9 réponses à ce sujet

[Bjond]

I've posted articles on this topic in most of the games I've played since EQ1 introduced the taunt mechanic to RPGs -- at least I think it was EQ1 that did it first.  I don't recall others using it before EQ1 and if so, I know I didn't bring it up until EQ1.

Why is taunt bad?  It quite literally destroys all AI.  It permits you to give orders to the enemy.  You get to tell them where, when, and who to attack.  This eliminates strategic choice from the enemy and reduces fights to a very few mundane tactical details.  By chosing to implement taunt, agro, threat, etc. into your game, you are from the outset making the decision to make a game that is completely unchallenging strategically.  How hard would it be to beat a computer at chess if you could tell it which pieces it had to attack?  It would unspeakably easy and dull, wouldn't it?

The reply from developers is generally along the lines of "it's too computationally costly to implement body-blocking" and thus shield-walls, fortifications, heavy-armor placement, and other historically used strategies to limit the enemy's ability to attack you, while you attack them.  This wasn't precisely true even back in the days of EQ1, but it was true to say that servers didn't have enough computational bandwitdth to implement decent AI -- they barely had enough to handle pathing.  It's very far from true now and especially untrue in any single player game.  DA:O actually supports collision detection (body-blocking), for instance, as do many other games.

The reason we are stuck with taunt breaking our AI in games now is quite different than back then.  I think it's simple cultural inertia.  Developers these days are too young and ignorant to realize that taunt and tanking are artificial concepts that are actually harmful to the game.  To the average developer today, taunt and tanking are as normal and customary to RPG combat as breathing water is to a fish.  All the games they played while growing up had it.

But but but .. how do I tank if there's no taunt?  My unarmored support will get focused and gibletized!  Yup.  That's how it works and that's a good thing.  You need to protect them, but you'll need some help from the game to do that.  The solution is to model things closer to history (real history, not gaming history ^~).

IMHO, if you want to free developers to actually make interesting fights with good AI, you have to implement some concept of a shield-wall.  I agree that forcing your classic 4-8 man party to form a real phalanxe for body-blocking or a pike-wall v. charges, or other precise models on historic tactics is a bit much and probably only fun to really serious wargamers.  Instead, implement the *concept*.  Add some skills to every character to permit them to join into a shield-wall in various ways.  Heavy armored chars would perhaps increase party armor, light perhaps increase dodge, and all of it should scale up quickly the more you add to the wall.  Turn the graphics team loose on creating some nice visuals for it.  I'm sure they can come up with hundreds of ingenious ways to make a party sparkle and glitter as a group when being attacked; eg. I'd love to see Dog frisbee-catch an arrow when part of a wall.

Voila!  Not only have you returned the most critical decision of "who do we attack?" back to the AI and the developer, but you've solved the problem of large-battle scaling and contribtion for MMOs.  In your traditional MMO boss fight, you only need one tank.  More than one are often literally useless and sometimes there are limits to how many of any particular class can contribute.  Not so if you have shield-walls (real or conceptual).  Now extras make your wall stronger.  You could even make the wall a bit of a mini-game in itself.

In ancient warfare, the hero wasn't some guy solo fighting a batallion.  It was the man who stood in the wall at all costs.  If implementing things this way ever becomes widespread, young gamers would grow up with a direct visceral understanding of why Spartan mothers supposedly told their sons, "Come back with your shield or on it."  Only it will probably become something like "Dude, you broke our wall!  You suck."

--

I'm sure any fish reading this can come up with far more fun and interesting ways to implement a "shield-wall" style party ability now that you've been lifted into the air and shown that there might be life outside the water.  I sincerely hope you make such a game someday.  I'll certainly buy it if you do.

Modifié par Bjond, 11 janvier 2012 - 11:56 .

[Corker]

I suppose you could drop an artificial potential field (APF) over each member who wanted to participate in the phalanx; possibly you'd need to drop one on the terrain as well. IIRC, that sort of technique was recently used by some APL guys using a team of mixed UAVs and UGVs at ARL to do patrolling and whatnot.

