Talk:Hero behavior/Archive 01

From Guild Wars Wiki
Jump to: navigation, search


Start

Some things are wrong in this article.... Im such a moron when I say Heros WILL infact overwrite weapon spells, altough early in the article it says otherwise 24.141.45.72 19:59, 18 November 2007 (UTC)

If it's wrong, change it for the better. Backsword 20:01, 18 November 2007 (UTC)
It says "Heroes will overwrite weapon spells with other weapon spells.", and has said so for months. -- Gordon Ecker 23:09, 18 November 2007 (UTC)
Oh, you're referring to the detection note. I'm fairly sure that, if someone has a specific weapon spell on them, heroes will not overwrite it with another copy of the same weapon spell. I've only seen them overwrite weapon spells with different weapon spells. If you know the note is wrong, you can correct the mistake, and if you think the note is unclear, you can rephrase it. -- Gordon Ecker 23:15, 18 November 2007 (UTC)
heros only referesh weapon spells upon the event of the weapon spell running out ,but if the moment calls, they will use a different weapon spell to override a current one (the same as an enchantment or item spell) they wait fo rthe full duration otherwise

Can you order heroes?

Can you order heroes to attempt to manually maintain an enchantment they will not automatically cast?

Yes, just disble (shift-click) that skill and use it manually, the hero will not re-cast or remove it on his own. 81.183.145.46 15:39, 8 February 2008 (UTC)

Hero Shadowsteps

I've never been able to force a shadowstep (clicking on the skill icon while targeting and enemy). Skills tried: Shadow Prison [e], Dark Prison, Death's Charge, Aura of Displacement [e], Beguiling Haze [e], Shadow Fang. Is this just a normal thing or an AI bug? --Risus 23:35, 31 January 2008 (UTC)

New Behavior Bug

Heroes when using certain skills with cast and cancel the skill until there energy has reached 0 making them almost useless at times. User Tenri My image.jpg Tenri 16:07, 8 August 2008 (UTC)

Skill ordering

Is it true that heroes' use of skills depends on their order in the skill bar? Can someone point me to a page that explains how it works? 64.202.165.132 10:07, 14 August 2008 (UTC)

it is true, same with doppleganger too, if u set a dagger attack chain in the wrong order they will not use it correctly, if u put self heals at the top of their bar, they will sit there and use them while in combat Annoying And Deadly 17:46, 24 August 2008 (UTC)

Spiteful Spirit

Does anyone know what the actual AI triggers are for SS? I personally use it at all costs, no matter what. Olias seems to be a little less generous with it and only seems to use it when there are four or five mosters grouped together closely, (a rare occurrence). I cast it on one target, especially healers and/or fighters that will change position to a place where it suddenly works like a charm!!! :) Why don't the heroes use it as liberally? --The preceding unsigned comment was added by User:90.216.131.140 (talk).

Endure Pain

I am going to remove the part about endure pain, unless I'm told not to. I have Jora, and she uses it whenever she drops low in health. --Unendingfear talk 01:24, 6 January 2009 (UTC)

I'm removing it tommorow. If anyone doesn't wantt me to, let me know quickly.--Unendingfear talk 02:18, 7 January 2009 (UTC)

You know, you can do basically whatever you want. If someone thinks you're full of it they'll just revert it. --Jette 02:19, 7 January 2009 (UTC)

Glyph of Lesser Energy - blocking low energy spells?

When I watched my Vekk today, I recognised that whenever he was under the effects of Glyph of Lesser Energy, he did not use any spells at all, he just attacked the called target. I switched to Zhed, he was fire ele, and it was just the same - except that he did use meteor shower when affected by the glyph. Was there an update that changed the heroes' behavior on the Glyph? If you can witness the same behavior, the line concerning the energy glyphs should be changed.

I completely agree. Whenever I give a hero GoLE whenever I see it under effect it is always fading out, due to non-use. ~~ User:Frvwfr2 frvwfr2 (talk · contributions) 01:47, 25 January 2009 (UTC)
Perhaps there was an AI change that makes Heroes not use spells when they're under GoLE, unless the spell's energy cost is big enough to benefit from it? Personally speaking, this seems to be 5-cost spells. Sousuke used to freely use Fireball, Liquid Flame and Rodgort's Invocation while under GoLE, but he would not touch Mind Blast at all. When I changed his build to use Glyph of Power instead, he then started using Mind Blast again, and frequently. (It's possible that the AI views Mind Blast as an energy management skill and does not use it unless they are running low on energy though...) More research is needed for this, I believe. - Zaxares 02:15, 14 February 2009 (UTC)

Ineffective vs. Inconsistent (Specific spell notes) -> maybe Limited Use instead

I'm unsure about creating a new subsection for just one spell (Ancestor's Rage). The current note sounds to me as if heroes do not use the skill as intended; they fail to use a skill when it would be appropriate/effective to do so.

So, maybe there's another way to divide the skills up without spawning more subsections. I think clearly we need something like the first two categories:

  • There are certainly skills that heroes fail to use; Unused seems appropriate.
  • There are many skills that heroes just waste (e.g. trying to rot non-flesh); Misused fits.

The third category should be those skills that are not used to their full potential. When heroes use them, they are used advantageously. The problem with these skills is that heroes don't always use them when they should. So, maybe, Limited Use.

So, I've recombined the last few skills under a new, common subsection title with a single explanation — if we can, let's stick with the explanation, even though the name might need to evolve.

