[MUD-Dev] Player Manipulation of Environment

Marc Hernandez marc at ias.jb.com
Fri Dec 7 00:36:35 New Zealand Daylight Time 2001


On Wed, 5 Dec 2001, Ling Lo wrote:
> On Tue, 4 Dec 2001, Hans-Henrik Staerfeldt wrote:

>> I think that you simply need more genes to make your point
>> better.

>>    A damage gene.
>>    A extra attack gene (bite, etc).
>>    A speed gene.
>>    An agressive gene.
>>    A flee gene.
>>    A size gene.
>>    A magic gene for spell XX.
>>    A weaponuser gene.
>>    A item gene (example; carries a sword).
>>    A magic resistance gene.
>>    A good hide gene (armor+price).

>> I would say that for this to work, they should not be excluding
>> eachother (at least not all).

> There ought to be associated development and maintenance cost for
> each of these.  Developing an extra head to bite costs time and
> energy and more again to maintain that extra chunk of biomass.
> Just look at certain high maintenance crops.  In ALife simulations
> where ants have some traits that can be increased without any
> costs, it's hardly surprising to discover that the ants have
> optimal genes at the end of the simulation run.  (I mean, if the
> ant could either see 2 squares, 3 square or 4 squares then at the
> end of it all the ants can see 4 squares...  that's not precisely
> earthshattering.)

I wrote a simple bug simulation.  In it I had a gene that controlled
how old a creature could get, just to see if I would get characters
that lived the max.  I also had a fixed size array.

At first the longer living ones would of course dominate, then when
the population reached the end of the array I would not allow the
birth.  Eventually the creatures would fluctuate below the max
length of the array.  The bugs had optimized for fitting in my fixed
size array!

This was an awsome and unexpected effect of this kind of thought
towards characters.  You could use server load to optimize the
genes, as well as memory.

You could also breed creatures based on what ones survive battles,
in addition to modifying their genes.  Then if frost giants (immune
to ice) are plentiful and people start getting a lot more fireish
spells/weapons they would die out and be replaced by firemen.


Another possibility is to decouple the genotype and phenotype.  This
could have all sorts of nifty effects.  Using Dr. Koza style genetic
algorithms might result in nifty effects also (effectively little
evolutionary programs).

<various rabiit things snipped>

> How about borrowing buffered solutions from chemistry?  Create two
> species, rabbits (alkaline) and bigbadassmonsters (BBAMs, acidic).
> Rabbits eat lettuce, of which there's a lot of.  BBAMs are rare
> and pretty powerful but they prefer to eat rabbits because they
> taste like ambrosia.  If the players deplete the rabbit
> population, the BBAMs can't find the rabbits and go for the less
> tasty but more numerous players, giving the rabbits a chance to
> recover.  When the BBAMs are depleted, the rabbits undergo a
> population boom and flood the area.  This in turn feeds the next
> generation of BBAMs (who breed like rabbits when not hungry).

This reminds me of...  (Spoilers from ... I believe it was a Niven
book, beowulfs children or somesuch (although I think that was the
sequel) It is a big spoiler for the book)

  Taking a cue from science fiction, on Beowulfs world they had a
  species of incredibly powerful monsters.  In addition they had
  various other creatures including a vicious fish in the water.
  Well, after being attacked by these creatures they decide to hunt
  them down and kill them. A few months after they kill all the
  creatures they notice a whole lot of activity in the water.  A
  little while after that they notice a huge hoard of those
  monsters.  The monsters kept these fish down in population, but
  the fish eventually became the monsters.  Needless to say they
  were in a huge amount of trouble.

> Fudge the above to a ridiculous extent and cheat because the
> presence of players is a cheat anyway.  They're a foreign
> substance that gets added, so why not boost the ecosystem by the
> same?  I don't see why there is a need to maintain a precise
> simulation with each and every entity in the game represented
> explicitly.  Do a search for population containers.

> For the players to appreciate the effect of ALife, there needs to
> be a community mind on behalf of the players with easy access to
> observations of past events and effects.  In plain English:
 
>   [Radio] Bubba: I'm going into wyld bam country, anyone wanna
>                  come along?
>   [Radio] Buffy: Didn't Bobbo just kill all the rabbits?
>   [Radio] Bubba: So?
>   [Radio] Buffy: Means there's going to be some hungry bams
>                  running around, they're going to be in packs
>                  looking for meat, be careful.
>   [Radio] Bubba: Hrmm... I'll need help then.

> Isn't the environment the evaluation function?  I would think, in
> a system like this, to use some magic system which keeps all the
> genes created instead of discarding good work.  Give each agent a
> journal, track the progress, create some scoring system and when
> the mating season comes round, process the genes and dump the new
> critters into the environment.  The score can be calibrated, so
> too many high scoring agents, pick some of the weaker gene sets
> for the transmorgifier in that season.  The better genes can be
> kept in reserve to help in bringing around a faster response when
> there is a danger of extinction.
 
> Genes don't have to be physical traits, it can (and for games,
> should) include behaviour as well.  For a game of minmaxxers,
> there are fads of what is the best class/race/whatever combo, an
> evolving combat agent could be more resilient to such flux.  It
> would be cool if the players decide that rangers are the best
> class, a flood of rangers appear and the critters adapt to cope.
> Of course, a sufficiently interesting combat system is required in
> order to evolve anything interesting.
 
> The neat thing about genes in nature is that the result pops out
> because of the mess of interactions.  A set of genes determined by
> the designer will only lead to a set of results that are limited
> by the designer's thoughts at the time.  In any case, improvement
> of genes with each generation follows a log curve, there's a
> definite point of diminishing returns.

Is this going to be true in a dynamic system?  Given the meta system
of monsters that do good getting more airtime, there might be an
effect like a pendulum with the various creatures being plentiful
and scarce.  Add in genes for the individual monsters themselves as
well as interactions between the various creatures and you might
have a sytem worth investigating.

Throw in other activities like pet raising and such and you open
up a huge number of different axis of advancement (rather than killing
creatures).

--
Marc Hernandez
Two standards are better than one.
Working at Oddworld.com (and not speaking for them).

_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev



More information about the MUD-Dev mailing list