[computer-go] cgos

[Don Dailey]

The website is out of date and it says so at the top of the page.

The pairing system was updated many months ago to try to schedule games
more frequently.  If you have a fast program playing another fast
program, you may get 2 or 3 games in while another pair is playing a
single game. 

However, I agree with you on this.  I believe it is better to schedule
ALL the games at once among all the available players.   This will
greatly improve the "fairness" of the pairings because the server can
make the best choices at once instead of being constrained by who
happens to be available.

I need better terminology.  The current server schedules in
"cycles" (which I think are 30 seconds apart.)   This means new games
can start every 45 seconds.   The proposed algorithm would schedule in
"rounds" which means that all pairings are decided at once, and the next
"round" does not commence until all the games are finished.  

You are correct about the slower K-factor change.  It is only a single
constant in the program.

For anyone interested, I would like to note that the K-factor change is
not constant per game change.  It decreases in proportion to the
opponents certainty.     The rationale is that if an opponent has a very
uncertain rating, there is no reason to justify a big K-factor change.

- Don


On Mon, 2006-09-18 at 14:24 -0700, Christoph Birk wrote:
> On Sun, 17 Sep 2006, Don Dailey wrote:
> 
> > So the question is, should I go for discreet rounds of play which will
> > greatly ease the ability to pair players of similar strength and still
> > provide variety, but will decrease  the number of games your program
> > will play, especially if it plays fast?  The decrease in number of games
> > won't impact the slow programs very much, but will impact programs that
> > play fast - but will enable "fairer" scheduling by criteria 1 and 2.
> 
> I am confused :-)
> Your website states that pairings are done in rounds already.
> And that's my experience from running programs at CGOS.
> It's also the better way (IMHO) to do the scheduling.
> After all, the better programs are more important.
> 
> About the rating system:
> I think Anchorman/ControlBoy are doing a good job, but I would
> like to see the K-factor change slower. That should be an easy
> modification (and not require a new server).
> 
> Christoph
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/