I also removed the note about unlike other skills as redundant (since that's the reason for the skill being in that section). I also rephrased the Ancestor's Rage note to emphasize the inconsistent behavior. Tennessee Ernie Ford 05:48, 12 March 2009 (UTC)

I think the description is too negative. Sometimes limited use is a good thing, for example heroes won't waste full party healing spells such as Light of Deliverance to heal a single party member. They will, however, use some other skills such as Healing Ribbon to heal a single party member, which I consider misuse. Also, I think the "do not use them in hero builds" advice for unused skills is excessive, they can still be ordered to manually use the skills, and heroes' use of Blood is Power is only problematic if the party contains a caster who needs energy and is wielding a martial weapon. -- User Gordon Ecker sig.png Gordon Ecker (talk) 09:07, 12 March 2009 (UTC)
I've updated the Unused advice to consider carefully; I agree, never was too strong.
I guess I would consider your evaluation of Light of Deliverance to be full use; the skill is used efficiently, as a human player might do. I'm not sure I would include it in this Specific Skills section; I think of it as highlighting only problematic skills. I would agree that Healing Ribbon belongs in misuse.
re: description of Limited Use — again, maybe that isn't the best title to fit the explanation. Perhaps, Unintuitively Used. Or, Over-cautiously Used. Or, is the problem that there are really four categories to fit a section that would include any skill that heroes use differently from what human players expect? Those never used; those uncontroversially misused; those used with too much caution (more/less the current third category); and those used effectively, but differently from how players tend to use them.
I don't have a particularly strong opinion about the titles, advice/explanation, or which skills belong where. I did have an objection to the original presentation, a set of unsorted skills with sometimes ineffectively-worded issue-descriptions. My sort/division, based on that original list, was to make it easier to see the specific skills and their specific issues. I imagine that more skills belong here and a further re-organization would be more useful still. I would defer to your editorial instincts, as you have more experience with the game and with what works well on this wiki. Tennessee Ernie Ford 19:51, 12 March 2009 (UTC)

Stances that Cancel on Skill Use

From my experience, it seems that when the heroes activate a stance like Dwarven Battle Stance (which will end if you use a skill), they will liberally use another skill without regard to the stance.97.118.163.253 09:07, 22 March 2009 (UTC)

Behavior Concerning Weapon Type

Maybe someone who knows about this could add a section to describe how heroes behave depending on the type of weapon (melee or ranged) they are holding. Will they always avoid melee if they are holding a ranged weapon?97.118.163.253 09:10, 22 March 2009 (UTC)

Heroes will go into melee range if they are equipped with skills like Flame Burst or Throw Dirt, even if they are wielding a ranged weapon. But, they seem to misjudge the actual range of PBAoE skills, as they will often cast them just outside of the proper range. Vili User talk:Vili 09:13, 22 March 2009 (UTC)

What about using skills off-target?

I know they are capable of using skills on enemies besides the called target, but I find it hard to believe that I have never seen any complaints about heroes that will not use skills off-target that should be used off-target (like any condition or hex when the target already has that condition or hex on them). Is it just that nobody has looked into it or are they pretty good at using skills off-target when they should? I've noticed my AoE eles will generally attack off-target if they can catch more enemies in the AoE by doing so. Necromas 04:16, 16 July 2009 (UTC)

If they aggressive, if no target is called, if they are not lockedon, and generally left alone to make decision at their own descretion, then yes they can and often will off-target as they see fit. However, most experienced players prefer heroes to ignore those in favor of following thier commands, likely due to the overwhelming necestity of concentrated firepower/efficient kill-ratios. --Falconeye 08:31, 22 February 2010 (UTC)

Maintainable enchantments, stances, preparations, rituals

I suppose that information about skills, which AI is able and prefer maintain indefinitely in and out of combat, is really important. I saw several times, when heroes refuse to use enchantments or stances which would give them real benefits, or they activate them only in battle, irregularly and with low priority. Current article doesn't contain the full list of such things and I think that need to spend some time observing the heroes behaviour and carrying out some tests; I will try to do this as possible. Because my current observations are seriously incompleted, I keep them here instead of main article until the list will become more or less representative. If you have something to add or fix, please feel free to do it. The expected table can contain: profession; enchantments, stances, rituals, signets and other skills; better in order of priority of usage. --Slavic 18:32, 9 April 2010 (UTC)
Ranger Ranger

Monk Monk

Mesmer Mesmer

Elementalist Elementalist

Ritualist Ritualist

Dervish Dervish

Echo

I couldn't find anything mentioned about this in the article or in the talk, so I'm starting a new headline here in the talk.

I have not extensively tested this for all echoing skills or all heroes, but General Morgahn wastes alot of energy recasting Aggressive Refrain when it's running out, even though another shout/chant would refresh it before it expires.

This would be an example of unwanted behaviour. --85.226.213.71 16:53, 10 June 2010 (UTC)

Formatting changes

I've slightly reformatted the newest sections to be more consistent with other articles, e.g. removing some white space, replacing ":*" with "**", and making the headers consistent with the article. Since the TOC is so long, I moved it to the right. I kept nearly all the text the same, with these exceptions:

  • condensed the update info at the top (no reason to repeat "Update from..." on each line) and removed the parentheses from the description;
  • slightly rephrased the introduction to Specific skills (and retitled); some of the skill use is detrimental and sometimes the article discusses types of skills (e.g. heroes use any available rez skills without checking to see whether anyone else is casting another rez).
  • moved the "Forced use" out of "specific skills," since it's parallel to the broad categories.

Given the length of the article, I also added wikicode whitespace (e.g. around == headers ==) to make the article easier for folks to edit. I apologize in advance to the folks that prefer the condensed format (e.g. ==headers==).  — Tennessee Ernie Ford (TEF) 21:36, 29 June 2010 (UTC)

Combining "specific skills" with "unwanted behavior"

I think we should combine the two sections describing Hero AI use of skills.

