Engagement Mechanics- Problems and Solutions

[Guest 4ward]

this video shows melee stickiness. English starts at 4:40. The section explaining which steps to reproduce in engagement starts at 9:55.

 

http://www.youtube.com/watch?v=_GA3U5f3WQQ

Edited December 2, 2014 by 4ward

[Chilloutman]

Can someone tell me what is so bad about engagement? Because for me its works as expected. Want to run away from melee enemy? well count with extra attack against you. Where is the problem?

[Captain Shrek]

Can someone tell me what is so bad about engagement? Because for me its works as expected. Want to run away from melee enemy? well count with extra attack against you. Where is the problem?

Good job getting into the thread after 6 pages and ignoring them all. 

 

*Frankly, I don't worry about engagement mechanics either.

[Chilloutman]

 

Can someone tell me what is so bad about engagement? Because for me its works as expected. Want to run away from melee enemy? well count with extra attack against you. Where is the problem?

Good job getting into the thread after 6 pages and ignoring them all. 

 

*Frankly, I don't worry about engagement mechanics either.

 

 

I read about first 3 pages, where people complaining about how bad engagement mechanic is, but noone state why

[TheisEjsing]

After reading this thread, I'm pretty sure the biggest problem with engagement is, how it makes people forget their manners. 

[Captain Shrek]

 

I read about first 3 pages, where people complaining about how bad engagement mechanic is, but noone state why

 

 

Fair enough. That is because most people discussing the issue vehemently have already reached conclusions and presented them in other threads. Even worse, they have all read each other's posts and know what is being said. 

[Chilloutman]

 

 

I read about first 3 pages, where people complaining about how bad engagement mechanic is, but noone state why

 

 

Fair enough. That is because most people discussing the issue vehemently have already reached conclusions and presented them in other threads. Even worse, they have all read each other's posts and know what is being said. 

 

 

Well so dont be suprised that I asked what is wrong with it. I dont have time to read all threads on forum

[Karkarov]

No, it would be first attacked by. Currently it's first engaged by. If an enemy is engaged by multiple characters, they attack the first one. In the case of first attacked by, if they were attacked by multiple enemies it would be first melee enemy they were attacked by.

 

I'll make a video this week and show you how it looks without Engagement (I have modded it out).

The only problem there Sensuki is dumb ai is no better than engagement.  If the enemy simply attacks the first target that hits them in melee how is it any different?  It isn't.  It just makes it easier for the player because now it doesn't matter if their initial deployment is crap and their mage gets engaged by someone else.  The trade off of course is you can have more tactical movement once combat starts but I am not sure that alone is worth the effort since if you deploy smart from the get go you should be set for anything you need to do outright regardless.

 

For engagement to be "good" it has to be improved though.  I stand by my long ago earlier point and say that engagement is meant to simulate MMO style mob control, well if that is the point, and I feel it is, then you should not be giving everyone engagement.  Take it away from all classes but fighters, barbarians, paladins, and monks.  Assign all enemies that don't have a class a role and if that role isn't "tank" then that enemy doesn't get engagement either.  I consider this just about a must have change to be honest for engagement to be decent.

 

Second... aoe slows are nice and all but again the point of engagement is mob control.  To force the enemy unit to stay put and or attack a specific person.  So take away the attack on disengagement (enemies will never disengage anyway), and make engagement an outright movement lock.  Once a unit is engaged they can not move out of melee with that unit unless the engagement ends for whatever reason.  They are basically "stuck" until the engaging unit dies, engages someone else and drops engagement on the target, or the target uses a disengagement power.  Truthfully this is actually almost how the system already works for the enemy AI.  Also engagement should just have a set "range" and be melee only.  Since there is no free attack involved anymore it doesn't matter what the range of the equipped weapon is.

 

Lastly one other catch.

 

If a unit is not engaged they should attack based on threat.  Yeah another mmo term.  In other words if they are not engaged and thus "locked down" they should attack whatever is doing the most damage to them.  So if a guy who is not engaged and not really being attacked in melee is getting raped by your ranger with a bow he shouldn't just stand there and take it, they should turn and go for the ranger.

