mechanics

04 Feb 2012

(was this recorded anywhere?)
healing—healing someone 'tags' them for a certain time (5 sec?), regardless of how much healing was done. during this time a portion of their damage goes to you for credit in PvP. tagging more people splits your attention among them equally, while other healers joining in could also potentially reduce your share of credit for damage.

no easy way to fix the 'stealthed rogue in BG' problem without affecting either someone who is CC'd, or giving healers an incentive to tag as many people as possible to be in combat with them (as well as for all other damage users to use AoE). Ignoring this second, simply counting everyone on damage list (including credit from healing) at time of death might work, if suboptimal since it still means either tagging as many as possible (AoE) or penalizing people who are CC'd (share within a group proportional to damage/credit). no reason to increase complexity of mechanics unless 'stealthed rogue leeching points' is a problem.


however, can also be extended to PvE! use for threat with local complexity.

skipping back to tanking though, 'boss attacks damage multiple people'. restores positioning as a complex mechanic with both positive and negative possibilities, instead of the purely negative (for the group) of 'frontal AoE cone'. Only being used for certain, flagged mobs (all bosses, maybe other powerful elites as well) helps with both roleplaying and server load.

not sure of justifications for tank mitigation so that 'a dodge doesn't lead to reduced damage on other targets'... separate from [the issue of balancing] other classes' relative mitigation but if a tank is being kept alive by only one healer, then a 1% marginal decrease in incoming damage is only equal to a 1% increase in healing for that single healer..? must be compared to dps as well though, since that's ultimately the concern for raid group composition. would it make sense for parry's damage reduction to be less than 1, but even lower for larger raid sizes..? (same with block) the question becomes do lower dodge rates make roleplaying sense (or would not doing so have gameplay benefit) especially if different raid sizes provide different parry reduction values.

if parry (and block) gave lower overall mitigation as a single target vs a raid boss compared to dodge, but allowed damage splitting when adjacent to the primary target unlike when dodging an attack, would it be balanced..? a rogue using Evasion would want to stand alone to avoid diverting attacks onto anyone else (..usually) and would have no reason to stand next to another class taking damage due to lack of total split damage reduction from dodge, while a class with parry/block/absorb and higher armor would want to.

not mentioned above, higher armor would mean taking a larger portion of damage even though the total pre-mitigation damage would always be higher when split. A divergence in functionality like this would be more acceptable if success was less balanced around maximum output including highest tank avoidance/mitigation, and more about the balance between multiple situations including when the tank does not have aggro. no idea about druid tanks and dodge, maybe they would just be less effective at protecting other raid members due to dodging attacks instead of being hit by them and parry/blocking?

in which cases would a lower overall single-target mitigation rate from parry/block make up for the greater predictability of incoming damage compared to dodge? if healer mana isn't a concern when the tank is taking damage, does this mean healer threat needs to be more difficult compared to a bear tanking, or would this be balanced by damage splitting in the situations where another player took aggro? can it be balanced at all for encounters where threat isn't a problem?

parry mitigation less than 1 means that the amount of parry+block must be more than marginal dodge + absorb, so it's inherently lower variance in incoming damage... if all raid sizes have at least some reduction in effectiveness of parry/block, then average mitigation probably balanced at a certain raid size... larger raids see lower total reduction but less spikyness for parry/block tanks, while smaller raids see greater total reduction as well for parry/block tanks. since incoming damage is fundamentally limited by cast time, RNG and tank health, larger raid sizes necesarily have to shift away from primary target damage anyway... just needs to be more fun and less "everyone must be perfect" (lack of both invisible and visible carrying).

did dodging bears ever make sense. any more than parrying a dragon's attack...

summary: avoiding damage split makes 'sense' for dodge and miss. this would mostly be bad for any tanks that stack dodge (which shouldn't really be that high, ideally, anyway). however it would only be bad if the primary tank loses aggro, which depends on other mechanics. furthermore only against bosses which have damage splitting mechanic for their attacks.