Both the "specific skills" and the"unwanted behavior" sections describe oddities in hero skill use; both cover specific skills, groups of skills, and type of impact on the player. Both categorizations have their merits, but I think having two sections distracts from the point of the article (to make it easier to adjust playstyle to the AI).

PROs: unwanted behavior

  • divides by relative impact (misuse vs limited use)
    • with name change to "unexpected behavior," allows for AI oddities that benefit the player
  • only 3-4 categories makes for a compact display


PROs: specific skills

  • divides by type of skill — easier to maintain and more aligned with how most players already think of skills
  • more flexible categorization

If we combine them, which style suits the article better?  — Tennessee Ernie Ford (TEF) 21:47, 29 June 2010 (UTC)

Well, I'm tired and drunk and it seems I can't make enough effort to properly understand your question. But it seems to me that whichever answer justifies your edit of the page must be the right one cause I managed to understand that even in my current state and it looks really good =P - VileLasagna 22:12, 29 June 2010 (UTC)
LOL - thanks! And have another beer :-)
The idea is to do a more substantial edit that would combine what is currently Section 7 (specific skills) and Section 8 (unwanted behavior). The first divides skills by skill type (spell, ritual, ...), the second by how the AI handles it (misused, used limited, ...). Hope that helps.  — Tennessee Ernie Ford (TEF) 01:01, 30 June 2010 (UTC)

Dismiss condition

"Heroes will use Dismiss Condition more for the healing than the condition removal, and will spam it."

I removed the above note because, as described, that doesn't sound like a terrible use of the skill. It might be less than 100% efficient, but I don't want the heroes to wait for the ideal scenario before using a skill; it's good for healers to heal. Of course, if they spam it too often, I can see that they might run out of energy keeping redbars above 75%; they need enough bluebar to heal serious wounds and remove conditions. In that case, the problem isn't Dismiss Condition misuse, it's poor energy management — DC becomes one more tragic example of a known, but generic issue.

This page starts to become less useful if we list every occurrence of heroes behaving badly, because it means that it will be too hard to see the skills with which heroes have the worst trouble. If I've misunderstood the issue, please discuss it here and/or add it back to the article with the altered phrasing. Thanks!  — Tennessee Ernie Ford (TEF) 00:08, 18 October 2010 (UTC)

I see that note was added back to the article. As many times as I stare at my heroes' refusal to use the skill on their own to heal someone, even if they have plenty of energy and that person will die from degen if they're not healed, I know that it's not even true; they do not use it as a heal (though, I kept wishing they would). I have never seen them use it as anything other than a condition removal skill. Has anyone aside from the person who added that note in twice ever seen the heroes use it as a heal? If not, I think the note should be removed again.--Res 05:05, 20 February 2011 (UTC)
I've removed it. Tested and I couldn't get the hero to use Dismiss Condition to heal my character even when he was enchanted and about to die. --Silver Edge 05:59, 20 February 2011 (UTC)
Maybe your hero had a falling out with your character and didn't care if you lived ;-) (More seriously: thanks for testing.)  — Tennessee Ernie Ford (TEF) 06:02, 20 February 2011 (UTC)

Different AIs?

I'm not perfectly sure about this, but I've noticed that each hero seems to have a differnt method of using skills assuming you give each hero the exact same skill set and equipment. for example, dunkoro seems to prioritize using healing magic over prot while tahlkora prefers prot. olias uses curses more than death, livia prefers blood, etc. anyone else want to test this or am i just being crazy?Nibelhim 15:06, 6 December 2010 (UTC)

I've seen this theory before, but I've yet to find any evidence supporting it. Here are some of the reasons that heroes can appear to behave differently:
  • Are they really set up identically?
    • Are they really identically kitted out? Commonly, people will provide slightly different runes/insignias or different placement of those (e.g. Survivor on legs instead of boots).
    • Are they identically skilled? There is some evidence that the AI prioritizes from left to right, so #1 slot might get used more often than #2 slot, all things being equal.
  • Assuming identical setups, do they really have the same tasks?
    • One hero will always be closer to a target (enemy/ally) than the other; the AI prioritizes proximity over many other things.
    • Two identical heroes in the same party will end up performing different tasks. The easiest way to see this is during resurrections: hero A starts the rez a beat ahead of hero B; when the rez finishes, B has 3/4s free (no aftercast) to start something else. Now, both heroes are out of synch, so if e.g. hero B started on healing, then hero A is free to handle prots.
  • Finally, we humans tend to remember supporting evidence of a theory and often miss contrary evidence. For example, we remember every time famous people die in threes, but very rarely notice when a fourth dies shortly after or remember two weeks later that the count stopped at two.
It's possible that this theory is true, but I haven't experienced any issues by using the heroes interchangeably. If you want to test this yourself, setup e.g. Livia and Olias identically with necromancer builds and take them to the Isle of the Nameless. Flag one hero to solo one of the groups there; record the order of skills used. Wait for the dummies to recover and flag the other. If there are different AIs, you should be able to see a clear difference after repeating this 5 times.  — Tennessee Ernie Ford (TEF) 17:39, 6 December 2010 (UTC)

Test results: no differences

I tested this in the Isle of the Nameless using pairs of heroes of the same profession with no armor upgrades, the same weapon (green Draconic Scythe for the dervs, nothing for the eles) and the same build, all set to 'Guard'. I flagged one back while the other attacked the 60 armor target alone. Then swapped them around. Each hero got 5 turns in each test.

