[MUD-Dev] RE: Understanding Simulation (was: Point of View)

[hart.s at attbi.com]

[Ted Chen]

> Comon', that's like saying "I meant to do that."  Of course you
> can flag anything as "successful" if you change the
> specifications/requirements for it.  That's not very useful for
> avoiding that same situation in the future.

I didn't advocate making broken models for some purpose and then
relabeling them. I said that what is a broken model for one purpose
is not a broken model for another. I won't dispute that there are
broken models and non-broken models, given some set of requirements.

Really it is only that perfect safety from extinctions, big
oscillations, etc. is not a universal requirement.

I'll unequivocally grant that, changing very little else about a
game like AO, if you just dumped in even a single-pop. logistic
growth model it would die:

  - That model can't recover (in terms of size) from zero population
  (although if you triggered an exception and spawned some or some
  cheat like that, it would just be bad because supply would
  generally be low, providing many players distributed near the
  critters - it wouldn't quite be useless if you wanted to reward
  players for going to poorly explored areas, for example).

  - NPC populations, both absolutely and relative to player kill
  capacity, are generally small. Maybe that's due to technical
  issues (polys on screen, cycles on server, memory, blah blah).

  - Players are generally permitted to kill npcs quickly. If it took
  a few hours to stalk and kill each critter, a lower birth rate
  would be closer to acceptable, smaller populations would be closer
  to acceptable... On the other hand, who wants to spend a few hours
  killing an NPC in a genre which is all about bagging tons of NPCs?

  - We don't even need to talk about adding predators to such a
  scenario.  Sheesh.

Of course, as has been discussed before, you don't need to have each
individual around explicitly, you could spawn in individuals at some
rate proportional to the size of the population, keep critters in
some (relatively) inaccessible reserve, etc. That could let you have
big populations relative to player kill rates.

Of course, that's not necessarily an improvement over spawning.  If
you have 100 billion goblins growing on an S-curve, and they get
spawned in (say) at a rate proportional to their basically
untouchable population size, you've set parameters so that your
model is (functionally) a spawn. So it doesn't make sense to do such
a model if you are not actively interested in letting players at
least bash the population down a touch (in which case you can adjust
parameters such that players spending 2% more time could remove the
goblins.)

The strength of that kind of model is not that you get really
anything for free, I'd say rather that it gives you parametric tools
for accomplishing a range of possible effects. But you may or may
not even want any of these effects. In the big games I don't think
it would help at all - there you probably want monster supply to be
a close approximation to monster demand, so no one goes away unhappy
that they didn't get a monster, but also so you don't have to spend
too much computation on critters that aren't doing any good. On the
other hand, having *local* changes in population size is no big
deal, and supports players moving around. But the most appropriate
structure in the typical design is probably going to be a thousand
variants on the spawn. And that's just fine, within the scope it's
appropriate to.

> Seeing a population go extinct is just as informative as seeing
> one last forever.  In MMOGdom, it's mostly yes.  You'll be hard
> pressed enough to come up with different species to populate your
> ecology.  Having to watch your work evaporate as they go extinct
> one at a time is not a very fun thing to do.

Yes, it would definitely be a bad idea to spend thousands of hours
doing animal art, and then set it up to be rendered unusable in a
few hundred hours. Of course, you could reuse art, or implement
procedural changes (e.g., switch to green and blue wolves). But
that's not necessary at all. If we are ready to contemplate
"respawning" individuals at the same locations, I think it's not so
bad to contemplate "respawning" whole species, e.g. in remote areas,
and allowing them to migrate.  I don't think it would be so bad to
have to at least pretend the goblins came from somewhere in
particular :)

Come to think of it, I seem to remember that some games have already
streamed in new critters, or at least slight variations, on a more
month to month basis. If you can work up a system under which
extinctions are relatively controlled or predictable, you could in
fact add new content from time to time (assuming you're doing such a
game. It wouldn't be too time consuming in a text game or using
stupid tricks like recoloring, retexturing, rescaling). Not too
bizarre for a monthly-subscription service to release new content
periodically (that's not intended as a cheap shot or anything).

> I do think ecology systems are doable in code.  However, I also
> get the impression that a majority of 'ecologists' on this list
> simplify /or ignore the dynamics.

Hmmm.

> The rest, already have an inkling of what components might be
> causing the dynamics (eg. predators) to avoid them.  But that's
> wrong too as that throws the baby out with the bathwater.

Well, I've often said that predators aren't really an appropriate
way of controlling population sizes. Part of what makes them
exciting is that they allow more sensitivity to interesting events
like wolf extinction by the players. On the other hand, that
sensitivity usually translates into disaster for games in which
everyone needs wolf meat. So maybe even the baby needs to go for,
say, Anarchy Online. :)

It's certainly possible to be judicious about what disasters you
allow, e.g. provide a functional alternative to wolf meat and let
the wolves be mostly for show and tell, cool scenery.  Or with
regard to predation, you can throw atmospheric predation in without
making the all-important rabbit population subject to fluctuations
in the wolf population.  That way I think you can get the big
automated fictional happenings without the dire functional
consequences for the players. I don't think it's throwning out the
baby because you do have to protect certain key resources in the
game (such as xp in games which favor it).

I've certainly questioned whether there's even a point to doing it
if the critters involved don't matter to players (e.g., if they
don't give xp). I think the answer is that they do matter, but not
for the same reasons - high-caliber immersion is in the details, and
objects, even game objects, have more existence and value than XP.
Will players quest to save the wolves? Who knows, but I don't think
it's my business to ensure that everything that happens should be
urgently important. Models and dynamics are interesting to me
because they change, are subject to control, may be unpredictable,
may be complex. To the extent that the simulation has input into the
availability of key resources, that control is often going to be
"exploits," the unpredictability is going to be a reason to leave,
and the complexity is going to make managing parameters hell.  Tie
the simulation to places where variation/control are tolerable or
even desirable and I think it becomes much more functional.

> In either case though, talk still degenerates into "if we use this
> lego-block with this lego-block, tune it, then it'll work"
> variety.

You're right, but anyone who gets down to working with said
lego-blocks is going to figure out the trouble spots pretty
quickly. (I assume!)

Not to imply that there's anything wrong with pointing out said
trouble spots to others.

> There's a grand difference between "including a factor" and
> "including a factor that improves the system," and that's where I
> hoped the System Dynamics references might be of some use to
> people.

Yes!

> And besides, I don't think human interaction nor interaction with
> other species is that much of an unnecessary demand.

Not at all, provided that the other properties of the world even
vaguely support the consequences of this interaction. At least,
that's the problem I've been trying to emphasize.

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