Complete.Org: Mailing Lists: Archives: freeciv-dev: May 2000:
[Freeciv-Dev] Re: AI (again!)
Home

[Freeciv-Dev] Re: AI (again!)

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: freeciv-dev@xxxxxxxxxxx (Freeciv developers)
Subject: [Freeciv-Dev] Re: AI (again!)
From: Reinier Post <rp@xxxxxxxxxx>
Date: Wed, 31 May 2000 13:00:19 +0200

On Tue, May 30, 2000 at 09:04:09PM -0400, ½²«íµØ wrote:
> I have seen enough arguing over the list on
> "this is unfair", "my idea will make the game more balanced", blah blah blah
> However, I find them rather inconclusive, more of personal gut feelings.

Of course they are.

> One way to scrutinize the rules would be playing them with adaptive
> learning AIs and see if they converge to different strategies.
> Current server AI is great, but I doubt if it can be modified for this.
> This is important before Freeciv can do away from Civ1/2 rules.

Agreed.  Perhaps a standard testbench system can be developed.
It would be a wrapper script that reads parameters from a config file,
including

  - which civserver versions to use (if a code patch is being tested)
  - which combinations of server settings to use (including rulesets)
  - which savegames to start from

then automatically runs all-AI games on all specified combinations,
and summarizes/wraps up the results.  With it we can do a serious
amount of easily reproducible testing.

Basically,

#!/bin/sh

REPORTS=/var/tmp/civtest

for civserver in `cat $1`
do
  for rcfile in `cat $2`
  do
    for savfile in `cat $3`
    do
      (
        DIR=$REPORTS/$civserver/$rcfile
        mkdir -p $DIR && cd $DIR && $civserver -r $rcfile -f $savfile &&
        summarize-it  # like the civserver.freeciv.org reports?
      )
    done
  done
done
 
-- 
Reinier



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