Threat capped?

[likesombode]

So we were on Majordomo heroic tonight and around 3/4 of the way through I found I was either threat capped or omen was bugged? it put me at 20000.0k threat without building more as the fight progressed. This wasn't really a problem untill the end where in the cat phase the summoned cats were going straight to the player that was lower to me on threat, he was at around 19000.0k.

Has anyone else encountered this or anything like it with the new threat changes?
Is there a threat cap?

I have included a screenshot

http://img823.imageshack.us/img823/7...hreatissue.jpg

[Airowird]

Due to the way binary numbers work, a programmer once capped threat at 20million, to ensure it wouldn't do funky stuff like overflow and put you into -20mil (and a bit, as it's actually only a problem once you're at 21 474 836,47 threat as we know it)

Note that at this point nobody can ever pull off you, as there is no possibility to gain more threat (thus out-aggro you). People were already afraid that this would happen by the time we hit Deathwing, but due to the new buff it's already an issue.

[likesombode]

Ah k sweet.

With the cat mechanic, transferring the highest threat to another target, should that work the same way? Or is it a little buggy?

[klausi]

Note that at this point nobody can ever pull off you, as there is no possibility to gain more threat (thus out-aggro you).

We regulary (weeks before the buff) hit that mark on the second Alysrazor ground phase, offtank just taunt and proceed tanking and we never had any issues with either of us drawing aggro (well i swap out of defensive stance all the time but on Omen our mage and both tanks swap positions frequently at that point).

But another question: how do you reach that cap on Major Domo? The fight ends after 10 minutes, threat multiplier is five so unless you are capable of pulling 40k dps and can't see you ever reach that. I even noticed that they changed "mage-md" on the fight, taunting off them while using mirror images for another threat bump didn't work last night.

[Airowird]

It's because Taunt mechanics makes the OT the main aggro target, so the old tank will never be able to pull off him again.
Basicly, once you cap, the only way to gain aggro is taunt. At that point taunt really is the aggro magnet, it saddens me a bit :s

[likesombode]

We regulary (weeks before the buff) hit that mark on the second Alysrazor ground phase, offtank just taunt and proceed tanking and we never had any issues with either of us drawing aggro (well i swap out of defensive stance all the time but on Omen our mage and both tanks swap positions frequently at that point).

But another question: how do you reach that cap on Major Domo? The fight ends after 10 minutes, threat multiplier is five so unless you are capable of pulling 40k dps and can't see you ever reach that. I even noticed that they changed "mage-md" on the fight, taunting off them while using mirror images for another threat bump didn't work last night.

Don't Know, I guess my omen must have bugged out then?

[leethaxor]

Tanks were getting hardcapp'd on threat on h cho'gall. Next tier it really will be a problem, next xpack it will be even larger.

[Quinafoi]

I'd note that it isn't 20 million but in fact 2 billion, Omen and other threat displays divide the number by 100 because Blizzard actually does 100 threat per point of damage base (in order for the percentage bonuses and reductions to work easily). 2,147,483,647 (or 2^31 - 1) is the maximum number that can be represented by a 32 bit signed integer.

[Airowird]

Hence I specificly stated "21 474 836,47 threat as we know it", because the community knows threat as 1 damage = 1 threat as base before Blizzard released their server-side threat numbers. I just didn't wanted to go into it too far, keep the explanation simple, you know? :P

[Coreinsanity]

Hence I specificly stated "21 474 836,47 threat as we know it", because the community knows threat as 1 damage = 1 threat as base before Blizzard released their server-side threat numbers. I just didn't wanted to go into it too far, keep the explanation simple, you know? :P

They managed to make the gold cap go up from what I understand. It used to have a value that was also capped as a 32bit signed integer. There are other variable types they could use to store threat if I recall (And keep in mind I don't know a whole lot about this kind of programming - I mostly do PHP). They could unsign threat, is there even a time when we have negative threat? that would at least double the current threat cap, if I'm not mistaken.

[Airowird]

I believe the gold cap was actually lowered (from >2mil to just 1mil) and the threat is alread unsigned iirc.
The largest problem imho lies with the fact you get 100 threat per 1 damage. In the current environment, only heals & power gains actaully use the last 2 digits, and even then healing when done while in combat with 3 mobs already causes the threat to be "imperfect", so they could easily increase the effective threat cap by 10 or 100 by decreasing the threat per action values.
But that probably requires a new game build ... which means a new xpack.

[swelt]

What they did by hotfix was reduce all threat generated in the staghelm encounter. Given that they've decided to make threat a non-factor, there's no reason why the numbers need to make sense to a human, or to preserve any kind of 1 damage = 1 threat relationship. They could just change that to 1 damage = 0.1 threat. Unlike gold, there's no reason to start extending memory spaces, they can just change the values without having any negative impact on the game.