it's possible to balance this by slightly penalizing/balancing for tanks with parry and block to have lower mitigation when tanking... but it means this is only balanced for certain expectations of losing aggro, and a tank which can avoid this might be better off being a dodge-based tank class. damage splitting roles could even be performed by secondary tanks. the risk from higher incoming damage variance on a dodge-based tank is similar to the risk when damage splitting, and in both cases it may not be possible to fully balance this against overall mitigation advantage for all encounters.

also see, positioning for threat and risk of pulling aggro.


back to PvP and healing:
Cataclysm PVP Feedback // aka 'too many healers and Blood DKs' (p39)
http://us.battle.net/wow/en/forum/topic/3967948445

splitting healer attention 'credit' in PvP: more is bad compared to just focusing on high-damage players. however in PvE: more is good when aggroing is from exceeding a threshold. tentative: while a mob remembers prior threat, healing threat is only added if it affects someone who damaged the mob in the last few seconds / is the current target..?

if overall healing threat is relatively constant (or capped at an asymptote with damage being similar) this leads to the strategy of sporadic aoe heals to tag for splitting threat. optimal mechanics might depend on server load..

the other possibility was not basing healing threat on the healing done at the moment of heal, but being based on the healer having tagged that player and a share of credit for their subsequent actions. completely different model but how would it work for healing throughput..?

is it cheaper to renew a timer with every attack ('attacked mob in last 5 sec') or to do a divide operation when a heal takes place?

[tangent] getting group threat levels on a tank asymptotic to dps from other players at large group sizes can be done with faster threat decay in larger groups. threat spikes from rng can then lead to single mobs peeling away from the group.


08 Feb 2012

'chained' relationships based on logical distance of an action. Healing would be flagged as 'originating from a threatening entity' when done to a mob's current target, or possibly 'a target in the last 5 sec'. That healer would then be ranked on a second-level list along with sources of incoming damage. Healing anyone on the second-level list would generate threat, but would not lead to being placed on the second-level list.

May be possible for total healing threat to continually increase with number of mobs for which healer is generating threat. This would be a solution to two contrasting situations, one where a player tanking a single mob would occasionally do a small amount of damage to another so that a dedicated healer has their threat split between those two mobs, the other where a player damaging a single mob does enough threat to a second, weak mob to keep it aggroed while attacking the primary mob to (once again) split healing threat. (These show that prioritizing healer threat based on which mob is targeting the healed player is not a workable mechanic.)

Other possibilities for unpredictability in large groups are if 'random component to aggro' is cheap enough to be done for all mobs, and capping total threat of non-damage aoe abilities (are there even any left?) the same way aoe damage is capped.

No conclusions on balancing number of healers in a 'realistic' way... even a system that reduced effectiveness of healing as time went on would not achieve the goal well. The perception of enrage timers delimiting competence and consequences of tank death from RNG are significant; if not for these then periodic events that led to loss of group members and weren't preventable by having more healers could be used more. When balanced off of raid performance (must kill adds with certain HP, or other dps check) must take into consideration range of group performance and use mechanics with which the group can compensate for individual player skill deficits. As with other types of class stacking the tradeoff of time and effort for a greater chance of success may be acceptable if the cost is clearly visible.

A system where capped aoe damage is not done equally, but instead the closest mobs receive full damage might have gameplay benefits that justify a higher server cost (and display complexity)... for example not feeling penalized on single target dps when using an aoe spell that hits multiple. Greater threat variability and increased difficulty tanking would be a 'bonus'. Being able to aim a spell so the most important targets receive the full effect would become a component of skilled play.

One interesting mechanic used in Aion is spell cancellation from taking damage, based on the damage from the attack. A counter-stat ("Concentration") causes a linear reduction in the size of an attack for the purpose of calculating interruption but it's probably broken/imbalanced especially with the game's scaling... but it leads to a very different feeling when a healer takes aggro even without the possibility of being locked out from a spell school. It provides an immediate danger from pulling aggro on a boss (multiple interruptions leading to death) without the vicious cycle caused by a healer healing themself and generating more threat or the long-term penalty of total mana used.

Examples of PvE healing threat and AoE tanking difficulty:
- boss goes crazy (period of high risk of pulling aggro); an injured player steps away and uses a self-heal ability. For the first few seconds they are still on the 'second-level list' and generate threat from self-healing; however since they were not the boss's primary target this healing does not keep them on the second-level list and they drop off of it, and the remainder of their self-healing does not generate threat on the boss.

