[MUD-Dev] Cognitively Interesting Combat

[Paolo Piselli]

--- cruise <cruise at casual-tempest.net> wrote:

>   http://www.casual-tempest.net/combat/

> The main qualities I was aiming for:

>   Choice and strategy.  Variation - sometimes complex, sometimes
>   quick and simple.  Easy to learn, but hard to game, with no
>   guaranteed win.

I like this version much better than the previous once.  It has more
of a feeling that you are reacting to a situation than that you are
picking a random tactic.  The probabalistic nature of the outcomes,
along with the stiuation-dependant utility of each tactic, make it
impossible to just find some optimal move sequence and repeat it
over and over.

Returning to my idea about analyzing the interaction based on a
rule-based model, I some thinking about the goals and rules that I
was using as I played.  At my "novice" level of play, I found myself
using about 6 of the available moves at some point during the
combat.  Rarely did I ever know clearly which action to take,
because I was weighing the value of managing my balance against
dealing out damage.  This means that my model of the game has at
least 6 rules total, with around 2-3 under consideration for some
situation, and at least 3-6 rules fired per combat.

This quick thikning about a novice model for your game does not take
into account any rules for planning or goal-setting, which may be
present in the way a more advanced player would play.  It is also a
model born from a breif play session - the model could reduce or
expand with experience.

For instance, if I eventually learn that there is an optimal move
for any given situation, then the active set will reduce to 1 rule.
However, given that there are many possible situations (combinations
of my and my opponent's balance and health), I think that the total
number of rules fired per combat would then approach the number of
different situations per combat.

On the other hand, I may learn the utility of the 3 moves that I
never used, and thus my model would expand to include at least the 9
rules for the 9 different actions.  Additionally, I could learn more
about conditionally managing high-level goals such as "maintain
balance" and "damage opponent" - this would require more rules for
selecting which goal is more important at a given time, then which
action to select based on the saliences of my various goals.

But IMO the really important thing is the "asymptotic" model -
i.e. the cognitive model that a player will converge on over time as
he learns the game.  My assumption is that "easy to learn, hard to
master" makes no difference, because combat will be repeated so many
times that the player will reach a very high level of expertise.
Thus the "asymptotic" model must have a certain amount of
complexity.

On a side note, I've found some professors who are willing to advise
a thesis on something like this :)

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