Talk:Hero behavior/Unexpected behavior Archive 1

2/16 Update

→ moved from Talk:Hero_behavior#2.2F16_Update

In addition to base-level AI changes, today's build made specific adjustments to the AI's use of every skill in the Unused Skills and Misused Skills sections of this page. If you've had problems with those skills in the past, I encourage you to retry them and see if the changes have at least improved Hero usage of them. Thanks to everyone who has contributed to this page; it's an invaluable resource for seeing player feedback about the Hero AI. - Joe Kimmes 22:59, 16 February 2012 (UTC)

THANK YOU! Do you have any ideas/suggestions on what we can do to make this page more useful for you in the future? ^_^ --Falconeye 23:35, 16 February 2012 (UTC)

It's pretty useful as-is, honestly, but it definitely makes it easier for me when things are directly identified as problem areas. For example, the Usage of Specific Types of Skills section has a lot of interesting data on how heroes use skills, but there's both good and bad behavior documented. The more clear it is to QA and I that an area of the AI is undesired/underperforming, the easier it is to prioritize changes. Documenting the good behavior is helpful too, of course. - Joe Kimmes 00:03, 17 February 2012 (UTC)

Was the melee Ai updated today? 217.129.113.149 00:55, 17 February 2012 (UTC)

Yes, we fixed issues that caused melee Heroes to freeze up during combat and made some general performance tweaks across the board to the Hero AI. Assassin heroes still execute their dagger chains slowly though, which is something I'm continuing to look at solutions for. - Joe Kimmes 00:59, 17 February 2012 (UTC)

My thought on that would be something along the lines of "if casting lead attack, follow up with off-hand+dual" instead of "if foe is under effect with the lead attack brush, use offhand" I think it'd be faster. Just an idea. 86.145.116.15 04:11, 17 February 2012 (UTC)Minion

Also, someone add all the teardown skills to unusable. Heroes will use them without flash enchantments (for example, stripping VoS when they must re-cast Staggering, Zealous Renewal or Sand Shards). 86.145.116.15 04:16, 17 February 2012 (UTC)

Use of profession specific skills >> its already listed. --108.38.126.227 09:25, 17 February 2012 (UTC)

I fear that the list prior the update was un-checked as in if statements were true or false. Cases like Leech Signet, Power Drain and Renewing Memories. And as you wrote "made specific adjustments to the AI's use of every skill in the Unused Skills and Misused Skills sections of this page"... I fear my favourite skills were nerfed instead... Please say you didn't touch those... Yoshida Keiji talk 10:31, 17 February 2012 (UTC)

Mr. Kimmes, for those of use illiterate in the ways of programming, is it easier to change the AI to use skills more efficiently, or is it easier to split skills and/or make skills more hero friendly? Enlightenment on the subject could lead to better feedback. ^_^ --Falconeye 08:58, 19 February 2012 (UTC)

Assassin's Promise

→ moved from Talk:Hero_behavior#Assassin.27s_Promise

"AI uses it on low health targets, even if it is vital to their attack chain" Can somebody explain the purpose of this note? Yoshida Keiji talk 09:10, 17 February 2012 (UTC)

Some assassin attack chains recuires that the target is hexed for them to work. Assassins promise doesn't get used by heroes on foes before theyre HP is below something like 25%. And therefore the assassin stands there doing nothing. It's a bad choice for a hero anyhow though 84.48.54.253 10:39, 17 February 2012 (UTC)

Thanks for the reply. I think here we have another potential wrong statement. Maybe the person who wrote this didnt put another hex spell on the same bar... Yoshida Keiji talk 10:51, 17 February 2012 (UTC)

...Why would he need to put another hex on the bar?

I understand why AP is used on low health targets, and in most cases I would agree, but there's at least one use case that warrants using it up front: Assassin combos requiring a hex. I used to run something like that on my character back in the day. Start off with AP, use an IAS and 4 dagger attacks and my foe was dead, my skills recharged and I got some energy back to boot.