1st: Melonni and Kahmu using (in order, left to right): Avatar of Melandru, Vital Boon, Aura of Thorns, Eremite's Attack, Mystic Sweep, Chilling Victory, Signet of Pious Light, Restore Life. Attributes: 1 Healing, 11 Scythe Mastery, 10 Earth Prayers, 10 Mysticism. (No reason for these skills, just a build I had to hand.) Without fail they always used the same skills in the same order - Avatar of Melandru, Mystic Sweep, Aura of Thorns, Eremite's Attack, Chilling Victory, Mystic Sweep.

2nd: Zhed and Sousuke using Glyph of Lesser Energy, Fire Attunement, Mark of Rodgort, Flare, Searing Flames, Liquid Flame, Aura of Restoration, Restore Life with 12 Fire Magic, 12 Energy Storage and 3 Healing. (Again just an existing build.) They always maintained Fire Attunement and Aura of Restoration and always activated Glyph of Lesser Energy before attacking. Both heroes used the attacks and Mark of Rodgort in a different order each time with no discernable pattern. (Effectively same behaviour from both.)

3rd: My own theory is that heroes "prefer" skills from their default attributes (eg. Dunkoro is a better healer, Thalkora a better protector). I used Zhed and Sousuke again (Zhed starts with a water build, Sousuke fire), with Fire Attunement, Water Attunement, Flare and Ice Spear with 10 Fire Magic, 10 Water Magic. (I wanted more skills but couldn't pair up fire/water skills. All the rest had extra effects or different energy costs, cooldown etc. and brought in too many variables.) As before both maintained both attunements constantly and both heroes used both attack skills seemingly at random.

In short my conclusion is that I couldn't see any difference in behaviour between characters of the same profession and any observed differences are likely to be a sort of anthropomorphism (we want to believe they're different individuals, not different models running the same program) or the fact that in other games where you can't completely scrap and re-do an NPC party member's build there are big differences between them. Much more extensive testing would be needed to cover every aspect of Hero AI (e.g. retreating from battle vs. using self-heals in the field, being quicker to start attacking, "favourite" skills or conditions etc but I'm satisfied that they're the same. 109.151.89.145 13:57, 21 October 2011 (UTC)

That seems to me like a substantive test regime. I am particularly impressed that you also tested theory #3 (I had forgotten that is a popular rumor).
It's true that there are all sorts of other possible variables that could play a role, but I think you have demonstrated that for any given player, it's safe to use heroes interchangeably. Any other effects (in the unlikely event that there are any) are going to be minor by comparison.
Well done and thanks for taking the time to run the tests and post your results. – Tennessee Ernie Ford (TEF) 21:44, 24 October 2011 (UTC)

No problem. I mainly did it because I was bored with using the same heroes and couldn't get a straight answer on if/how they'd be different if I swapped them for others. So when I had some free time I decided to find out for myself. 109.151.89.145 10:06, 28 October 2011 (UTC)

This could simply have to do with the order the heroes are added in. For instance when I add 2 identical esurge mesmers to team, the top one always uses more energy. Probably has to do with the code preventing 2 heros from casting the same hex / enchantment at the same time at the same foe. 84.48.54.253 10:31, 17 February 2012 (UTC)

I would like someone to test this since I haven't captured this skill yet

Signet of Illusions, do heros use them correctly if you give him arcane mimicry/inspired hex/arcane thievery or other spell stealing skills (as in use them after SoI if they link to non-illusion magic attribute and before SoI otherwise)? K61824 06:20, 24 January 2011 (UTC)

Copy/Echo/Mimic skills confuse hero AI; your better off using Keystone Signet/Symbols of Inspiration since its so simple a 5 year old can play it. ^_^ --Falconeye 10:40, 19 February 2011 (UTC)
This is whats so fucking wrong about wikis. It's the blind leading the blind and pretending they know something. Keystone signet builds rely on the heroes to spam skills, something they simply don't do. So keystone heroes fall in under the same category as assassin heroes, they simply don't work because of flaws in the AI. Not to mention Symbols of Inspiration opening up yet another can of worms.. --The preceding unsigned comment was added by 84.48.54.253 (talk).
And this isn't apparent in everywhere else in the whole world? Konig/talk 16:10, 17 February 2012 (UTC)

Mend Ailment note

"Heroes will use Mend Ailment intelligently for condition removal and bonus healing."

I tested that statement out in the Isle of the Nameless. I took Tahlkora with all skills except Mend Ailment disabled and took Razah with my ritualist. Both my rit and Razah went as Rt/E with blank skill bars, no weapons, and no attribute points spent (just in case she favors anyone based on any of those things.) I moved my character between Burning and Bleeding and had Razah move between Cripple and Blind (to do this, I had to flag him between Cripple and Poison, but flag him back once he reached Blind but before Poison.) She would use Mend Ailment on Razah, not on me. Then I had my rit and Razah switch roles so that I was running between Blind and Cripple and Razah was running between Bleeding and Burning. She then used the skill on me. Since neither Blind nor Cripple cause loss of life, the usage of that skill appears to be based on which conditions the AI considers worse, not whoever needs healing.

I did a few more tests to understand better how she chooses a target. I brought Xandra with me (same setup Razah and my rit had). I had Razah and Xandra stay stationary by different ailment students while my rit ran across multiple ones. If Razah or Xandra was standing on Blind or Deep Wound, it didn't matter to Tahlkora whether my rit was about to die to degen, she'd usually cure whichever one of them was suffering from Blind or Deep Wound, even if I ran over one of them and had that condition on top of the rest. Only once did she cure me while one of them was standing on Blind. Every other arrangement I tested that did not involve Blind or Deep Wound seemed to indicate that she favored curing me (I didn't test every single possibility, though.)

To see whether she favored removing Blind or Deep Wound more, I flagged Razah onto Blind and Xandra onto Deep Wound. She cured Xandra. I had Razah and Xandra swap, and Tahlkora cured Razah. So, she seems to prefer curing Deep Wound. I had my rit run back and forth from Blind to Deep Wound, and even when I had both, she still prefered curing the one who had just Deep Wound.

