(Editor’s
Note: Every so often the developers behind the upcoming fantasy MMOG
The Chronicles of Spellborn take
a moment to write out what they’ve been working on and the
type of gameplay we’ll find in their game. The Ten Ton Hammer
staff was fortunate enough to attain their latest dev journal, which
focuses entirely on the complex mob behavior found in Spellborn.
It’s a fascinating look at developer theory, so please enjoy
the article!)



By El
“Selachii” Drijver



Although the title might suggest that this is purely an AI related
journal, I’m going to stick to the game design side of the
subject. Or at least, try to.



I’m also going to try and not just give the situation of
Spellborn, but rather talk about possibilities and why I think some
choice are good or bad.



Goals

href="http://tth.tentonhammer.com/modules.php?set_albumName=album196&id=TCos_feb_2007_1&op=modload&name=gallery&file=index&include=view_photo.php"> src="http://media.tentonhammer.com/tth/gallery/albums/album196/TCos_feb_2007_1.sized.jpg"
width="200"> src="http://tth.tentonhammer.com/modules/gallery/images/pixel_trans.gif"
alt="" height="1" width="1">

style="font-style: italic;">When they were
initially creating the mob behavior for Spellborn, the developres came
up with the goals of their team.

First some goals must be set, what do we want to achieve when a player
fights a non-player (mob). Should the mob fight back or maybe run away
in fear? If there is a fight, what should the player experience?



A challenge? Okay, but easier said then done, challenge because? Mob
hits hard or mob is hard to hit, mob heals itself or does the mob have
a couple friends waiting around the corner? Besides, making something a
challenge doesn’t necessarily make it fun. So, the goals will
have to be thought through and might evolve a bit during the early
stages of the project to achieve the desired
‘vision’ which they all contribute to.



Combining all these goals will actually decide a large portion of the
core gameplay so getting them down on (digital) paper is of the
essence. Goals should be clear and simple, so everybody can easily
understand them, yet they should cover every aspect and leave no
questions unanswered.



As an example I will handle a (well known) goal set for Spellborn.



Never should a player be just as effective with his hands in the air as
on his mouse, keyboard.



Working with Goals



The mentioned goal sounds pretty easy, but in the world of
MMO’s this means that all old fashioned stun/polymorph
effects aren’t allowed. We could decide to allow these types
of effects and add a counter mechanic, so players have something to do
while their avatar is incapacitated. Maybe some button mashing
mini-game, popular in a lot of console games currently, to break out of
the effect early.



However, we prefer to remedy the cause, rather than the effect, so
players will always be able to do something. Slow downs are still
present, but no complete root effects. Players might get disarmed but
then there’s always magic and the other way around. Another
popular mechanic is ‘fear’, our equivalent are the
numerous ‘teleporting’ type skills, which displace
a target across a short distance, effectively disrupting aim.

href="http://tth.tentonhammer.com/modules.php?set_albumName=album196&id=TCos_feb_2007_2&op=modload&name=gallery&file=index&include=view_photo.php"> src="http://media.tentonhammer.com/tth/gallery/albums/album196/TCos_feb_2007_2.sized.jpg"
width="200"> src="http://tth.tentonhammer.com/modules/gallery/images/pixel_trans.gif"
alt="" height="1" width="1">

style="font-style: italic;">In making complex,
dynamic monsters, the developers decided that they wanted to advance
the current AI found in most MMOGs.

Goals might influence subjects you wouldn’t expect after
giving them a quick thought. This goal single-handedly makes sure that
big bosses will never be allowed to stun a player to make it harder for
the rest of the group, let alone stun entire groups for phase
transitions (not counting cut-scenes).



Implementing goals, making them work.



So, we have our complete set of goals, now the game designer will have
a good chat with the AI designer. This set of goals will have to be
translated to an AI machine which can make choices, which should result
in the effects the goals want to achieve…



What if a goal said that the combat experience should be
“dynamic and realistic”? Really dangerous terms and
they sound like something players will want. Dynamic as in, mobs will
pick the weaker target during the fight, make sure there’s no
healer in the back, finish off players with low health, ignore players
who are good at dodging and so on. Realistic because you want a bear to
be a bit slower then a tiger, but have more biting strength. Swords to
cut and maces to break bones, but where’s the limit?



I’m sure a couple of you readers now think “cool,
yes I want that, of course, who doesn’t”, yet most
MMO’s you enjoy work with very transparent aggro systems.
Having these complex ‘dynamic’ and
‘realistic’ systems in place, combat could become
very chaotic, unpredictable and especially, not a fun player experience
at all.



So it needs to be translated to something transparent and somewhat
predictable. Players want be able to anticipate moves, create
strategies, want to know what kind of reaction(s) they can expect from
their actions.



In general, players enjoy it if they can achieve real control over the
fight. This doesn’t mean that the behavior should be dead-on
like:



-‘hit the mob,
get aggro, mob won’t go anywhere else’.



But rather an enhanced version of these simple mechanics.
It’s one of those cases where less can indeed be more, in
terms of player experience. So, let’s apply that on the above
behavior.



-‘hit the mob,
get aggro, mob will go to another target if previous target keeps being
healed’.

href="http://www.tentonhammer.com/node/13596" title="adinthalos"> src="/image/view/13596/preview"
width="200">

style="font-style: italic;">According to El,
fights should only be predictable on  a basic level.

Although this is ‘dynamic’, it’s far from
the poohah-dynamic AI you read so much about on game box covers and
feature lists, but will in fact add a great deal of gameplay.
Let’s add another little simple rule.



-‘hit mob, mob
checks which nearby target is the weakest and aggro’s, mob
will go to another target if previous target keeps being
healed’.



Simple additions which can already cause countless of
‘dynamic’ combat situations. Because these rules
are rather basic, players will be able to predict the behavior after a
couple of encounters, so they can be expanded with more of these basic
rules. Some designers prefer to add chances here and although I think
some use of ‘chance’ can make the experience a lot
better, excess use of it will make the fight chaotic and eventually
destroy the fun all together.



So, should fights be predictable? The basics, yes, to a certain level,
even though it makes the situation not so super-dynamic-high-realism.
It will create a better, more manageable user experience and players
will be able to estimate what can happen…to a certain level.




This subject is far from covered completely, but I’m going to
close this journal now. I hope my message came across and maybe even
sparked some interest. I’m open to well constructed questions
and opinions, which I might answer or reflect upon in a following
journal.