I think the person that wrote that sentence wanted to do something like that. Mind you this wouldn't work since heroes would be too slow with the combo chain, but that's a totally different matter. Optimal use of AP would be for the hero to check if there is a skill requiring a hex on his bar, and if there is he should use AP as a starting hex if no other is available. Otherwise the behavior should be the way it is right now. If I had to chose one or the other though I'd stay by the current one. — Poki#3 (talk) 04:14, 18 February 2012 (UTC)

The mechanics upgrade you speak of would be great. Though since we are speaking of a hero, and heroes dont go out without the player, we are talking about team build. With a minimun of 4, why cant the whole party provide an extra hex? Specially while awared of adversaries removing AP would render the character activating that elite...useless... If applied at target reaching 50% health, then the more time opponents have the chance to wear it off. Yoshida Keiji talk 11:45, 18 February 2012 (UTC)

I can see someone not wanting their hero to stop the attack chain in the middle to cast AP (especially since that particular chain I was using required the enemy to be knocked down in the middle of it). There are pros and cons of both solutions. — Poki#3 (talk) 22:40, 18 February 2012 (UTC)

"I Will Avenge You!"

→ moved from Talk:Hero_behavior#.22I_Will_Avenge_You.21.22

"AI uses it whenever a party member dies in range, even if others (especially pets) are close to death, wasting its maximun potential." I don't know who added this as the skill page has no reference in its talk page. But...attempting to understand the author... If the AI were to wait for more to die (which is very inconvenient)...when will it know to use it? Yoshida Keiji talk 09:20, 17 February 2012 (UTC)

Blackout

→ moved from Talk:Hero_behavior#Blackout

Heroes never seem to use this skill either, even after the recent update. Can we include this fix in the next AI update? Thanks.DarkSpirit 17:25, 17 February 2012 (UTC)

That skill is the purest definition of a judgment call, which computers are notoriously bad at. Joe could spent 8 weeks working on a script for it and you'd still be better off trying to micro it when necessary.

Unless you mean they never ever use it, even when you tell them. Then you can add it, yeah. –Jette 04:25, 18 February 2012 (UTC)

No we are not talking about targeting. The AI should target the foe that I called and use Blackout on that foe when I have equipped the skill on his bar. Players shouldn't be expected to micro every hero skill, much less a simple skill like Blackout.DarkSpirit 23:34, 18 February 2012 (UTC)

Blackout is the polar opposite of a simple skill. –Jette 01:41, 19 February 2012 (UTC)

If the hero prioritizes Blackout over any other skill, then they have no skills available for 5s every 12s. Some players might object to that. I agree with Jette: it's a skill that requires good judgment to use effectively (heck, I don't think I would trust a lot of players to use it well). – Tennessee Ernie Ford (TEF) 04:55, 19 February 2012 (UTC)

Ebon Dust Aura

→ moved from Talk:Hero_behavior#Ebon_Dust_Aura

This is very funny. This skill is listed both under "Enchantments" and "Unused skills", controversial. Will remove both until its behavior is clear. Yoshida Keiji talk 13:35, 18 February 2012 (UTC)

Ritual Lord

→ moved from Talk:Hero_behavior#Ritual_Lord

I don't know what they fixed about this skill (Update: 16 February 2012), but the AI still uses it on other non-Binding Rituals spells/skills and therefore wastes it. Only when the AI will use the Ritual Lord skill on the Binding Rituals only, will the skill be fixed, and only then. -- Zayeban (93.103.156.42) 14:20, 18 February 2012 (UTC)

I have actually noticed that the ritualist will not use non-binding ritual skills while Ritual Lord is up. However, my experience is that the hero very often still does not activate it before using a binding ritual - even sometimes activates it AFTER a binding ritual, meaning it will not use any other skills for the duration of Ritual Lord. A solution would be if the hero activated RL at the same moment as a binding ritual, i.e, there was no attempt to make the hero queue it or prioritise it (which it seems to have issues with). 131.227.154.148 12:15, 23 February 2012 (UTC)

I've been experimenting on my own today during vanquishes, and I think I've come up with some insight as to what the hero might be thinking while using this skill in a build. 1) Any time they already have a spirit up in spirit range (maybe compass, but it doesn't seem that drastic) they won't recast it, even if it's obviously too far away from combat to do any good. Disadvantage is, not all spirits are Passive, and they need to know the difference so they can use offensive ones closer to their targets. 2) When active in combat, they often cast Rit Lord if at least one ritual is recharged, thinking they might be casting that next (even if combat is almost over and the spirits over the hill behind them aren't going to die). 3) They still don't always cast Rit Lord when casting a ritual, even though they could do so mid-cast. Obvious bug here.