[Coreinsanity]

I believe the gold cap was actually lowered (from >2mil to just 1mil) and the threat is alread unsigned iirc.
The largest problem imho lies with the fact you get 100 threat per 1 damage. In the current environment, only heals & power gains actaully use the last 2 digits, and even then healing when done while in combat with 3 mobs already causes the threat to be "imperfect", so they could easily increase the effective threat cap by 10 or 100 by decreasing the threat per action values.
But that probably requires a new game build ... which means a new xpack.

I always though the gold cap was 214k Gold (Stored as copper, which is around 2.1b as 32bit signed int). Recently was told they raised it to 1m (Maybe they decreased it from 2 before that?).

Either way the gold cap being 999,999 = 9,999,990,000 Copper. The max size of an unsigned int is 4.2b.

Are you saying threat has a decimal value (Last 2 digits)? If so then it can't be an integer. Integers are whole numbers afaik. If not, I don't see why they couldn't change the variable like they did with gold. But yes, it would probably require an expansion to change that, depending on how they develop the game I guess.

They could also make it so classes not in tank have reduced threat and tanks have reduced threat. This way both have reduced threat numbers overall but the tanks retain their higher threat that they were aiming at after the hotfix.

[Airowird]

You're right, I was stuck with the 2.14b in my head, it's indeed the MAX_VALUE of a signed int. An unsigned int wouldn't allow you to reach the 1mil (dunno if it's exactly 1mil or 1g short of it, but regardless...), so they probably did make the gold to be stored either in a separate int or (most likely) in a long, which has 64bits & basicly provides enough extra room. The 1mil cap was just so they don't need to provide giant frames in the UI for the gold farmers.

Threat in the game is different from the threat we calculate.
Before we had actual threat values from the server or even threat meters for that matter, the theory crafting community 'arbitrarily' chose to represent threat as 1 threat = 1 damage (unmodified by talents/stances/buffs). It was not untill Blizzard made their ingame numbers directly available that we noticed that 1 damage = 100 threat ingame. But nowadays the threat modifiers are all 'X int', only mana gain & heals are causing threat on a per 50 basis (and healing is divided among enemies in cmobat, so 3 mobs will still cause rounding).
As we said in shoutbox earlier, cutting the threat value to be 1 damage = 1 threat ingame (easy conversion, all threat mods simply changed can be done in a minor patch) would allow them to atleast release content for now untill we hit a major mechanic change patch, where they can either change it to do that decently, increase the cap by changing the type of variable threat is stored in or even easier ... make threat binary (initial pull & taunts = you gain aggro, threat dump = you lose it in favor of whoever is closest to the mob, similarly to RoS)

[Coreinsanity]

You're right, I was stuck with the 2.14b in my head, it's indeed the MAX_VALUE of a signed int. An unsigned int wouldn't allow you to reach the 1mil (dunno if it's exactly 1mil or 1g short of it, but regardless...), so they probably did make the gold to be stored either in a separate int or (most likely) in a long, which has 64bits & basicly provides enough extra room. The 1mil cap was just so they don't need to provide giant frames in the UI for the gold farmers.

Threat in the game is different from the threat we calculate.
Before we had actual threat values from the server or even threat meters for that matter, the theory crafting community 'arbitrarily' chose to represent threat as 1 threat = 1 damage (unmodified by talents/stances/buffs). It was not untill Blizzard made their ingame numbers directly available that we noticed that 1 damage = 100 threat ingame. But nowadays the threat modifiers are all 'X int', only mana gain & heals are causing threat on a per 50 basis (and healing is divided among enemies in cmobat, so 3 mobs will still cause rounding).
As we said in shoutbox earlier, cutting the threat value to be 1 damage = 1 threat ingame (easy conversion, all threat mods simply changed can be done in a minor patch) would allow them to atleast release content for now untill we hit a major mechanic change patch, where they can either change it to do that decently, increase the cap by changing the type of variable threat is stored in or even easier ... make threat binary (initial pull & taunts = you gain aggro, threat dump = you lose it in favor of whoever is closest to the mob, similarly to RoS)

I don't really see why they have the threat numbers themselves so high. The growth rate of them is exponential. I would think they would realize this is going to cause problems, especially with what they have been doing to scaling. The difference between a level 80s damage and a level 85s damage is -massive-. Bigger (it seems) than the difference between 70 - 80. I mean, I death strike crit on my DK for about 39 - 45k sometimes. That's (at 40k crit) 4million threat right there if 1 damage = 100 threat. Not even counting the spam of other stuff I'm doing.