Self-organizing swarms of autonomous agents is an active area of research in AI. Getting them to line up in some sort of regular matrix is problem one; getting them to move in concert is problem two; getting them to then do anything useful is problem three.

The technology may be ready for video games; there's a lot of cheats in a virtual world that roboticists don't get to use in the real world. But you'd still get some really weird, buggy AI behavior when the APFs overlap in unexpected ways. If you're willing to put up with that some percent of the time, you might be able to have your shield wall for the rest of the time.

http://www.swarms.org/ if you want to read more on the latest research in the area.

EDIT: I've contended that you ought to be able to set it up as a virtual proportional-derivative controller - essentially, connect each member of the shield wall with an imaginary spring.  But - especially if you want to have nonlinear effects like the wall breaking - that's a whole lot of ugly math.  And I don't have the slightest idea how it would map to animations.  My impression is that there's animations for "walk" and "run" and you can speed up "run" if you're hasted, but there's no way to represent fine-grained motion control (which is what you'd need if you were implementing a controller).

Modifié par Corker, 11 janvier 2012 - 01:36 .

[Bjond]

I feel like I've patiently explained how to play "catch" and some kid picks up the ball and hurls it 20 miles. Having real AI like you describe would be absolutely wonderful, but it's a lot more than I was imagining. It doesn't need to be a real wall or even resemble one -- just a teamwork ability for mitigating incoming DPS. I want the AI free to choose it's own target because until it can do so, it has no hope of ever being AI at all.

Using a shield wall would have to protect all members, too, not just those participating. In later warfare, one of the points of formations was to offer protection for the unarmored folk so they could cheerfully attack from range. Either way, it's obvious you understood exactly what I was talking about. ^^

[Corker]

*nods* I've been a sword-and-board, protecting our squishy spears behind me.

I suppose it seems kind of unfair to me, to maintain individual mobility while reaping the benefits of a shield wall. But I suppose if it were an armor or defense-buffing sustained spell that multiplied per caster, I wouldn't blink at it. So maybe that's not so bad...

[Tryynity]

I am little confused about how taunt/intimidate/bluff ruins AI - I apologise if I have misunderstood your post OP

I have played EQ2 and I have use the "pull aggro" mechanic as a fairy paladin (it is really very humourous if you think about it) to keep as much aggro off my Druid fairy friend LOL  (enemy can save)

In DDO I play an Intimitank (Paladin) - that keeps the Boss's attention so the casters can take it down. (enemy can save)

In DDO I also play a bluff rogue assassins, which can successfully pull whatever enemy NPC it wants from a mob from sneak, they come and then are assassinated. (again enemies can save)

I find this is tactical play - rather than whole party charge a mob and wipe.

In DDO you can effectively make shield walls - we block doorways and casters stand behind and nuke.  There is no save, except some NPCs have been known to break thru/jump over, but this is realistic.

To some extent these tactics also work in DAO - but with more fail rate than I like - but they are program controlled so its still pretty good, considering. I still prefer playing with NPCs like this than a random PUG in most MMOs I played.

Modifié par Tryynity, 12 janvier 2012 - 12:58 .

[Bjond]

Heh, it's OK.  Glad you read it and are trying to see my point.  AI is fundamentally about decisions.  If you prevent your enemy from being able to choose who to attack, you are eliminating the primary and most important decision it can make.  That's what taunt does.  It forces target selection on the enemy.

If you think about any real-world fight, the enemy is always free to choose it's own targets.  This is true even for games with PvP.  Players are never so dumb as to say, "Hey'all, let's try to kill the most heavily armored & least dangerous (dps-wise) target first!"  OK, sometimes they are, but they learn very quickly that is fatal.

