitive approach, think that if interrupts happen every ~X seconds at random, then doubling your cast speed would half the % of the spells you lose, so if you originally lost 10% of your casts due to interrupts, then you would now only lose 5%, thus increasing # of spells casted from 90% to 95%. That's a bit more than a 5% increase in spells casted by gaining 100% haste, which is quite tiny in comparison to the straight-up DPS increase 100% haste gives (which is 100% increase (20x more) if mana and GCD aren't becoming a factor). In this situation (10% of casts interrupted, 100% haste) haste would increase in value by a factor of ~0.05, so if haste was originally worth, say, 1.3 sp - it'll now be worth 1.365. In reality you're probably interrupting a lot less than 10% of your spells, though. Again I'd like to see more accurate math but this already shows that the expected added value for haste is going to be pretty damn small on a real fight.

Zaldinar · 03/27/09, 12:59 PM

Originally Posted by galzohar

A long time ago someone showed some math for the effects a faster cast would have on "interrupts" of the kind that makes you cancel your cast (void zone style). While obviously faster casts result in fewer spells canceled, the increase in # of spells not canceled caused by 1% haste increase is much much much smaller than 1% (for various interrupt rates), and therefore can be safely neglected. If anyone wants to do more accurate math again to show it I'd love to see it.

In a more intuitive approach, think that if interrupts happen every ~X seconds at random, then doubling your cast speed would half the % of the spells you lose, so if you originally lost 10% of your casts due to interrupts, then you would now only lose 5%, thus increasing # of spells casted from 90% to 95%. That's a bit more than a 5% increase in spells casted by gaining 100% haste, which is quite tiny in comparison to the straight-up DPS increase 100% haste gives (which is 100% increase (20x more) if mana and GCD aren't becoming a factor). In this situation (10% of casts interrupted, 100% haste) haste would increase in value by a factor of ~0.05, so if haste was originally worth, say, 1.3 sp - it'll now be worth 1.365. In reality you're probably interrupting a lot less than 10% of your spells, though. Again I'd like to see more accurate math but this already shows that the expected added value for haste is going to be pretty damn small on a real fight.

I did a piece on this a while ago that has some limited application, since its really all about relative advantage periods in a timed interrupt scenario. Most of the thought process still works out though. Unfortunately the main WoW forums ate it, and it was before I was saving threads in text locally before posting them.

It was all based around the concept from this timeline:

http://i142.photobucket.com/albums/r...s/timeline.jpg

If we consider two actors putting out X DPS at different intervals, there will be periods where one will have put out more damaging events then the other, and thus have put out more damage. An easy time to think about this is a rogue in melee with a boss vs you casting at it, you start a 3 second FFB 1 second before the boss dies, you wasted 1 second of DPS time. The rogue had more opportunity to put out damage since their attacks happen more frequently. That's the basic idea.

Using BC stat conversions, if you compare a hasted vs unhasted scenario and measure these periods of advantage, you can end up with some really whacky looking graphs:

http://i142.photobucket.com/albums/r...1-788_5-55.jpg

Being the wide angle, and

http://i142.photobucket.com/albums/r...-250_25-55.jpg

Being the close in on the range we typically care about. Remember this is all relative, so it really can only be used to compare two actors directly for a sort of advantage understanding. Their relative frequency of damage events is what matters, disparity in their actual output numbers can be accounted for by treating the DPS in the equation as a 'nominal' value and adjusting the later values accordingly (An actor with a double advantage and half the DPS is on equal footing, sort of).

Really though that entire exercise is just intellectual thumb-sucking, it doesn't truly accomplish anything for our modeling efforts, it just helps explain some of the disparities we observe. Originally it was intended to explain why Frost was doing comparable to fire on some fights that the theorycrafting wasn't predicting, back when Frost was just WE + Frostbolt and Fire was 8 FB + 1 SC.

Zerstorung · 03/27/09, 3:14 PM

One of the issues I have is that everything is calculated under ideal conditions, meaning, then RNG doesn't suck and you actually get your crit rate.

Personally, I find myself on the lower end of the rng more often than not. My Fireball crit rate is 60-63%, but I'll be damned if I ever break 57% in a fight. Perhaps this is because the fights are shorter and doesn't give enough time for the rng to balance out, but in a case where you aren't getting your actual crit rate, I'd imagine haste would become the superior stat. Haste is always there, crit is not.