 

With these changes I think engagement would work just fine.

Edited December 2, 2014 by Karkarov

[Sensuki]

 

 

The only problem there Sensuki is dumb ai is no better than engagement.  If the enemy simply attacks the first target that hits them in melee how is it any different?  It isn't.  It just makes it easier for the player because now it doesn't matter if their initial deployment is crap and their mage gets engaged by someone else.  The trade off of course is you can have more tactical movement once combat starts but I am not sure that alone is worth the effort since if you deploy smart from the get go you should be set for anything you need to do outright regardless.

 

You don't have to have dumb AI, If you position melee characters at the front at the beginning of combat and assign them to attack enemies, that should pretty much handle who is being attacked by who in melee. That is exactly what happens in the Infinity Engine games too. I keep my Warriors at the front, enemies in front of them end up attacking the Warriors. 

The times when I don't have my Warriors in the front and my Mage is out of position, I can move my Mage behind my Warriors and intercede, and pretty much every time I am able to control the AI so that my Mage doesn't get hit.

[Chilloutman]

The only problem there Sensuki is dumb ai is no better than engagement.  If the enemy simply attacks the first target that hits them in melee how is it any different?  It isn't.  It just makes it easier for the player because now it doesn't matter if their initial deployment is crap and their mage gets engaged by someone else.  The trade off of course is you can have more tactical movement once combat starts but I am not sure that alone is worth the effort since if you deploy smart from the get go you should be set for anything you need to do outright regardless.

Agree, problem is dumb AI

 

For engagement to be "good" it has to be improved though.  I stand by my long ago earlier point and say that engagement is meant to simulate MMO style mob control, well if that is the point, and I feel it is, then you should not be giving everyone engagement.  Take it away from all classes but fighters, barbarians, paladins, and monks.  Assign all enemies that don't have a class a role and if that role isn't "tank" then that enemy doesn't get engagement either.  I consider this just about a must have change to be honest for engagement to be decent.

Dont agree, engagement is not there just for mob control. Its tactical aspect even player have to take in account - I think its aimed to situation where you run away with low health characters to safety to get healed. Now you cant easily get out of combat without of consequences

 

If a unit is not engaged they should attack based on threat.  Yeah another mmo term.  In other words if they are not engaged and thus "locked down" they should attack whatever is doing the most damage to them.  So if a guy who is not engaged and not really being attacked in melee is getting raped by your ranger with a bow he shouldn't just stand there and take it, they should turn and go for the ranger.

 

With these changes I think engagement would work just fine.

Agree, but it doesnt rule out engagement. IF AI will improve they should make decision if its worth getting disengagement attack to chase ranger or if they should focus on current enemy. This could make interesting combat and unpredictable

[Chilloutman]

You don't have to have dumb AI, If you position melee characters at the front at the beginning of combat and assign them to attack enemies, that should pretty much handle who is being attacked by who in melee. That is exactly what happens in the Infinity Engine games too. I keep my Warriors at the front, enemies in front of them end up attacking the Warriors. 

 

The times when I don't have my Warriors in the front and my Mage is out of position, I can move my Mage behind my Warriors and intercede, and pretty much every time I am able to control the AI so that my Mage doesn't get hit.

While I understand your concern about 'not as in IE' I believe that copiing simplest dumb AI should not be the goal

[Sensuki]

Icewind Dale Heart of Winter did not have dumb AI. The Pillars of Eternity AI is simpler than all of the Infinity Engine game AI currently, I am merely asking for at least on par, if the AI ends up being better - then great, but I asked Josh Sawyer about this topic in an interview and he said that since they have to build the AI from the ground up, the PE AI will probably be more on the simple side for the first installment.

[Chilloutman]

Icewind Dale Heart of Winter did not have dumb AI. The Pillars of Eternity AI is simpler than all of the Infinity Engine game AI currently, I am merely asking for at least on par, if the AI ends up being better - then great, but I asked Josh Sawyer about this topic in an interview and he said that since they have to build the AI from the ground up, the PE AI will probably be more on the simple side for the first installment.