There are still tons of tests that can be run here, but it seems that the way the monks use condition removal is based more on which conditions the party's suffering from rather than, at least in this case, what triggers the conditional. Up to a point, anyway. It seems that some conditions aren't given much priority, so in that case, she seems to prefer curing whoever's suffering the most conditions. I have absolutely no idea why she behaved the way she did in the last sentence of the above paragraph, though.

If I get the chance, I'll try to run some tests to figure out how the AI ranks the conditions and with different condition removal skills. But until then, I removed the note I quoted above. Testing aside, it was a bit redundant considering what the article says under Status Effect Cleansing, anyway.--Res 19:50, 24 February 2011 (UTC)

You could also test whether heroes will still prioritize removing Deep Wound when one hero suffers from Deep Wound and another hero suffers from Blind while wielding a martial weapon, since AI tend to choose targets for certain skills based on what weapon they are wielding (see the Notes sections for Backfire and Throw Dirt). --Silver Edge 22:28, 24 February 2011 (UTC)
Haven't tested that as thoroughly as I could have yet, but I did a few tests (haven't gotten around to ranking the conditions yet, though). Took my rit (channeling weapon), Tahlkora (DF weapon, Dismiss Condition, Mend Ailment, Mend Condition on her bar; rotated which one was enabled), Dunkoro (healing weapon, same skill set-up as Tahlkora), and either Koss (scythe) or Goren (hammer). My rit and whichever warrior I took swapped weapons for parts of the test. Tahlkora always stood on Blind, and Dunkoro always stood on Deep Wound.
Test set 1: Blind: Koss, DW: my rit. Regardless of whether Koss or my rit held the scythe, the monks always used DC on Koss, and MA/MC on someone suffering from DW (Dunkoro would always cure me, but running in and out of GToB sometimes made Tahlkora switch up whether she favored curing my rit or Dunkoro. But once she picked who she preferred to cure, she stuck with that until I returned to GToB again.)
Test set 2: Blind: my rit, DW: Koss. Regardless of who held the scythe, Tahlkora would always cure Dunkoro. Dunkoro would always use DC on himself, but targeted Koss with MA and MC.
Test set 1* and 2*: Set up was the same as 1 and 2 above, except that my ritualist and Koss both had scythes, scythe builds, and points in scythe mastery (instead of just Koss). Results were the same as Test set 1 and 2. Tried again with Koss and my rit both on Blind, and DC from both monks targeted Koss. MA and MC from Tahlkora targeted Dunkoro, but Dunkoro used them on Koss.
Test set 3: Blind: Goren, DW: my rit. I wanted to see if taking a warrior with a warrior weapon changed anything. It didn't. Same result as Test set 1, just with Goren instead of Koss.
Test set 4: Blind: my rit, DW: Goren. Mostly the same result as Test set 2, except that during one run, Dunkoro targeted Goren with DC instead of himself. Changed back to using it on himself when I returned to GToB.
Results: Dismiss Condition seemed to favor curing blind on a warrior regardless of their weapon, and cured someone with DW otherwise. Since they didn't cure blindness on my ritualist even when he held a melee weapon, had a scythe build, and had points in scythe mastery, it seems to be profession-based rather than weapon-, build-, or attribute point-based. Mend Ailment and Mend Condition favored Deep Wound over Blindness, regardless of weapons, unless they had no "available" DW target (reason for the quotation marks is explained in the next paragraph), in which case, they targeted the warrior.
I noticed that even though Mend Ailment is not "target other ally," Dunkoro never used it on himself during the above tests, and always chose the same target for it that he chose for Mend Condition (which is "target other ally.") So, I left Dunkoro on DW and took my rit and my other heroes off and waited for the conditions to wear off everyone but Dunkoro. I enabled Mend Ailment on him. He cured the Student of Poison. I moved Dunkoro from condition to condition to see if he'd cure himself, and he always cured the Student of Poison. To see if he'd cure himself if he had no other targets around, I gave him a melee weapon and a couple healing skills and sent him to fight the Master of Axes alone. When he got hit with Eviscerate, he would not cure Deep Wound. I waited for him to get hit by it again, and had him return to me to see if he'd cure it when not being attacked, and he wouldn't. He just waited for it to end. To see if it was a typo in the skill description or a bug in the AI, I told him to use it on himself. He did. So, it seems the AI is incapable of using Mend Ailment on themselves unless forced to. Their AI for it is the same as it is for Mend Condition.
So, it seems that if you want the monks to be able to cure conditions on themselves, or if you want them to favor curing blindness on a warrior (might also be true for other martial-weapon using professions, but I haven't tested yet), take Dismiss Condition over Mend Ailment. Mend Ailment isn't a skill I have my monk heroes use all that often, though, so if anyone else has actually seen them use MA on themselves, it'd be great to know. Otherwise, I'll add a note to MA's page and to this one that they won't target themselves with it (what section would that go under, anyway? Status-effect Cleansing/Condition Removal or Limited-Use?). And, when I get around to testing some other martial weapons (especially ranged) and professions, if they turn out the same as the tests with Koss and Goren, I'll make a note about the usage of DC vs. MA and MC.--Res 17:54, 25 February 2011 (UTC)
Awesome testing. I'm summarizing your results below; User:Res has permission to edit any misunderstandings:
  1. Monk heroes never use Mend Ailment on themselves. (incorrect behavior)
  2. Monk heroes prioritize using healing Deep Wound & Poison over Blind, regardless of skill (somewhat expected)
  3. They never use MA on casters (pending other results, this is incorrect behavior), regardless of skills or weapons.
I've used monk heroes since you tested using Dunk and it's possible (although unlikely) that a NeMo might behave differently. Points (1) and (3) strike me as worth nothing on this page as part of the exception section (towards the end) and on MA's page; point (2) seems more general and worth noting towards the top of this article (but not on the skill's page).  — Tennessee Ernie Ford (TEF) 18:32, 25 February 2011 (UTC)
I was going to strikeout 2 and 3, but I can't figure out what the wiki code is for doing it.  :/ Dismiss Condition preferred curing Blind if Koss or Goren was suffering from it, but cured DW otherwise. MA and MC always cured DW. The tests also didn't show much of anything regarding poison, except that of the students they could target, they preferred curing the poison student. The blind student is an enemy, so they can't cure that one. I tried enabling all of Dunky's and Tahl's condition removal to see the order they'd cure the students (the conditions aren't always immediately reapplied), and it went in order of Poison, Cripple, Burning, Bleeding, Deep Wound (in other words, students as they went from left to right/south to north.) I don't think this means it's the order they normally prefer to cure the conditions, though, since Deep Wound is the last one on the right, but it got priority in several cases involving party members. So maybe they have a different way of determining who to target when it comes to non-party members. For #3, they did sometimes use it on casters. Test set 1, for example, has Dunkoro using it on my ritualist, and Tahlkora sometimes using it on Dunkoro, and sometimes on my rit.
So, the list would be more like:
  1. Monk heroes never use Mend Ailment on themselves. The AI for it appears to be the same as for Mend Condition.
  2. Dismiss Condition prioritizes curing Blindness over Deep Wound if a warrior (and possibly any martial class, but I haven't tested that yet) is suffering from it, regardless of weapon, and cures Deep Wound otherwise. Mend Condition and Mend Ailment prioritize Deep Wound over Blindness, regardless of who's suffering from each one.
For the NeMo, I wasn't sure if you meant N/Mo or Me/Mo, so I tested both with the students and with the Master of Axes. They used Mend Ailment the same way the monks did.--Res 20:53, 25 February 2011 (UTC)

