Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!tank.news.pipex.net!pipex!warwick!news.dcs.warwick.ac.uk!not-for-mail
From: N.Parekh@dcs.warwick.ac.uk (Neelesh)
Subject: Re: AI in Soccer Games
X-Nntp-Posting-Host: nop
Message-ID: <1995Aug22.124348.9223@dcs.warwick.ac.uk>
Originator: N.Pa@nop
Sender: neepar@nop (Neelesh Parekh)
Organization: Department of Computer Science, Warwick University, England
References: <1995Aug9.193917.27920@dcs.warwick.ac.uk> <40skdn$ek2@mirv.unsw.edu.au>
Date: Tue, 22 Aug 1995 12:43:48 GMT
Lines: 66


In article <40skdn$ek2@mirv.unsw.edu.au>, fraser@cse.unsw.edu.au (Fraser Rohan Tully) writes:
|> Neelesh (N.Parekh@dcs.warwick.ac.uk) wrote:
|> 
|> : Does any one have any ideas on what sort of AI a soccer game could use?
|> Ascii graphic:
|> 
|>                    x1-----------> o       
|>                     .    y     .
|>                      .      .
|>                       .   .
|>                        ..
|>                         x2
|> 
|> key: o - ball
|>      x - the x team
|>      y - the y team
|>      -----> path that player follows
|>      ...... path that ball follows
|> 
|> The first x player would need to detect that in order to advance the ball,
|> he will have to go close to player y. The player then makes a decision based
|> on <player statistics> on whether or not to dribble past y. If its too risky
|> then the one-two might be a valid alternative. The move might consist of:
|> 
|> 1. Check that there is a clear path to x2
|> 2. Pass to x2
|> 3. (While ball is travelling?) tell x2 to get the ball, but hold onto it, x1
|>    wants it back!
|> 4. Run past y, keep running till there is a clear path back to x2
|> 5. x2 Sees this path, passes to x1.
|> 6. X2 tells x1 'Hey, Come get the ball, I'm passing back like you wanted'.
|> 
|> This sort of stuff isn't easy, probably why I haven't seen a soccer game


hmmmm.... seems like an interesting problem.... although i would tend to try to
solve it in a mathematical sense as opposed to the test-and-set method
described above.

Well, from the diagram (nice one fraser...) it appears that the two participating
players should be manouvering themselves to keep the opponent in the center of
this triangulaton. Once the x players have been forced to retreat the ball is
passed to the second x plater who immediatly forwards the ball to the calcuated 
position for a rendezvous with the first x player. This of course is a set play
by members in a team and would have been reheased many a time and so can be
relegated to matrix transformations remapping rather than adding an inference
engine to accomodate for these features.
 
What do you think? This could system could also be scaled and rotated to
allow long passes from midfield into the box for the striker. This sort of
passing is that made Gary Linakar (sp) famous for getting the goals, its seems
very similar to the assist rule played in Ice-Hockey.

I've probably over-simplified the mathematics in this situation somewhat but i
don't think its unreasonable.

BTW, has anyone seen ACTUA-SOCCER by Gremlin? I've seen a rolling demo but
haven't played it to assess the AI. Looks great but waht does it play like?

	Nels.  
-- 
_______________________________________________________________
Neelesh Parekh			Department of Computer Science.
neepar@dcs.warwick.ac.uk	University of Warwick. England.
_______________________________________________________________
