[Freeciv-Dev] Re: AI strategy

[vze2zq63@xxxxxxxxxxx]

Raimar Falke wrote:

> On Tue, Dec 04, 2001 at 04:44:44AM -0800, Raahul Kumar wrote:
> > --- Raimar Falke <hawk@xxxxxxxxxxxxxxxxxxxxxxx> wrote:
> > > 3) multi player with civbots on the clients without human player
> > >    (AI-vs-AI): no problem (there is the problem that the human player
> > >    may help the civbot ;) )
> > How?
> In a civbot-vs-civbot game it is possible that one of the civbots is
> really a human player. But you can catch this because you can request
> the source (or a binary) of the civbot and can test it independently.
This should be allowed.

That is to say, if it's possible to unhandicap a civbot by giving them 
extra bonuses (everything in the enumeration), it should be possible to 
unhandicap a human player in the same way.
It is possible that one of the civbots is really a human - but there 
would be little point in cheating this way, since you could never claim 
credit for your victory (for instance, you'd never make it into the 
rankings).  Really, though, a civbot fired up by a player might as well 
be a human - everything about it, down to the source code, is controlled 
by them.  What is needed to avoid this is have the server control the 
startup of civbots - but this brings us back to the question of 
efficiency, where the "civbot" implementation is much slower than a 
server-side AI since communications must go over the network protocol.
Here's another problem: if we handle handicaps as above, and an AI 
player (civbot or server-side) is replaced by a human player, does the 
human player inherit the unhandicaps?  Even in the current situation 
where that doesn't happen, the AI has an advantage in the very early 
game (say, first 20 turns at least), so it would be to the advantage of 
a human player to join a game 20 turns in.
I'm really not sure what the correct solution is here.

jason