Ah, thats a shame!

[Jon of the Wired]

 

You don't have to have dumb AI, If you position melee characters at the front at the beginning of combat and assign them to attack enemies, that should pretty much handle who is being attacked by who in melee. That is exactly what happens in the Infinity Engine games too. I keep my Warriors at the front, enemies in front of them end up attacking the Warriors. 

 

The times when I don't have my Warriors in the front and my Mage is out of position, I can move my Mage behind my Warriors and intercede, and pretty much every time I am able to control the AI so that my Mage doesn't get hit.

 

 

I think this may actually be the crux of some of the misunderstandings in the debate around engagement. We're all saying we want smart, good AI, and don't want dumb, bad AI, but I think when we say that we mean very different things.To elaborate a bit on the scenario you present above, let's say you have a single enemy, a fighter, and a wizard. If you position your characters such that the fighter is close to the enemy and the wizard is farther away, the enemy then has two choices:

 

1) Attack the fighter.

2) Run past the fighter and attack the wizard.

 

Based on your above statement, you consider choice 1 to not be dumb and therefore, I assume, smart and good (please correct me if I'm misrepresenting your statement). However, I, and I think other posters in this thread, would consider choice 1 to be very dumb (and therefore bad) and would expect a smart AI to make choice 2. In choice 1, the AI decides to attack the high defense, low offense fighter, when it could just bypass it and attack the low defense, high offense wizard. I would have a hard time describing any AI that makes that choice as "smart", and it's not how I want the AI to behave. I'd like the AI to do its very best to defeat me (knowing that any game AI is only going to be so good).

 

Then, since I want smart AI (i.e. AI trying its hardest to kill me) and I want a front line and a back line, where my front line of characters can protect my back line of characters, it becomes obvious that you need something like the engagement mechanic, to prevent running past the fighter from always being the optimal choice.

Edited December 2, 2014 by Jon of the Wired

[Sensuki]

Based on your above statement, you consider choice 1 to not be dumb and therefore, I assume, smart and good (please correct me if I'm misrepresenting your statement).

Yes, I have to correct you. That is untrue. Enemy AI in Pillars of Eternity (or the Baldur's Gate games) does not attempt to get to your backline at all. This only happens when one of your backline characters is out of position and aggros nearby enemies. In the Infinity Engine games you have to correct this via positioning. It depends on which game you're playing (it's different for the BG games and the IWD games, because they have different AI) but you can usually, but not always correct your mistakes through tactical positioning and blocking. Stuff that requires actual player input.

 

The Melee Engagement system facilitates dumb AI because most of the time, enemy melee AI targeting will be overriden by Melee Engagement. Melee Engagement consists of an AI targeting clause that makes the enemy stop and attack the first character it is engaged by. Whatever smart targeting that existed before is now nullified completely, simply because they ran next to a melee character of yours - THAT is dumb AI, but necessary because the AI is penalized if they do not do that (and so are you).

 

However some melee enemies in Icewind Dale: Heart of Winter did try to do this, and you the player had to respond to those AI targeting clauses and try and stop the enemies from getting to your backline, by using formations, blocking and if needed, CC abilities. Some enemies in IWD:HoW are scripted to attack the party member with the highest AC (which is usually the Wizard), and thus you have to respond to that.

 

The Melee Engagement system simplifies this process by only requiring you to position your melee members in front of your backline, and the enemy will automatically attack them if they have enough free engagement slots (which is usually always). If this is going to happen most of the time, there isn't really a need for Obsidian to develop the AI to be very smart, if it's so easy to override by the player.

 

So no, I don't see how you can want smarter AI, but want Engagement as well. One of the whole points of the Engagement system as specified in the original Kickstarter update was to make it easy for the players who struggled to control the AI in the Infinity Engine games to do so. I didn't struggle at all, and a side effect of this system is it removes pretty much all tactical positioning and movement from combat after the initial melee has begun, as well as being ripe for abuse (incoming thread soon).

