[Freeciv-Dev] Re: the contents of savegames (was: Preparations for other AIs and humans)

[Stan Shebs <shebs@xxxxxxxxxxxxxxxxx>]

Reinier Post wrote:
> 
> On Mon, Jan 15, 2001 at 03:27:34PM -0800, David Pfitzner wrote:
> > *2«íµØ <iquin@xxxxxxxxxx> wrote:
> >
> Couldn't a good AI depend *mainly* on the state of the current game
> for its performance?  What you'd lose is history information
> ('2 turns ago there was an enemy city here with 3 musketeers').
> 
> How bad is this in practice?  Would people force restarts just
> to have an easier time against the AI?

In Xconq, there is one crucial piece of information to save
for AIs, which is the AIs's own initial strength estimates of
its human opponents.  This is because the AI will resign if
it thinks it's losing, the heuristic being when its opponent
(including allies) is 4 times stronger than at the beginning
of the game.  Since random games can be started with various
handicaps, and scenarios may have all kinds of different force
ratios, if the AI doesn't record its initial estimate, it may
either resign instantly after a saved game is restored, or refuse
to resign when its cause is clearly lost.

(Incidentally, there is a command to reveal all of your units
to another side, one of whose purposes is to show your enemies
that there is no hope.  Of course, if they're not as weak as
you think they are, you've just made a critical error, heh-heh.)

It may also be useful to record unit spottings, so if an AI
sees a Fighter pass by once, it knows that one of its enemies
has achieved Flight, can use that to raise the priority of
countermeasures.

Stan