Heck, during heroic boss fights (provided the DPS isn't above average, E.G: all < 10kDPS) I get to about 8000k - 10000k threat on omen with my DK it seems.

Personally I love the idea behind the threat increase recently. It makes my life in heroics so much less stressful. I can focus on surviving on my DK and I don't have problems with hunters and other ranged starting to DPS mid charge on my druid.

[jere]

Are you saying threat has a decimal value (Last 2 digits)? If so then it can't be an integer. Integers are whole numbers afaik. If not, I don't see why they couldn't change the variable like they did with gold. But yes, it would probably require an expansion to change that, depending on how they develop the game I guess.

Threat is still held in an integer, but it is intuitively treated as a 2 decimal place value. It's a special type of fixed point notation that is used in a few other languages (Ada has support for it for example).

They reason they did this was because of partial threat multipliers (like healing, the old blessing of salvation, the old paladin healing threat modifier, etc.). It gave them 2 decimal places to handle things like multiplying an odd number by a 50% or 25% modifier.

[Coreinsanity]

Threat is still held in an integer, but it is intuitively treated as a 2 decimal place value. It's a special type of fixed point notation that is used in a few other languages (Ada has support for it for example).

They reason they did this was because of partial threat multipliers (like healing, the old blessing of salvation, the old paladin healing threat modifier, etc.). It gave them 2 decimal places to handle things like multiplying an odd number by a 50% or 25% modifier.

That's interesting, didn't know you could do that. I am guessing now with such high TPS from tanks and DPS compared to heals plus the fact that we are dealing with threat in the magnitude of millions/billions of threat from tanks, decimals seem to be just something that is adding extra complication to a system they appear to want to simplify. I mean really, paladin has Hammer of the Righteous; Death Knight has Death and Decay, heart strike and disease spreading;Warrior has thunderclap + Rend spreading if he has the talent;Druids have thrash, claw and mangle I don't really see a point to threat being divided among mobs to the point of needing to care about decimals (There should be a way they just omit that, I would think?)

[Airowird]

Threat is only divided for global actions, such as buffing, healing or gaining mana/rage/focus/whatever (not the passive regen etc, but procs that grant +X power)

This is because in the times healing threat was actually relevant, 3 mobs hitting tank required 3x the healing, but the result would be that each mob saw as much healing threat as if it's friends didn't exist (if that makes sense).

Edit: Think I read that wrong: Yes, we don't really need to have decimal (or rather <100 accuracy in Blizzard's PoV) for heals assuming threat at lvl 15 initial dungeons is fine. As I said previously, reducing all threat gains by *0.1 would solve the threat for this expansion atleast and they have proven with Staghelm that they can hotfix that in if they wanted to.

[Coreinsanity]

Threat is only divided for global actions, such as buffing, healing or gaining mana/rage/focus/whatever (not the passive regen etc, but procs that grant +X power)

This is because in the times healing threat was actually relevant, 3 mobs hitting tank required 3x the healing, but the result would be that each mob saw as much healing threat as if it's friends didn't exist (if that makes sense).

Edit: Think I read that wrong: Yes, we don't really need to have decimal (or rather <100 accuracy in Blizzard's PoV) for heals assuming threat at lvl 15 initial dungeons is fine. As I said previously, reducing all threat gains by *0.1 would solve the threat for this expansion atleast and they have proven with Staghelm that they can hotfix that in if they wanted to.

Yeah, that's basically what I'm thinking.

Seems to me they have a lot of control over their client/server and what they can push in a fix. They added the random dungeon finder in the middle of wrath, that was a pretty significant feature. Seems to me if they can do that they could fix the threat variable type by patch 4.3 and not a new expansion. Unless it's not as simple as changing a variable - this starts to jump into areas of coding I don't really know. About the most complex thing I work with is OOP PHP right now. So I really have no idea how large scale applications like the WoW server and client have stuff like that handled, or where/what they would have to change. Though, if they can't (And even as a hotfix) doing *0.1 would provide a nice buffer until they can. That takes it from 2b to 200m, and given you need a significantly long fight to reach that I doubt it will go even double that for a while.

Edit: Also yeah, that makes sense. I was actually thinking something like that earlier, and how the amount of AoE tanking abilities has actually made that die off.

[Airowird]

Actually, they'ld take the cap from 20mil to 200mil (which would then still be 2b) and because the threat is all server-side, it should be an easy fix. Put perhaps they have a reason not to hotfix it (wonder what that reason would be, if buffing tank threat was worth a hotfix).

Also, I still vote for binary threat as an implementation in 4.4ish