- a tank/healer hybrid is healing players taking damage from separate mobs when one of those mobs aggros the healer. After taking some amount of damage a specialized tank manages to recover aggro, and the healer begins to heal themself. This does not generate threat on mob A which was doing incidental damage to player A without targeting them, whom the healer did some healing to in the past few seconds; it does generate threat on mob B which is targeting player B whom the healer healed recently; and it also generates threat on mob C which recently attacked the healer. The amount of threat on each of those two mobs is not half of what it would be for a single mob, but instead an intermediate amount somewhere between half and full. After several seconds, heals to self or to the player tanking mob C would no longer generate threat on mob B [see note 1].

This is similar to what would result from "healing threat based on proportion of each mobs' total threat list with threat decay" but would be much cheaper for the server if updating timers is cheaper than several divide operations for each heal.

- a tank is attacked by twice the usual number of mobs that do half normal damage. The tank's aoe attacks (or splash threat from single-target attacks) do more total threat but significantly less than double the total amount of threat due to caps/reductions on aoe threat and damage. AoE damage from other players in the group scale similarly to the tank's threat, but faster threat decay with twice the usual mobs and unequal distribution of AoE damage with some mobs taking full damage (different for tank vs other players due to proximity to ability origin) means they are more likely to pull aggro. The healer is doing the same amount of total healing since the mobs do half normal damage, but the total amount of healing threat is increased. So the tank is doing roughly normal threat to some mobs (the closest for aoe attacks) and significantly reduced threat to other mobs for reduced average threat and higher variance; the healer is doing reduced threat per mob but the same threat to every mob, and no upper bound on the total amount of threat as the number of mobs increases for the same healing done. These curves for average tank threat vs healer threat should use complementing formulas, but the expected values of tank dps and mitigation and mob damage should mean that the healer pulling aggro is assured for large groups.

- tank is attacked by a large group; healer pulls aggro from mob A and takes damage. Tank recovers aggro. Healer continues to heal tank and is healed by hybrid dpser; this generates aggro only on mob A, which has also become more sensitive to aggro (threat decay) due to switching targets, and so hybrid dpser pulls aggro on mob A. ...well this doesn't work, since the healer would be on the 'second-level list' for all mobs, although on first list only for mob A (and hybrid dsper is already on second-level list of mob A due to using aoe attacks)... but say if hybrid dpser had been attacking something else, then they would have generated a small amount of healing threat on mob A and as a result... not pulled aggro. So this is more a 'threat decay' thing, while the purpose of first and second-level 'threatening target' lists (separate from current threat amounts list) would be only to cause mobs to see healers as 'threatening' in some cases without the healer needing to cause damage and without making the healer permanently threatening by self-healing and renewing that threatening status.

Example of that last: player A is tanking mob A away from the boss. Player A periodically damages the boss to maintain some uptime on their second-level list, so when healed the healing threat is split between mob A and the boss (although closer to full than half). This is fine and intended use of mechanics.

However, healer pulls aggro on mob A from healing player A despite the reduction of healing threat from healing mob A's damage. Healer takes some damage from mob A before player A recovers aggro. They need to heal themself at this point, and will generate less threat on mob A if they can get on another mob's list of 'threatening targets'. They can either... 1) cast a quick spell on the boss, or 2) throw a quick heal on the boss's current target, while still healing player A, and then heal themself with some healing threat being split with the boss instead of going to mob A. This reduces the chance of pulling aggro from mob A again. While if simply attacking a mob put you on their first list (instead of being targeted by that mob), then player A would already be on the boss's first list, the healer would be on their second list from healing player A, and healing threat from self heals would be split between mob A and the boss with no action necessary on the part of the healer, lowering their skill cap. It would also mean that either you would be penalized for self-healing just before dropping off the first list (if you could normally avoid going from first to second list), or... hmm.. it would mean that a healer who was trying to act as a threat transfer agent for another healer would need to occasionally attack the boss (which would put them on the first list) instead of just healing other players (second list)... anyway maybe there aren't any reasons it would need to be "being targeted puts you on first list" instead of "healing puts you in a status where being healed doesn't put that healer in the same state"... so troublesome.~ If you were a dragon, would you care about all of this?


