From: fc@all.net (Fred Cohen)
Date: Fri, 29 Dec 1995 02:43:06 +0800
To: weidai@eskimo.com (Wei Dai)
Subject: Re: Reputation capital: FIBS case study
In-Reply-To: <Pine.SUN.3.91.951228011912.27043B-100000@eskimo.com>
Message-ID: <9512281414.AA21403@all.net>
MIME-Version: 1.0
Content-Type: text


> On Wed, 27 Dec 1995, Lou Poppler wrote:
> 
> > The thorniest problem in our reputation economy continues to be the
> > case of the player who drops out of a match when clearly losing, to avoid
> > the decrement of his rating number (based on match results only, not on
> > individual games).  [......stuff deleted...]  The best defense we have
> > found against the match dropper is complaining in the newsgroup.
> 
> It seems to me the easiest way to solve this problem is to list for each 
> player the number of games he dropped and didn't finish along with his 
> rating and experience.  Why go for elaborate social solutions when a 
> simple technical solution exists?
> 
> Wei Dai

It seems to me that not finishing a game is the same as knocking over
the board.  It's a loss for the player waiting to move, and a non-game
for the other player.  This should solve the not-finishing-a-game
problem in short order.

For the problem of playing a fake unrated player, try variations on this
scheme.  Track the players each player plays and reevaluate all players
scores in relative terms.  To get a rating, players must engage in games
with enough other players to form a valid statistical basis - at least
100 games with rated players for 10% accuracy. 

	Start with provisional ratings in the 1-10 range based on
	comparrison with other players regardless of who.

	As players play other fully rated players, add a temporary
	rating based on relative performance and post both relative
	and 1-10 ratings.

	When players reach 100 games within the rated group, they get
	an official rating.

-> See: Info-Sec Heaven at URL http://all.net/
Management Analytics - 216-686-0090 - PO Box 1480, Hudson, OH 44236