I wouldn't say they should be programmed to not use a non-ritual skill next, I just think the player needs to be smart enough to give them 4-6 rituals in order to make the chain go off more smoothly. This worked for me. I used 5 binding rituals, and Xandra did a very good job (not flawless, but at least 75% efficient) skill chaining correctly. She cast Boon of Creation without RL up, and she cast 4 or 5 spirits using RL to buff. It makes me sad they don't maintain Boon 100% as they are able to, though. Better yet, it makes me sad that RL doesn't recharge instantly after every. single. cast. Maybe more players would use it. These two fixes wouldn't make an OP build, but it would be a significant competitor to Soul Twisting. I'll consider making the suggestion official another time. --Starfleck Almighty 04:56, 27 March 2012 (UTC)

Disrupting Dagger

→ moved from Talk:Hero_behavior#Disrupting_Dagger

"Heroes will use to interrupt spellcaster profession enemies rather than use it frequently against non-spellcaster targets."

We need to check this statement. Maybe the hero is wiser than the player and won't activate because it knows the projectile will not be fast enough and that melees usually dont cast spells. Yoshida Keiji talk 15:24, 18 February 2012 (UTC)

Destructive Was Glaive

→ moved from Talk:Hero_behavior#Destructive_Was_Glaive

"Destructive Was Glaive is only used when there are foes adjacent to them. Heroes are likely to drop rather then hold when there are foes adjacent to them."

Now that I think of this, I have never disabled an item spell on heroes...I dont think they will keep it as they do with bond enchantments but... The pre-cast can be an issue as they would get interrupted, but the "drop" comment...im unsure. Opinions? Yoshida Keiji talk 16:56, 18 February 2012 (UTC)

Teardowns/FE problem and dagger chains

I have a theory to solve both of these issues. Firstly, it can't be too hard to tell the AI to only cast a teardown when it will be stripping a Flash Enchantment. Easy, that's done. Assassin chains, I would like to know how they work out when to use the next attack, but I presume it's to do with checking if the foe has been hit with a Lead in the past x seconds. That system just needs to be simplified; making them think less. Make them just chain from lead->offhand->dual together, then check next time. (e.g. if foe is not suffering from a lead attack, use full chain) Although this would only work with short recharging chains...Minion 21:48, 18 February 2012 (UTC)

Infuse Condition

While enchanted with Infuse Condition, heroes will spam Draw Conditions and Foul Feast on minions suffering from conditions. 118.209.128.177 13:27, 24 February 2012 (UTC)

Never mind, I must have been imagining things, or there were other allies around. Olias didn't seem to do this in isle of the nameless versus disease, daze, and weakness. 118.209.128.177 14:16, 24 February 2012 (UTC)

Check again outside the isle. I have noticed heroes sometimes behave differently on the isle (I haven't tested this since the most recent update, but ANet didn't mention any other changes to the testing area). – Tennessee Ernie Ford (TEF) 16:45, 24 February 2012 (UTC)

Okay, I vanquished the falls today and this was happening again. The spiders there use apply poison, and olias would constantly lag behind spamming draw conditions and foul feast on the minions. So it looks like either they behave differently in Isle of the nameless, or they avoid using draw on disease or something. But it was definitely happening with poison in the falls. 118.209.128.177 22:40, 25 February 2012 (UTC)

Documentation Fail

A few days back TEF tagged skills with {{verify}} but then someone else came to remove them or will remove in the future. What prevents the same skills from not being "mistakenly" re-enlisted again if there is no record of which skills were there before and why they were removed? By mistaken, I mean skills that were wrongly noted due to poor observations as in false statements. That is without considering changes in their behavior. We have been using this talk page to discuss skills before incorporating them to the main page. We should do the same when taking them out. Yoshida Keiji talk 01:46, 27 February 2012 (UTC)

We can just document the ones that were removed on the talk page. It's going to happen a lot — most people will not be meticulous about documentation. – Tennessee Ernie Ford (TEF) 02:00, 27 February 2012 (UTC)

Temporaly, until the booking is complete. We shall re-tag those skills that were not given any explanation. Otherwise this page will look a joke. As in: Verification was removed because I say so (by editor). Yoshida Keiji talk 02:07, 27 February 2012 (UTC)

