"Magik" <rickglover DeleteThis @paulhastings.com> contributed wisdom to
news:1134535483.465521.233560@o13g2000cwo.googlegroups.com:
> Randomness is definitely a factor.
>
> Also the amount of time that a computer opponent has to process their
> turn would determine the skill level. An novice AI player may only be
> allowed to complete its turn in 1 minute while an expert level AI would
> have 10 minutes.
Some of the greatest capabilities of PbEM play are things that I feel have
been under-utilized. Since players have all day to do a turn and their
turns are processed by a host, many limitations built into other games for
the same of speed can be bypassed. So it allows for epic sized maps, more
players, more options, more game-changing random events, and a lot more ai.
I would love to see a game that actually takes me a day to make all of my
turn and the hosting runs on my computer all night, as long as it makes use
of even ONE of those features. Of course VGAP4 does allow for more players
than most games and some pretty large maps so it is making good use of the
PbEM.
> Also, AI decisions should be put into a database to keep track of what
> it is doing from turn to turn so it continues on a particular target
> plan over the course of turns.
Thats one way but the multi-level randoms I spoke of were another way of
handling that. The races randoms are rolled at the beginning and used as a
new random range for each action. So one race might be running in one game
with numbers always at the high end and in a small range, and in another
game at the high end in a large range, or low and low, or low end and high
variable, etc.
> Another way to simulate some randomness
> there is to randomly delete one of those entries in the database, so
> that the computer player "forgets" what it was doing and has to process
> that again.
Agreed. Especially in games that support different styles. If a game
supports ai that are barbaric swarm, and ones that are defensive
researchers, and ones that are agreessive expansionists, etc. Then a great
option is for the computer to be able to realize that its present tactics
are losing ground, and reselect. If I find a defensive researcher and start
picking away at their defenses while stealing their tech, its a big shock
to have it suddenly get pissed off and become an enraged swarm. Its also
alot of fun if the game allows for the insane emperor who just randomly
reselects everything at any random moment.
> In my opinion, AI actions need to be put into priorities based on race
> and game situations. Those that are highest take precidence in getting
> processed and put into the long term plans database. The rest may not
> get processed due to time constraints.
Ive always liked that. I would love to see the bird men be treated
different than the lizards. And then still have some random variables
within that. But I highly suspect that only a game supporting player
scripted races would ever get there. Something like what Space Empires IV
(and soon V) did. I love the fact that the game went from 8 races in the
distributed disk to being over 300 races with portraits, ships, and tactic
styles (and its still growing). I dont think vgap4 will get there. Maybe
vgap5?
As for time contraints, yes it would be nice if the ai chose the fast and
dirty actions, and then went back if there was time to rethink and
consider. A setting for the max run-time for the host would be sweet. Id
set it for fast and stupid while I was learning then give it more time to
"think" as I progressed.
Gandalf Parker