[computer-go] Where and How to Test the Strong Programs?

[Michael Alford]

I have lurked here for a long time, I find this newsletter very 
interesting, because I play Go, I am not a programmer. I agree with 
everything Mr Foltand had to say, but would like to add a small bit 
about IGS ranks. A few years ago, IGS changed its ranking system so as 
to anchor from the top, ie a 9d on IGS was a 9d pro, making it the 
closest to "real world" rankings. I think the strength of IGS ranks has 
slipped a bit since cyberoro came online, but is still probably closest 
to "real world" ranks. I would suggest not using AGA or KGS or any other 
  ranking system for establishing your CGOS ranks, recruit some IGS 
players with solid (over 100 rated games) ranks, and have them play with 
your various engines.

back to lurking,
Michael

David Fotland wrote:
> I think AGA and KGS are pretty close.  AGA is a "real" rating system in that
> ratings are earned in sanctioned tournaments so they are not disrupted by
> casual games.  http://www.usgo.org/ratings/default.html  
> 
> European ratings (also from tournaments) are perhaps 2 stones tougher.  Many
> think they are more reasonable, since most feel the top of the amateur
> rating should be 7 dan or lower, and top AGA ratings are higher than that.
> Top pros that have participated in AGA tournaments have ratings about 10
> dan, and there are many amateur 8 dans.
> 
> Japanese ratings are less tough.  Japanese amateur ratings can be purchased,
> with a test, so there has been inflation, and I don't think there is a
> national rating organization like in USA and Europe. 
> 
> Korean and Chinese are very tough, since they think a 1 dan amateur should
> be close to professional strength.
> 
> So, I'm AGA 3 dan, but I would have a tough time playing as a 1 dan in
> Europe, and I play at 4 dan in clubs in Japan.
> 
> I tried playing in a club as 3 dan in China once, and got totally crushed.
> 
> My preference would be a scale that is fixed at the top, with 9 dan pros at
> 9 dan.  This would put 1 dan pros at about 7 dan and top amateurs at 6 dan,
> with a few 7 dans.  This is pretty close to the European scale.  Top
> tournament pros almost never lose to pro 1 dans, so there are 300 or more
> ELO points between the top amateurs and the top professionals.
> 
> So perhaps top human play is 3500 or more on the cgos 19x19 scale.  That's
> 12 ranks above 2000, with the higher ranks having more ELO points per rank.
> 
> David
> 
>> -----Original Message-----
>> From: computer-go-bounces at computer-go.org [mailto:computer-go-
>> bounces at computer-go.org] On Behalf Of Don Dailey
>> Sent: Tuesday, December 11, 2007 6:37 PM
>> To: computer-go
>> Subject: Re: [computer-go] Where and How to Test the Strong Programs?
>>
>> I feel that we probably need several more players to have much
>> accuracy,  but I don't mind starting the best educated guess we can
>> muster - it can be modified at a later time.
>>
>> How do AGA ratings compare to other systems?   Is any particular system
>> considered (defacto or otherwise) more of a "standard" than some other?
>>
>> How do AGA ratings compare to KGS?
>>
>> - Don
>>
>>
>>
>> Christoph Birk wrote:
>>> It looks like my (3k AGA) CGOS rating (tast-3k) is converging around
>>> 2000 ELO. That gives us a zero-point but we need at least one more
>>> rated player (better more) to get the scale.
>>> If you would like to use my GUI please contact me by private email
>>> at ccbirk at gmail dot com.
>>>
>>> Christoph
>>> _______________________________________________
>>> computer-go mailing list
>>> computer-go at computer-go.org
>>> http://www.computer-go.org/mailman/listinfo/computer-go/
>>>
>> _______________________________________________
>> computer-go mailing list
>> computer-go at computer-go.org
>> http://www.computer-go.org/mailman/listinfo/computer-go/
> 
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>