home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!sunic!chalmers.se!etek.chalmers.se!fy.chalmers.se!smidt
- From: smidt@fy.chalmers.se (Peter Smidt)
- Newsgroups: rec.games.go
- Subject: Re: Internet Go Server Status.
- Message-ID: <1993Jan21.152030.15620@fy.chalmers.se>
- Date: 21 Jan 93 15:20:30 GMT
- References: <1993Jan15.004720.13397@adobe.com> <1j728cINNmo2@darkstar.UCSC.EDU> <4fKMaTm00gpm5B5ep0@cs.cmu.edu>
- Organization: Chalmers University of Technology
- Lines: 38
-
- In article <4fKMaTm00gpm5B5ep0@cs.cmu.edu> Wilfred.Hansen@cs.cmu.edu writes:
- >jonathan@cats.ucsc.edu (Jonathan Cano) writes:
- >> While we are on the subject of the automated rating system I have a
- >> few questions about the algorithm used.
- >
- >I've posted earlier versions of a rating algorithm, but the version below
- >has some new comments. I believe this algorithm is similar to that
- >used by both the IGS and the American Go Association.
- >
- >The algorithm does
- >
- >> Have you posted the formula that the rating algorithm uses?
- >
- >Perhaps Tim will respond with an estimate of how close the algorithm
- >below is to the rating system in use on the IGS.
- >
- >-- Pseudo-code for a ratings algorithm
- >
- > ( 280 line algorithm follows )
-
- My theory is that the algoritm on IGS is wrong in some sense. I have
- got some very odd results when playing/loosing against other
- players. I'm almost certain that the algoritm must have some fault,
- because I've heard of player that have won a lot but NOT got
- a positive rating and vice versa.
-
- I'm also very suspicious to the fact that the rating is update
- after approx 1 day. I've played matches and my rating have NOT
- change or change in a very strange way in relations to my match
- results. I think that the 1 day delay is made to hide that the
- algoritm is wrong. I've mailed and asked why the delay is present
- and got no answer...
-
- /Smidt
- --
- +=======================================+
- "The whole valley is like a smorgasbord."
- -- TREMORS
-