Split?

Should Use of specific types of skills be moved to Hero behavior/Use of specific types of skills? --Falconeye 03:11, 27 May 2011 (UTC)

I don't think that's the most useful split we could make. On the whole, this article has lots of data but it's very hard to feel informed after reading it; it would be better IMO to divide into (a) generalized predictable behavior, (b) unusual (but still predictable) behavior, and (c) unusual behavior that doesn't fall into one of the patterns. There are specific skills that would be examples of each of those.
However, until that happens, I would be happy to see that most of the specific skill usage is moved out of the primary article so that it is easy to see the main patterns of hero AI. — Tennessee Ernie Ford (TEF) 04:49, 27 May 2011 (UTC)
This page totally needs to be better organised, the suggestion of "Specific type of skills" makes a lot of sense to me but mostly I think what we need is a list by professions, of skills that AI fails to use correctly...and more over if you ask me... I shall also recommend a hero AI section in each skill pages (which I think you call that formating). Plus the small icon as listing dots.User Yoshida Keiji Signature.jpg Yoshida Keiji talk 20:30, 2 January 2012 (UTC)
See Guild Wars Wiki talk:Formatting/Skills#NPC AI. for a previous discussion about creating a section on each skill page about AI usage. --Silver Edge 23:05, 2 January 2012 (UTC)
So... what are we going to do? Robert Gee says here, TEF says not here... and now somebody went to put the same note in my Project space...I wander if its the same person... anyways...you all get together into a ring and the one that comes out, speak. User Yoshida Keiji Signature.jpg Yoshida Keiji talk 00:17, 8 January 2012 (UTC)
I'm saying that we need an article that is focused on general hero behavior; this one is too unwieldy for anyone to read. ANet is saying that they want a one-stop-shopping spot so that they don't have to dig through individual articles to be able to address specific oddities.
Both those ideas are compatible with the suggestion to split this article into two: one for the general behavior and one that is for specific skill use. This allows to help more people to more easily learn what they should expect from heroes most of the time and also understand the specific situations to which the AI responds and make it easy for ANet to find issues to address. – Tennessee Ernie Ford (TEF) 02:28, 8 January 2012 (UTC)
Okay, so you are telling me that this is just a matter of one guy stepping foward to solve it for the rest? User Yoshida Keiji Signature.jpg Yoshida Keiji talk 04:24, 8 January 2012 (UTC)

(Reset indent) What I would suggest is to have this article about the generals of hero behavior, with a note somewhere stating something along the lines of "Heroes have odd behavior with certain skills" followed by a link to a subpage of this article (I suggest Hero behavior/Specific skills, which is a list page of all the skills and how each skill is treated, done in this kind of format:

  • Shock Shock — Will cause heroes to go into touch range, but will not use.

Or something of the sort. Konig/talk 06:45, 8 January 2012 (UTC)

That's +/- what I'm proposing now (and suggested 8-9 months ago): general on the primary article, specifics on the subpage. (However, heroes do use Shock; see discussion above.)Tennessee Ernie Ford (TEF) 20:03, 8 January 2012 (UTC)
With the recent edits, the page has been halved. There are still many stubbed sections and once filled they will certainly extend longer but still I get the impression that maybe it won't need to be splitted. Where do we stand now with the split notice on top of the page? User Yoshida Keiji Signature.jpg Yoshida Keiji talk 04:51, 19 February 2012 (UTC)

Alternative Layout

Should this be split? --Falconeye 03:11, 27 May 2011 (UTC)

Dervish Heroes

In general heroes and mercenary heroes make decent dervishes, running into melee combat and using their abilities more or less effectively. However, since the dervish is one of the more complicated classes to play thanks to the teardown mechanic dervish heroes do have some problems to be aware of.