I wouldn't re-add things removed w/o comment; this page is at its highest value if it contains only testable and verified claims. If the idea was removed by a registered user, let's just ask them to explain what they did to test. On the other hand, I would do think we should remove undocumented behavior notes. – Tennessee Ernie Ford (TEF) 02:17, 27 February 2012 (UTC)

When I saw "verify, remove this if tested" I did not assume that meant "post on the talk page detailing your testing procedure, but don't remove this once you've verified it". I could not get my heroes to use Dark Aura or Deadly Paradox under any circumstances. Whether it's the only skill on their bar, or they have Deadly Paradox and and some assassin spells, or a mix of Deadly Paradox, attack skills, and spells, in or out of combat, it was never used. Dark Aura was never used on an ally that was actively using the touch sacrifice skills, even if Dark Aura was the only skill on that ally's bar, when the saccer had high health, low health, was surrounded, fighting a single enemy, or saccing with no enemies around. I tested Aura of the Lich, Heal as One, Masochism, (which are simpler cases) and whatever else I changed as well. Manifold 03:48, 27 February 2012 (UTC)

It's more a question of trying to avoid seeing the page bloat again with opinions about what the AI should do as opposed to facts about what it does outside of what we should expect. A lot of people have added stuff without explanation ...and, more often than not, that has turned out to be misleading. YK is asking that we take some extra time to explain what we've done so that others don't try to re-add or re-remove the same point repeatedly (as has happened in the past).

Even though there are people who only post when they have researched something extensively (including Manifold), it's helpful to at least include a note in the edit summary, (Plus: any of us can make a mistake or miss a special case.) – Tennessee Ernie Ford (TEF) 04:40, 27 February 2012 (UTC)

We are already awared that AI observations are sometimes documented (with proper researchs or not) in specific skill pages as well as in this page (Hero behavior). Since book-keeping is a maintenance job that most people skip, then shouldn't we agree to only have one place for that? As in... Are you up to keep moving talks from skill pages to here for eternity? Also, if only documented observations are to be listed in the main page...book-keeping would mean to remove notes and transfering them here as well...which starts to become a pain... I have just went to read what sysops do about page protection from IPs (non registered users) and full protection that only administrators can edit. Maybe you apply for such tools to at least have the main page safe and save the rest of us from asspain bookeeping? So far the Project AI skill usage has not been employeed by anyone but me, I guess it will only serve if a specific discussion needs meticulous analisys then? Yoshida Keiji talk 12:36, 27 February 2012 (UTC)

(Reset indent) "I wouldn't re-add things removed w/o comment; this page is at its highest value if it contains only testable and verified claims. If the idea was removed by a registered user, let's just ask them to explain what they did to test. On the other hand, I would do think we should remove undocumented behavior notes. – Tennessee Ernie Ford (TEF) 02:17, 27 February 2012 (UTC) " A few months later... If we are to remove undocumented behavior notes...this page will face a major cut-down... Shall we use Guild Wars Wiki:Projects/AI skills usage to verify, document and only then post researchs in this article's main page? Yoshida Keiji talk 07:25, 23 May 2012 (UTC)

Problems with heroes

i have been having problems with my heroes since the update, heroes will go long distances, around obstacles to aggro enemies, they no longer respond to called targets, and i often find them neglecting to use any skills. this has caused some content to be much harder to play, due to the low number of Players, and problems with heroes. please help! --The preceding unsigned comment was added by 97.83.124.21 (talk • contribs) at 03:47, 13 March 2012 (UTC).

And how do you fix lame ass comments that where placed in the wrong update page, has no indents nor signature as in bookeeping? Yoshida Keiji talk 07:28, 13 March 2012 (UTC)

New category

Created a new category specifically for spell breaker as while heroes to some degree use it correctly they misbehave in relation to the target when it's active not casting any spells / prots / heals of any form on the target. Despite the fact that they can. --The preceding unsigned comment was added by 84.48.54.253 (talk) at 00:17, 2012 April 10 (UTC).

The new category mentioned above is Hero_behavior/Unexpected_behavior#Skills_that_cause_unwanted_secondary_effects, a split from Misused skills. Yoshida Keiji talk 02:55, 19 April 2012 (UTC)