[Cubiq]

If the enemy can get past your fighter then your wizard can get away from the enemy as well.

So you can just shoot him down with arrows. And he will do less damage to both of you than if he just attacked your fighter.

So it's not really the optimal choice for him to do that in the first place.

You don't want to run around like an headless chicken? You don't have to, because the AI would be smarter if he just stayed on the fighter to begin with.

[Jon of the Wired]

I agree that the current PoE AI is pretty dumb. My expectation for a smart AI is that it would have more responses to engagement than simply stopping and attacking the character that engaged it. I'd expect a smart AI to use abilities like knockdown or escape, as appropriate, or even to run past and take the attack if it has sufficiently high deflection and endurance (in that case my backline is threatened, but at least I got a free attack out of the bargain). Even the IWD:HoW AI you're presenting as a positive example very rarely meets my criteria for being smart, by your own admission.

[Jon of the Wired]

If the enemy can get past your fighter then your wizard can get away from the enemy as well.

So you can just shoot him down with arrows. And he will do less damage to both of you than if he just attacked your fighter.

So it's not really the optimal choice for him to do that in the first place.

You don't want to run around like an headless chicken? You don't have to, because the AI would be smarter if he just stayed on the fighter to begin with.

 

On the other hand, you could just as easily leave the wizard in one place and run the fighter around so that the enemy never gets an attack in. Kiting is a hard problem for the AI to deal with no matter what you do.

[Sensuki]

Even the IWD:HoW AI you're presenting as a positive example very rarely meets my criteria for being smart, by your own admission.

Sure but it's what I would like to see in the game at a bare minimum. Encounter design can also make up for AI deficiency. The Encounter design in the beta isn't great, but at least Adra beetles have a couple of ranged attacks now.

Edited December 2, 2014 by Sensuki

[Jon of the Wired]

 

Even the IWD:HoW AI you're presenting as a positive example very rarely meets my criteria for being smart, by your own admission.

Sure but it's what I would like to see in the game at a bare minimum. Encounter design can also make up for AI deficiency. The Encounter design in the beta isn't great, but at least Adra beetles have a couple of ranged attacks now.

 

 

Sure, but I have to assume that if all the enemies in IWD:HoW were actively trying to get to your backline characters (using the highest-AC targeting clause, lets say), that would dramatically change the feel of the game when compared to its current state, where most enemies are content to waste time attacking people in plate armor. I don't think IWD:HoW gives you the tools you need to deal with enemies that are that aggressive on a constant basis (which is where the passive, soft CC of engagement is, to my mind, beneficial).

[Cubiq]

On the other hand, you could just as easily leave the wizard in one place and run the fighter around so that the enemy never gets an attack in. Kiting is a hard problem for the AI to deal with no matter what you do.

 

Like this?

You could but that would be wasteful of spells, which are now more important with camp supplies.

(i stopped the video because i ran out of spells.)

[Sensuki]

Not all enemies in IWD:HoW do that, I'm not 100% sure how the targeting clauses work, I am going to do some testing tomorrow, but different enemies seem to prioritize different targets, but they also have pretty snappy target re-acquisition - which is the important bit for me.

 

If all enemies bolted straight for your backline and chose not to attack your frontline, you would come to expect it and simply use the backline characters to run around in circles while the other characters smashed them around.

 

Here is me having a bit of fun with the AI in PE atm

 

[Chilloutman]

This is what we talked about in PMs Sensuki. Good AI should make decisions if its worth getting diengagements attack to run to your mage. Even smarter AI will for example try first get e.g. knockdown before disengage. Smart AI will not run to backline if it means multiple disengagement attacks. Smart AI will try to avoid AoE damage zones. You are calling for more tactical combat and at same time promoting the most simplest dumb AI which attacks first target on sight - I dont get it

[Osvir]

This is what we talked about in PMs Sensuki. Good AI should make decisions if its worth getting diengagements attack to run to your mage. Even smarter AI will for example try first get e.g. knockdown before disengage. Smart AI will not run to backline if it means multiple disengagement attacks. Smart AI will try to avoid AoE damage zones. You are calling for more tactical combat and at same time promoting the most simplest dumb AI which attacks first target on sight - I dont get it

About AI... I kind of imagine it as... for instance Medreth's group, how would he and his cowled followers act? How would you act if you were the Player controlling Medreth's group, and an adventuring party came to attack you or betray you for Nyfre's sake?

