Complete.Org: Mailing Lists: Archives: freeciv-dev: October 2003:
[Freeciv-Dev] (PR#6566) Re: Re: Barbarians
Home

[Freeciv-Dev] (PR#6566) Re: Re: Barbarians

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: undisclosed-recipients: ;
Subject: [Freeciv-Dev] (PR#6566) Re: Re: Barbarians
From: "Christian Knoke" <chrisk@xxxxxxxxx>
Date: Sun, 19 Oct 2003 05:10:25 -0700
Reply-to: rt@xxxxxxxxxxxxxx

On Sun, Oct 19, 2003 at 12:21:08AM -0400, Ross Wetmore wrote:
> 
> I think part of the practical problem is whether you have one barbarian
> player type (for name, icons, behaviour weights, etc), or 10 barbarian
> players (fresh one for each spawning), or 80 barbarbarian players since
> in reality each unit should pretty much function on its own, scattering
> to cover the maximum amount of territory and pick off the most stragglers
> and weakly defended targets they can. But then individual barbarian
> players do not want to attack each other or form alliances either.

Well, the latter is a good description of what I have in mind. :)

> There really should be almost no coordination between barbarian units,

Why not?

> with the possible exception of a leader and one bodyguard. And this
> means a lot of the AI code that coordinates attacks, does statistical
> evaluations of health/cost/worth, etc is just not pertinent. A barbarian
> should cover as much territory as possible, and attack almost anything
> it runs into regardless of the odds until it runs out of hit points or gas.

Each barbarian horde can be conrolled by a seperate AI player. All barbarian
AI players are teamed. They are relatively weak because they are few and
they are scattered. They need not to be weak by representing them with weak
units, though. Give them average strenge.

When you play with normal AI players on the same continent from the
beginning, they have a good balance between aggressiveness and coordinated
settling and development, up to a certain point. When I play gen 1, I *have*
to see for my defence. That makes me think that normal AI can be a quite
good implementation for barbarians, more than the current one. But I can be
mistaken. We'd have to test it and compare current barbs with AI barbs, of
course.

Computer power shouldn't be the problem.

I'd think (as a non-coder, please forgive me), if you  set up minimal
restrictions for barb AI's that might be necessary, and if you keep those as
simple as possible, then you'll even end up with an easier maintainable code.

Christian

> Thus, much of the AI code is useless, a lot of the rest needs to be
> tailored specifically for barbarism, rather than (collective) sanity,
> and only a minimal thread running through it all is likely common code.
> 
> The key question to resolve is whether to include barbarian behaviours,
> or lack of, as optional elements in the range of any player's personality,
> and have the options checked as one resolve each player's activities, or
> one optimizes the checks out and has a high level split into a barb and
> non-barb activity path.
> 
> The latter, of course means that many elements will need to be duplicated
> and the divergence over time of the two relatively independent branches may
> become a problem (i.e. a coder didn't bother to update the barb path so
> they start losing out on new abilities like teleport unit handler code or
> something). But there is no reason why a normal player shouldn't be able to
> drop a bunch of suicide units off in another's back field, and turn them
> loose to do as much automated damage as possible, i.e. to go into auto-barb
> mode, like auto-settlers.
> 
> There are tradeoffs plus and minus either way :-)
> 
> Cheers,
> RossW
> =====

-- 
Christian Knoke            * * *            http://cknoke.de
* Dear M$, I know your OS sucks, you really don't need to  *
* tell me 10 times a day ...                               *
* * * * * * * * *  Ceterum censeo Microsoft esse dividendum.



[Prev in Thread] Current Thread [Next in Thread]