  • Heroes will wait until combat starts to cast Avatars and will not maintain them out of combat. You may wish to manually have them use avatars before attacking.
  • Heroes give no consideration for tactical use of the teardown mechanic. You must make sure they have skills that will compliment each other. If you intend to give them enchantments which only give benefits while enchanted such as Grenth's Aura, you may wish to avoid giving that hero a teardown skill.
  • Heroes will not spam certain skills like Whirling Charge, so they are not effective ways to feed Avatar abilities on heroes.
  • Heroes will not use certain combos like Grenth's Grasp and Grenth's Fingers together efficiently. In order for a hero to effectively use Grenth's Grasp you must equip them with a weapon that deals cold damage.

Pious Renewal

in response to the dervish changes, i'd like to report that hero dervishes properly spam pious renewal in between enchantment stripping attacks with a few exceptions. they wont put up pious renewal before using an enchantment stripping stance (Pious Fury, Pious Haste) and VERY rarely miss it when activating adrenal skills, likely due to PR's cooldown. have yet to test skills that use enchantments being up without stripping them (Aura Slicer). --The preceding unsigned comment was added by 72.128.88.51 (talk • contribs) at 11:08, 10 June 2011 (UTC).

Order of Importance

I think I read somewhere that heroes tend to prioritize skills on their bar from left (most) to right (least). Is this true? If it is, or if the reverse is true, can we put it in the article?--Ph03n1x 02:47, 10 September 2011 (UTC)

I was curious about this, too, since I've heard it often. I did a test a few months ago, but realized that I didn't properly control the variables, so I never posted about it.
I put Stone Daggers, Flare, and Ice Spear on Soukse's bar, in that order. They were all at attribute 10, which gives them all 50 damage. I chose these since they are so similar, that they shouldn't have different AI attached, and no recharge makes for fast testing. I took off his weapon and ordered him to attack dummies on the IotN. He used Stone Daggers 86 times, Flare 102 times, and Ice Spear 95 times. Unfortunately, I realized afterward that he was wearing a fire focus and had a fire rune on his headpiece. I was afraid it may have had an effect on the results, and told myself I'd restest, but I never got around to it. Manifold User Manifold Neptune.jpg 03:40, 10 September 2011 (UTC)

Incorrect

"Heroes do not seem to use Falling Lotus Strike or Falling Spider on their own." This is incorrect, at least when I tested them on an Assassin hero. They do use these skills; however, they sometimes hesitate which causes the target to be standing before the attack lands, which causes it to "whiff". They do use these skills, but incorrectly. 24.13.211.110 08:39, 6 December 2011 (UTC)

Just to clarify, they do use the skill correctly approximately 40-50% of the time. 24.13.211.110 08:41, 6 December 2011 (UTC)
(1) If you're convinced that other ppls will experience the same results, please remove (or otherwise edit) the relevant text in the article.
(2) Assassin hero? I didn't know that anyone actually used them. Is it just for lols? Or do you have a specific team role for them? – Tennessee Ernie Ford (TEF) 08:46, 6 December 2011 (UTC)
I've been doing a lot of testing with Zenmai just to see if I could feasibly salvage any sort of use for an Assassin hero, just to have the occasional variance. So far the answer is no. Due to the "heroes are aware of Exhaustion" update they flat out ignore Shock unless you give them caster-grade energy pools, and they don't really use Wastrel's Collapse very well. They tend to use most shadow steps immediately after combat begins which usually results in a death so I've been giving them Dash instead, which seems to be a bit better since they don't die instantly. To be totally honest they're about equal to Dervish heroes in terms of survivability, assuming that you've got them in runes, insignias and decent gear, and avoid all forms of shadow stepping and teleportation. They use Hidden Caltrops well. 24.13.211.110 08:57, 6 December 2011 (UTC)

Shock

Heroes do not use Shock, as I correctly edited in earlier this morning. I'm surprised to find it edited out by Tennessee Ernie Ford. Before I revert it back to include the data I added, I want to find out why in particular the change was made. TEF states that it isn't in his experiences to have heroes not use Shock and goes on to claim that it would be odd when other AI uses the skill. I agree that it is odd that heroes do not use Shock, but speculative comments like that should not replace research and testing. For the past couple of weeks I have been toying with Shock on my heroes and noticed that it wasn't being used, so I began testing and confirmed that it is not used under the following circumstances:

  • Heroes of any primary Profession wielding melee weapons will not use Shock if they're in melee range
  • Heroes of any primary Profession wielding ranged weapons will run directly next to the target as if to use Shock, but will not do so (which seems to be roughly in line with how Blackout is treated)
  • The attribute rank of Air Magic appears to be irrelevant in the use of this skill as no hero attempted to use it when tested at each individual rank
  • Improving energy through runes and insignias did not alter hero behavior; having the hero hold favorable weapon sets did not alter hero behavior
  • Heroes don't seem to care if they're in normal mode or hard mode and will continue to ignore Shock (should be obvious but was tested anyway)
  • When under the effects of Blood Ritual, Blood is Power, Succor, "The Power Is Yours!", or any combination of the above, Shock is ignored
  • When facing enemies with increased resistance to lightning damage, Shock is ignored
  • When facing enemies with decreased resistance to lightning damage, Shock is ignored
  • When all other abilities are recharging and while the hero has 20+ energy remaining, Shock is ignored

The final test I did before inserting the line stating the habit heroes have of ignoring the spell involved taking one hero of each primary profession, putting them into both favorable gear and "clean" (unmodified) gear, and placing only Shock on their bar. I took them both individually and in groups to the Isle of the Nameless and auto attacked the Master of Damage once before stopping (to get the heroes attacking) and observed their bars. During multiple 3 minute tests, Shock was not used once. To ensure that it was not Exhaustion causing the spell's disuse, I placed first Shock + Chain Lightning on each hero's bar, then Gale.

