Complete.Org: Mailing Lists: Archives: freeciv: August 2003:
[Freeciv] Re: FreeCiv
Home

[Freeciv] Re: FreeCiv

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: freeciv@xxxxxxxxxxx
Subject: [Freeciv] Re: FreeCiv
From: Horn Gábor <Horn.Gabor@xxxxxxxxxxx>
Date: 07 Aug 2003 13:40:33 +0200

Hi!

There were some conversation about this topic. You are right, ai should
be harder, but it's not easy to code at all. If u take the "classic"
stages of game, (expansion,exploration,invasion) u can see:

expansion: this is the most important maybe, and the hardest to code.
The problem is it's rather "intuitive" how to settle well, it almost
would need an artifical intelligence to settle as a good player does.
First we have some basic rules (in 95 percent only build settlers till
mapmaking, don't irragte till full island is settled, etc). But the good
settling also need other things: correct judgement between few movement
and avoid being crowded w/ the subsequent settling, recognize of
possible shortcuts, recognize of science bonus ("horizontal,vertical" vs
diagonal - especially important in pre republic stage). This problem is
similar to judging/refeering a map, and it's even a hard task for
humans. Maybe we should apply some kind of fuzzy logic here to be able
to measure each "interesting" field (eg the ones in 5 filed distance  - 
that's usually the maximum distance which might worth to move to w/o
settling) due to multiple factors w/ different weight. When u decide if
u settle a field or not u generally consider the following factors imho:

1, distance of it
2, special on it
3, special usable from it
4, "logistic" factors (can it be a shortcut? can it produce extra
science? will it block me from founding other good cities later? might
it be a "stronghold", eg an inner city from which i can liberate the
near coastal ones?)

And note 2 and 3 aren't independent factors, as there r powerful
combinations (as wheat+whale or wine+whale) which should be recognized.
(eg wine itself is not that good field to settle on imho, but almost a
must-to-do w/ whale in pre-republic)

Maybe if we could provide a kind of algorithm or table for the devs
about how to settle well, they could work on it.

2: exploration: rules r a bit easier here imho. If u can see islands w/
explorers only, check if the shared resource is already occupied (if
possible), and explore them asap. Do 2 distance trireme exploration
around your island. When u have nav explore as much directions as u can
to find your little expansion islands asap.

3, invasion: not easy to do good. But there are a few simple rules
again:

1, first attack the coastal cities w/ best ship u have. bring 2-3
attacker ship TOGETHER and 1 w/ horsemen on it.
2, for cities w/ defense bonus (on hills, forest, mountains) u might
consider useing dipo instead of military units.
3, try to protect veteran units, withdrawn them and heal if injured.
4, later protect the transoprts w/ bodyguards if possible.

Maybe applying these would be enough for first step.


note anyway but there are other things to decide for the ai to play
"good" or challanging againts a human player. Most important is the
wonders. AI really handles them interesting. Here r some basic rules
here imho:

1, only build wonders in non coastal cities
2, don't change where u build it every 2 turns :) (ai often do that,
which is only good for wasting prod. Maybe it "rethinks" every turn and
selects the most optimal city at that turn? It's a mistake, once decided
it shouldn't have to be changed)
3, focus on important ones: if u have prod/sci lead in most turns till
nav try to get magellans. If u r poor and has more players in the game
(eg not duel) try great lib. If somebody reaches magellans when u tried
it, do another useful one (pyramid,s maybe  suntzu a few turns later, or
the j.s.b/michelangelo's). 
4, try to do wondrs in 1 turn. So stock the needed numbers of caravans
around it, and when u have enough then do it in 1 turn.

And of course it's not the end, there're other aspects: diplomacy, techs
questions (research trade or nav first? go for happy wonder's techs?
when to stop research and go for war?), etc. Well, it's a complicated
game :)

bye, hirisov


> Hello everyone,
> 
> I have a suggestion for any newer version of FreeCiv,
> 
> there should be more difficulty levels... perhaps the current ones should
> be 4,5,6, on a scale from 0 to 10
> 
> there should be an option determining the aggressiveness of AI
> civilizations.
> 
> 
> Is it just me, or is using 31 (AI) civilizations on Easy impossible to
> beat?
> 
> 
> 
> 




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