Complete.Org: Mailing Lists: Archives: freeciv-ai: June 2004:
[freeciv-ai] Re: Coments on AI
Home

[freeciv-ai] Re: Coments on AI

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: freeciv-ai@xxxxxxxxxxx
Cc: Per Inge Mathisen <per@xxxxxxxxxxx>
Subject: [freeciv-ai] Re: Coments on AI
From: Ross Wetmore <rwetmore@xxxxxxxxxxxx>
Date: Sun, 13 Jun 2004 14:43:36 -0400


You are highly unlikely to find solutions that cover all eventualities. So the best you can do is try to identify common scenarios and rules that make a reasonable outcome most of the time.

Rules like the last defender should never attack might be simple first approximations. An attacking vs defending type might choose differently if it could assure a victory and next turn city survival because its odds highly favored the attack.

In any event, one should compare the chance of winning a battle vs defending against the attack on the next turn as a choice for a city defender, and not just a one turn one-on-one.

Setting up a system of defenders (with increasing numbers falling under the last defender rule based on counting attackers) and one or two attackers (since attackers can heal in a barracks on the off turn, thus two attackers means one full force attack per turn), might be an interesting compound rule or strategy scenario to develop and apply to key cities.

Even better is to code simple routines that check for a number of conditions and/or compute things like attack parameters, code them as callable macros and let the end-users experiment with developing their own attack and defend rules. Different gamers with different rules might prove quite a useful testbed from which the default game rules could be extracted and builtup over time.

Cheers,
RossW
=====

Per Inge Mathisen wrote:

On Thu, 10 Jun 2004, Christian Knoke wrote:

Since the commit of the rampage_want and related functions, the ones
that allow an armour to attack an objective two-three tiles away in the
same turn, the units defending the cities use to attack more often. I
think this is bad!

You can always have two units to attack a city. This first will die
or survive, but the second will conquer the city, because the defend is
damaged or has dead in the previous attack.

And worst, they often attack even though the attacking unit is in a
hill or forest!!

In a game 1-2 weeks ago I saw this more aggressive habit, and I thought this
not being too bad. But I also saw that good defended cities do this until
they are *empty*. The city should keep at least some good defending units.

Even if attacking gives a better chance to win, this is not necessarily a
better chance to survive: to overcome a defender you need two moves, one to
destroy it, and another to enter the city.


Which is the better strategy for one-to-one unit combat can easily and
definitely be calculated. The server autoattack patch did this (except it
did not take into account power-reduction-due-to-movement).

However, things get more complicated when there are or may be multiple
attackers...  More ideas on what the AI _should_ be doing are welcomed.

  - Per




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