 

Turn it around, instead of thinking "I am the adventuring party fighting Medreth", think "I am Medreth fighting the adventuring party".

 

What would be the smartest strategy? Go for the one who's blocking all damage? Or run around the engagement rules and ambush?

 

I wouldn't mind seeing opponents looking like they are fleeing, but they are really just running back away from the engagement, and then rounding you (high-tailing it 180 degrees, before engaged, and then circling and coming in from behind).

 

For Medreth's group, for instance, I could see one or two of the enemies running up towards the river, and then round the rock/bushes and come in from "behind" in an attempt to flank. Perhaps even have enemies "spawn" behind the church (to the left of the door) to come in from a surprise flank attack. If it is possible for Obsidian to make this kind of triggers or scripting, that'd be awesome. But I reckon it can be really difficult to get right, and could be exploited and/or very predictable.

 

One has to consider "But what if I catch up to them mid-flank script, do they bug out? Disengage attack? Did it solve anything?" etc. etc.

Edited December 3, 2014 by Osvir

[Sensuki]

I am not promoting dumb AI. The Engagement system is, and the people that enjoy it.

Refer to previous post in thread
 

Yes, I have to correct you. That is untrue. Enemy AI in Pillars of Eternity (or the Baldur's Gate games) does not attempt to get to your backline at all. This only happens when one of your backline characters is out of position and aggros nearby enemies. In the Infinity Engine games you have to correct this via positioning. It depends on which game you're playing (it's different for the BG games and the IWD games, because they have different AI) but you can usually, but not always correct your mistakes through tactical positioning and blocking. Stuff that requires actual player input.

The Melee Engagement system facilitates dumb AI because most of the time, enemy melee AI targeting will be overriden by Melee Engagement. Melee Engagement consists of an AI targeting clause that makes the enemy stop and attack the first character it is engaged by. Whatever smart targeting that existed before is now nullified completely, simply because they ran next to a melee character of yours - THAT is dumb AI, but necessary because the AI is penalized if they do not do that (and so are you).

However some melee enemies in Icewind Dale: Heart of Winter did try to do this, and you the player had to respond to those AI targeting clauses and try and stop the enemies from getting to your backline, by using formations, blocking and if needed, CC abilities. Some enemies in IWD:HoW are scripted to attack the party member with the highest AC (which is usually the Wizard), and thus you have to respond to that.

The Melee Engagement system simplifies this process by only requiring you to position your melee members in front of your backline, and the enemy will automatically attack them if they have enough free engagement slots (which is usually always). If this is going to happen most of the time, there isn't really a need for Obsidian to develop the AI to be very smart, if it's so easy to override by the player.

So no, I don't see how you can want smarter AI, but want Engagement as well. One of the whole points of the Engagement system as specified in the original Kickstarter update was to make it easy for the players who struggled to control the AI in the Infinity Engine games to do so. I didn't struggle at all, and a side effect of this system is it removes pretty much all tactical positioning and movement from combat after the initial melee has begun, as well as being ripe for abuse (incoming thread soon).

If enemies were programmed to make decisions about whether to run through Engagement zones or not, all I have to do is understand what the clauses for it are, and then it is ripe for the same abuse that are currently present in the game (if not, even worse).

If I know that enemies are going to try and run through Engagement zones, I can also play around that and cast Fleet Feet on my Wizard and kite them around while my other party members turn them into diced meat.

 

Enemies going for the backline is, most of the time, not a smart decision and not representative of smart AI at all.

 

What's more, there are only three months left on this project and in my email interview with Josh Sawyer from June, he said that since they have to build the AI from the ground up, we probably won't be seeing anything too complex.

I am asking for something practical, within the scope of the project. Others, are not.

Edited December 3, 2014 by Sensuki