From: lwp@conch.aa.msen.com (Lou Poppler)
Date: Thu, 28 Dec 1995 18:45:32 +0800
To: cypherpunks@toad.com
Subject: Reputation capital: FIBS case study
Message-ID: <2sY4wMz2BcRC083yn@mail.msen.com>
MIME-Version: 1.0
Content-Type: text/plain


I am currently engaging in a reputation adjustment of a certain player
on FIBS, the First Internet Backgammon Server (see my homepage).
Reputation is important on FIBS, and I thought it might be informative
to examine the ways it can work in the [more-or-less] real world.

FIBS allows various levels of nymity, and various modes of reputation.

Each player picks a pseudonym as his FIBS name.  These can range from
the crypticly anonymous nym, to some variant of the player's real name.
The server will also disclose the FQDN from which the nym is connected,
or was most recently connected.  If she wants, the player may include
any information, typically such as an email address, in an "address"
field in her player profile.  Or something cute, or nothing.  Players
may also choose to have a web page listed under their nym, on a page
full of such listings.  Players may participate in the newsgroup
(rec.games.backgammon), identifying themselves in their posts by 
any or all or none of {their FIBS nym, their real name, their email
address}.

Reputation has several components on FIBS.  Of course, personal 
experience with another nym is hugely significant, when it exists.
When that is absent, or not conclusive, there are other sources of
reputation which may be consulted.  The server reports two
numbers for each player, representing distinct reputation criteria:
the "rating" is a weighted score representing the won/less record;
the "experience" is a count representing how many games this player has
played.  Before I mention other reputation indices, I want to amplify
on these two.  When negotiating with another player about possibly
starting a new match, and about the length in points of the potential
match, most players will think about various aspects of the reputation
of the prospective opponent.  Depending on the circumstances, the
difference in ratings may be decisive, in straightforward ways, for
evaluating an invitation.  Of particular relevance to past discussion
on cypherpunks, the experience number is usually very important also.
Those with a high experience level, and perhaps a name that you recall
having seen before, are trusted more than nyms with little investment
in their reputations.  Experience matters a lot, because one of the
key strategies is to avoid becoming involved with unpleasant or 
dishonest players.  If someone has a huge experience number, you can
be fairly sure that you would have heard about it if they are somehow
undesirable.  

As FIBS is an open server, operating without charge to the users,
we get a wide variety of newcomers every day.  Some persist, and
become better known, and others lose interest and don't return.
Most are [more-or-less] agreeable and [more-or-less] ethical; however
some nyms will exhibit certain recurring problem behaviors, and quickly
begin to accrue negative reputation.  On common problem is the use of
profanity or the advocacy of various forms of bigotry, in "shout"ed 
comments or private messages, and the closely related hounding of 
players with female-sounding nyms by unwanted come-ons.  This behavior
is generally countered with "shout"ed and "tell"ed replies, and with
the "gag <name>" command which kills your copy of any further remarks 
from that user.  The negative reputation resulting from this behavior 
is pretty much instantaneous and self-documenting, with reinforcement 
from anyone who is enraged enough to remember and to heckle the offender 
at a later login.  Some of us are more sensitive about this than others,
so the offensiveness is just another variable in reputation, not always
the decisive one.

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).  Such players follow a strategy of playing winning
matches to completion, which increases their rating, but bailing out of
losing matches before they end.  When a match is interrupted, the server
remembers it, and the two players can resume later, by mutual agreement.
Almost all players are ethical enough to resume even losing matches, and
eventually to complete them.  The problem folks are the ones who ignore
or refuse requests to resume.  Avoiding such players is a key strategy
for serious FIBS regulars.  One indicator here can be a complex function
of the rating and experience numbers: a player with fairly low experience
but an unexpectedly high rating is one of two things -- either a very
strong player, or else a match dropper.  (Or else very lucky, but we
prefer to think of that as strength of play.)  The best defense we have
found against the match dropper is complaining in the newsgroup.

I just posted a warning in the newsgroup about a new match dropper which
I had the misfortune of playing against.  I accepted an invitation from
him last Saturday, to play a 3 point match.  I looked over his info
before I accepted, and decided to give him a chance: no email address,
experience = 57, and rating showing slightly more wins than losses, from
a site at a college in Britain.  I won the first game for 2 points, and
was pretty clearly winning the next game, which would make me win the match.
When "dorion" rolled a game-losing double five, he immediately dropped
his connection to FIBS.  The server saved the match, and I waited around
for awhile to see if he would come back.  He didn't.  Now at this point,
I'm not yet ready to start in on his reputation.  After all, we all have
host problems, and phone problems, and personal problems.  It's still 
possible that he and I will meet later, and finish the match.  (I'm 
skeptical, because the timing of it is just too convenient, but I hold
my tongue).  Finally Tuesday, I was hanging on FIBS when dorion returned.
I asked politely if we could finish, but dorion immediately logged out
without replying.  I stayed connected to FIBS, while I did some other 
work on other VC-s.  

Before long, dorion was back, my little watchdog macro beeped for me, 
and I again politely asked him to resume.  Instead of replying, 
he started a fresh match vs. a brand new player, experience
equal zero.  I started watching their match, where I asked again if
we could resume sometime.  While I wasn't getting an answer, I looked 
into the new player's information: connected from exactly the same
site as dorion!  This of course, is a warning sign of the other common
form of ratings cheating -- creating two nyms for one person, and 
playing against oneself, with one nym always winning and the other
always losing.  Well, again I got no reply to my requests, and both
nyms suddenly disconnected without finishing their match.

At this point, I set about trashing dorion's reputation.  I posted
a (PGP signed) article to the newsgroup setting forth the facts.
When I came back on today, there was one followup from another player
who had a similar experience with dorion.  When I signed onto FIBS,
I got a "tell" from a trusted, veteran nym saying that dorion and his
alter-ego nym BatesMotel had just been on, and played a couple of 
one-sided and not at all credible matches vs. each other.  I checked
their ratings/experience and followed up to my own article with a
definite advisory against playing against either nym.

Now, despite many elaborate proposals over the years, discussed at length
in the newsgroup, there is no automated mechanism for dealing with this
sort of negative reputation.  I have a personal "don't-play" list I keep
up to date, as do many others.  My complaint surely got dorion and
BatesMotel into many such personal lists, and neither nym is likely to
last on FIBS.  

I'm not really asking for suggestions here or anything.  It's likely that
most of them have already been debated to death on r.g.b.  I just wanted
to show you a case of reputation markets in action.

::::::::::::::::::::::::::::::::::::::    Thank you VERY much!  You'll be 
::  Lou Poppler <lwp@mail.msen.com> ::   getting a Handsome Simulfax Copy 
::      http://www.msen.com/~lwp/   ::   of your OWN words in the mail
::::::::::::::::::::::::::::::::::::::   soon (and My Reply).