With taunt, the choices you leave open to the enemy are just movement and skill (type-of-attack).  That's a pretty pathetic list of choices.  It makes every single fight sequence incredibly dull for the player and puts a huge burden on the developer.  It's really hard to make a creative boss fight when the most important choice your boss critter can make just isn't available to you.

--

BTW, the other tactics you listed have their own problems.  IMHO, "pulling" shouldn't work.  Even animals are too smart to fall for that and yet the classic game AI portrays supposedly human-smart enemies as being so dumb as to charge out in a solo attack instead of rounding up a bunch of buddies and stomping your butt into pixelated goo.  This is a slightly different topic than the taunt issue, but ultimately it devolves to the same thing: choice of target.  In the case of pulling, the critter obviously isn't being forced to attack you and only you, though.

The pulling issue is pure laziness on the part of the developers -- not a predisposition towards assuming it's the only way to design games.  It would be utterly trivial to make critter AI smart enough to flee fights it can't win -- something nearly every animal in existence will do.  This contrasts with every single critter in most every game.  They're all insane kamikazi who will attack even the most overwhelming odds and then fight to the death.

Whatever game-tactic works is perfectly valid and even fun in the sense of out-gaming the game.  I just think the mechanics used for this in today's games are seriously flawed.  It would be refreshing to play a game where the developers at least tried to make the enemies smarter than rabid squirrels.

[Tryynity]

I still kinda disagree - because all the NPC's I am used to have to roll a save to not be tricked/coerced etc...

re: Intelligence/Charisma/Constitution/Strenght/Wisdom/Dexterity score - I think that is fair.

Some endgame bosses you have to build your character to beat those with ridiculous scores.

The trouble is often the mechanic gets broken somehow - AC was useless at higher levels the last I played DDO - my top level rogue was in Monk Robes LOL i.e. zero AC from clothing like robes

Personally I prefer - tactics like pulling, bluffing, intimidate, charm, etc, as well as using the area around me in approaching a battle.

Ranging from high places - only casters and rangers can hit you back.
Hiding behind hard obstacles/barriers.
Finding ways around traps other than disarming.
Swimming underwater to lose people.

That makes it more realistic to me - I like to strategically plan out a battle.

What I hate is when a class becomes so riduclously overpowered they can solo high level content without the aid of extras.  But then the developers re-compute the scores and the class breaks for awhile.

RE:  AI intelligence in NPCs - DDO have developed this.  I do not know how, but the AI is way smarterer than it used to be.

If you have not played this MMO - I seriously suggest giving it a shot.  So far it beats every MMO I played.

I have yet to play SWTOR fully, it may top it in this regard, but I doubt it - I think their focus was on RPG which is fine by me, it is hard to incorporate EVERYTHING in a game (perhaps) .

DDO with all its greatness, once you have done all the Dungeons 1 mil times (which I have) until new content is released (which it is quite regularly now) it gets boring.  This is way I favour storyline over actual dungeons crawling.

Modifié par Tryynity, 13 janvier 2012 - 01:09 .

[Bjond]

For Corker: being part of the shield wall shouldn't be without drawbacks in a game sense -- and ideally they should "feel" reasonable; eg. reduce movement ability and dps for the character in the wall (using their "join wall skill") traded for increased damage mitigation for those just using their "benefit from wall defense" ability.  It would definitely need to have tradeoffs and require a "tough" decision by the attacker -- do I try to break the wall or overwhelm the squishy it's trying to defend (or do I just flee cuz it's too hard)?

----

For Tryynity: I don't regard a single dice roll (a "save") as an acceptable substitute for AI, but unlike taunt&agro the lures you're describing don't prevent AI.  They highlight decisions oportunities where AI is lacking, but it's not outright prevention like taunt.

Example1: Kid on playground versus your candy lure.  Kid might be distracted enroute to van.  (Chance for AI).  Teacher (neighboring critter) notices kid walking to van.  (Another potential chance for AI).  Van is really scary.  (Yet another chance for AI).  etc. etc..  My bet is that the only reason your lure is working in DDO is that the developers have zero AI at the reasonable decision points surrounding the lure.  They've basically lumped ALL those checks into a single dice roll.  However unlikely, they could still replace the dice roll with real AI at some point.  The game system they're using doesn't block them from doing so if they ever have sufficient desire and a (very) active imagination could conceivably fill in the gaps and explain the observable game behavior.