  • When presented with the option to use Shock or Chain Lightning, the heroes (of all primary professions, gear quality and holding any weapon) continued to ignore Shock but had no issue casting Chain Lightning, even Warriors.
  • When presented with the option to use Shock or Gale, the heroes (of all primary professions, gear quality and holding any weapon) continued to ignore Shock but had no issue casting Gale.
  • When both Chain Lightning and Gale were placed on the bar alongside Shock, the heroes alternated between Chain Lightning and Gale but did not use Shock.

This leads me to conclude that it is Shock and not any surrounding factors or mechanics that the heroes have issue with and refuse to use. I have not observed a hero using Shock once on their own over the span of just over three weeks, and I eagerly await for anyone who can prove that they act otherwise as well as explain how usage was achieved. Heroes not using Shock coincides with their odd behavior with other touch skills like Blackout, in which the hero will approach the target and remain next to them in melee range, but will not actually use the skill. Please either revert the article to include the proper information again, or present evidence to the contrary, rather than using assumption to fill the wiki. 24.13.211.110 21:41, 10 December 2011 (UTC)

The original edit made a bald statement without presenting any evidence (not even in the edit summary). In contrast, the post above provides considerable evidence. I'm not 100% convinced that heroes will never use the skill, but I think the OP has provided a strong supporting case.
Be careful about relying on Isle of the Nameless: there are all sort of skills that heroes will never use there (including Dragon's Stomp and its twin Earthquake — I recently put those on hero ele skillbars to test whether skills were prioritized L→R. Both heroes refused to use either skill against any target, whether called or locked-on, whether aggressive, guarding, or passive). I admit that I haven't tried out Shock in the last year, because I don't trust caster heroes with touch skills and exhaustion on a warrior hero is... counterproductive. But I distinctly remember it being used.
That said, I'll restore the note pending further testing. – Tennessee Ernie Ford (TEF) 21:56, 10 December 2011 (UTC)
Okay, I just did the easiest possible test: Jora with an axe and Shock: she used it consistently anytime she was at full energy. Since her kit offers her only 20Energy, each use left her with 10 points...and she wouldn't use it again until she had 20. I'll do some more testing and report back. Short story: heroes do use the skill, but perhaps not anything close to often enough to make it worthwhile to put it on a skillbar. – Tennessee Ernie Ford (TEF) 22:07, 10 December 2011 (UTC)
(conflict)I only conducted a small number of tests in the Isle of the Nameless; most of the time I was dragging the heroes along with me while I played through normal content. In one example, Acolyte Sousuke ignored Shock on his bar while I was vanquishing Mount Qinkai. He did run into melee, but he did not use the spell. 24.13.211.110 22:11, 10 December 2011 (UTC)
What was the situation surrounding Jora when she used the spell? My Koss wasn't using it when he used an axe. 24.13.211.110 22:11, 10 December 2011 (UTC)
Pretty sure heroes use it on moving foes. User Chieftain Alex Chieftain Signature.jpg Chieftain Alex 22:17, 10 December 2011 (UTC)
(Edit conflict) Further tests: with 6 non-energy skills (5 adrenal and rez signet), Jora favors Shock over Burst of Aggression. With two 5Energy skills equipped, the behavior was the same. Jora uses Shock as long as she's not exhausted (even when she had only 7 energy left).
I left from Fisherman's Haven with only her and a healer (no damage skills). I tried her set to aggressive and on guard; doesn't seem to matter. (Managed to get some skale fins out of the test ^-^). She's otherwise using a Primal Rage axe build: axe=12, air=3, so Shock is only doing 20 dmg. Obviously, this might be a poor example (relatively weak foes), but the point is: heroes will use the skill. The question is: why won't they at other times?
The Sousuke behavior is interesting. It might have to do with a separate problem that heroes have using touch skills generally: their AI says, "get into melee range to use a skill," but when they close the distance, the AI says, "oh, no, new priorities: you gonna get kilt, so get out of there" and they fail to use the skill altogether. Probably a better test would be during one of the challenge missions where you stand around waiting for foes, so the ele hero puts up wards and only thinks to use a touch skill when the target is already in melee range. (That's consistent with Jora's behavior: she's close to foes when she activates; she's not activating from a distance.)
Shorter story: I'm okay if you want to put up a note that says that heroes often fail to use Shock as expected, since your test results prove that. At best, my results show that heroes will use the skill, but they say nothing about how often. – Tennessee Ernie Ford (TEF) 22:26, 10 December 2011 (UTC)
Nope, it's not a generalized touch skill issue. I tested with an ele hero and they were using it consistently unless they were below 20Energy (with or without exhaustion), so there must be other complications involved. – Tennessee Ernie Ford (TEF) 22:34, 10 December 2011 (UTC)
Sousuke does not like getting kilt. -- Armond WarbladeUser Armond sig image.png 00:24, 11 December 2011 (UTC)
In my experience, melee heros will attempt to use shock as an interrupt, and also against moving foes. 118.209.11.123 06:49, 17 February 2012 (UTC)

AI programmer

Robert Gee mentioned that the AI programmer (I assume he is referring to Joe Kimmes) checks this article before making changes to the AI. --Silver Edge 02:21, 17 December 2011 (UTC)

I'm checking it as we speak! In all seriousness, articles like this are very useful. - Joe Kimmes Talk Page‎ 02:30, 14 February 2012 (UTC)