Right, so trying that again. The point of giving more total threat as the number of mobs increases is to try to balance it so the total risk remains roughly the same. This is only possible if there's a non-zero amount of risk despite a decrease in threat, so that the lowered risk on the 'primary' mob to which the threat would normally go without unnatural casting patterns is balanced by increased risk on the mobs to which threat is being split. Since such a casting pattern already has a cost in attention and time compared to what would be necessary if threat weren't an issue, the risk on the secondary target doesn't have to be as much as the decrease in risk on the primary target but it does still need to exist for actions in the immediate temporal continuum despite a lack of activity earlier in the encounter (so threat decay).

The purpose of special treatment for heals on a mob's current target, on the other hand, is as a heuristic for the 'importance' of those heals and consequently to denote the attention the mob pays to that healer. This is because a tagging system, while it may simplify the information or choices presented to a player and reduce server resource costs, does not give any other way of distinguishing large healing throughput.

That's almost convincing. Since this additional layer of complexity would only have any effect in the following situation: boss attacks a target. That target is healed by a healer. That healer takes damage, but not because the boss directly attacked them; it's incidental damage or from another mob. Then they heal themself or are healed by someone else (less likely). Result: this healing threat goes to the boss. Almost ask why so much focus on healing but this entire system is about healing so... one consideration is someone might feel it's too risky to do any damage at all to a boss if trying to reduce threat from self-healing, so treating healing on a mob's current target differently causes some healers to be already tagged to generate threat on self heals, while giving other healers a choice between the two distinct situations of either avoiding directly attacking the boss and therefore demonstrating a sort of deference which leads to no threat on the boss when self healing damage from whatever source, or using whatever abilities are efficient to use in that spec to damage the boss and risk pulling aggro when self healing.

In other words, in the rare situation of healing someone who hasn't directly damaged any mob in the past few seconds (which is usually themself or another healer), it would normally be desirable for that heal to either generate threat on no mobs or on as many as possible. Proper mechanics will cause that number to be one or higher in at least some cases, for the reason that a threat mechanic exists at all. When threat is a problem and a healer can't just assume that they've already attracted the attention of all mobs over the course of the fight through their previous actions, then it makes sense to keep casting patterns that maximize that number from feeling necessary by also having the possibility that it could be zero.

Summary: no 'first-level list', only a current target. Healing that player would generate threat on the boss, but it would seem there are gameplay and roleplaying advantages to not requiring players to be on the 'threatening targets' list despite being attacked. A dps player might simply try to escape without attacking or trying to tank, and after losing aggro could be safely healed after less than the usual amount of time has elapsed; the more unusual case is a healer who generates threat from healing players other than the primary target, and after pulling aggro refrains from healing themself at all.

[note 1] — a careful reading of the above text reveals a hidden assumption: that a healer being on a mob's 'second-level list' would cause all healing they do to have its threat split with that mob. In fact, this sentence, "May be possible for total healing threat to continually increase with number of mobs for which healer is generating threat" was originally something like "...number of entities on the second-level list". That seemed to be a mistake since it didn't make sense and was corrected, but it may in fact have resulted from that assumption about threat. This might not be a bad way to do it, if threat is split so the total amount of threat increases when split. It would mean that healing threat depends not only on the recent actions of the target of healing, but also on the recent actions of the healer and it would give a significance to a healer being on the 'second-level list' that otherwise isn't really present. The significance of a mob's current target, then, would be that healing that player would be the only way for a healer to get on the second-level list other than being targeted by that mob... or attacking that mob, I guess...

since right now healers can pretty much assume they're already on every mob's aggro list (at least the boss at all times) I think it was an understandable assumption to make..!~ When talking about it though, it's possible to forget that being on the 'second-level list' would only last for a few seconds and in many cases everyone you healed (other than self heals) would already be on it so your healing threat would already be going to the boss. This tagging system would be meant for more complex situations, for the specific purpose of causing healing threat to go to as few mobs as possible (and keeping it high even in the cases where it's being split). Yes I think it makes sense now.