Example2:  I'm a dragon.  It's easy to lure me.  You're tiny and you look tasty.  Rawr.  I'll catch you my pretty and roast your little dog, too!  Whoa, where'd this guy in a fire-resistant suit come from?  No matter, that mage is more dangerous and I can just kill it with one burp.  Wait, fireguy is insulting me.  I hate insults, I'll breath fire on him.  Ouch, there's a rogue stabbing my ankle.  GRRR, fireguy is REALLY insulting.  I'll breath more fire on him first.  Hmm.  Maybe I should stop that cleric from healing fireguy.  Naw, I'm really mad now.  Fireguy must die!  Ugh, is that lake all my blood?  I'm dying, should I flee?  NO, leave my Momma out of this fireguy, you're really gonna get it now!

There's just no way you can translate any fight from a game system that uses taunt/agro/etc into even moderately believable prose no matter how hard you try to do so.  That's because taunt is an aspect of the game system that prevents developers from ever making an enemy smarter.  The dragon is always ludicrously stupid and can only do what the player, it's enemy, tells it to do.  This is why I think "agro" based game systems are built to fail.  They prevent any AI from ever being added and stress the suspension of disbelief well past the breaking point.

[Tryynity]

You are assuming the taunter is using only words - he could throw stones/sticks. It would be only a distraction and I see your point.

But AI is getting smarter and the developers are making it happen - there may be an understandable limit though.

Personally at least the taunt skill does not bother me, on how it affects AI - I am just happy when it works and I save the main DPS character, especially if I hit worse than a girl (I am a girl and so feel justified in using that statement LOL)

I would be just as happy if it wasnt there - but was replaced with intimidate, if that is more believable for you. Diplomacy is another, I like... my rogue talks the guy outta hitting her and as soon as his back is turned BAM backstabbed hehehe

[Bjond]

I prefer more realistic approaches.  Taunt, intimidate, bluff, etc. are all psychological controls.  Psychological controls only work before things get physical and they're highly situational to being with.  Once things go physical or even once people think they might go physical, your words are meaningless.  You could make a false show of a physical threat, but that should at most last for one round.  As soon as the target figures out he's been tricked (which will be right away), he'll revert to picking better targets.  I suppose it's really a matter of taste.  We're talking about games here.  How much realism do you want?  Games (and P&P RPGs) are all designed around epic-fantasy / hollywood principles where the hero wins no matter how absurd the situation really is. 

Your "talk your way" example is one of those unrealistic situations.  It only works when starting from an implicit assumption of trust; ie. I don't think your talky rogue should be permited a bluff check versus a suspicious target.  True deception of a suspicious target takes immense preparation and planning.  You can't just waltz up and shout "bluff check" in a guard's ear.  That's a GM or game failure unless you're on the hollywood set where the guard thinks he's getting lucky when the pretty girl comes up to vamp him while he's on post.

I've become increasingly bored and intolerant of the epic attitude,
myself.  You'd have to be exceedingly clever in your deception if you
want to fool me into thinking something is a threat when it's patently
not a threat.  In game terms, if you want believable encounters, you
have to grant the enemy as much understanding of the game-rules as the
player.  This is particularly important when you determine how the enemy
does threat assessment to pick it's targets.


The biggest barrier to adding real threat assessment is the concept of a "tank" currently embedded in developers and gamer's minds.  They see it as a single point of defense, which means tanking doesn't work and the entire class system breaks down if you don't have a way to force things to attack the target that should be attacked last.  If instead that thinking can change to more of a shield-wall / barrier provider, you can still keep the "tank / defense provider" mentality while adding the most critical AI feature back into games: choice of target.