[computer-go] So many MoGo run on cgos 9x9

[Hideki Kato]

Thank you Don,

Don Dailey: <47081235.2040506 at cox.net>:
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>I don't have any reason to believe there is a problem.    The rating
>system is not based on how well you do against the anchor player, it's
>just a device to prevent long-term rating drift.   If the pool of
>players were to inflate or deflate for any reason, the Anchor will not
>drift with them and so WHOEVER plays the anchor will get corrections
>which will propagate to all the players.

Conceptually, I agree.  But apparently the ELO values of highly rated 
and longly playing programs such as MonteGNU are less these days than 
previous.  Do you have day-by-day rating of a program?  It may support 
my observation.  

Of course, as this onfusion is temporal and will converge in a 
few weeks I guess, it may not be a big problem.  It, however, would 
be better to not throw so many new programs into cgos in the same time 
for everyone anyway.

>I have long considered setting up a lower-end anchor.  To do this, I
>would let some fixed weaker player play for several weeks and then
>average his rating over time to arrive at a good guess.   It should be a
>substantial number of games to be accurate.  I think it takes about
>50000 or more games to get within a couple of ELO points if we were
>doing straight performance ratings, but I would setting for less - it
>wouldn't be that critical for a low level player (and we could even
>adjust it later based on more games.)

Don't you try the same for high-end anchor using released version of 
MoGo? 

-gg

>- - Don
>
>
>Hideki Kato wrote:
>> Hi Don and all,
>> 
>> There are many variants of MoGo run on CGOS 9x9 
>> now  # http://cgos.boardspace.net/9x9/standings.html
>> 
>> I guess they force confuse the rating system because MoGo is the 
>> strongest, programs have a match more frequently against MoGo than 
>> anchors and get rating decrease in average.  If we had many anchors 
>> running this would not happen.  Someday, in theory, it will converge a 
>> balancing point but it may take so long time.
>> 
>> As I'm not familiar with the match-making and rating system of cgos, I 
>> may be wrong.
>> 
>> gg (Hideki)
>> --
>> gg at nue.ci.i.u-tokyo.ac.jp (Kato)
>> _______________________________________________
>> computer-go mailing list
>> computer-go at computer-go.org
>> http://www.computer-go.org/mailman/listinfo/computer-go/
>> 
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v1.4.6 (GNU/Linux)
>Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
>iD8DBQFHCBI1DsOllbwnSikRAgP6AJ0TCdF1/bWVTUT6lUhzpamRwvsc3wCfSZuc
>iD0x/AU9o8Zu+i1kiOYA8XY=
>=NWUA
>-----END PGP SIGNATURE-----
>_______________________________________________
>computer-go mailing list
>computer-go at computer-go.org
>http://www.computer-go.org/mailman/listinfo/computer-go/
--
gg at nue.ci.i.u-tokyo.ac.jp (Kato)