RE: [Bug-gnubg] Computing winning chances when self and opponent'sexact Error rate mEMG's are known

[Ian Shaw]

Adi,

 

There was some recent discussion about this at bgonline.

 

It started as an offshoot from this thread,

 

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=66264#67042

 

Resumed here:

 

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=66724#67113

 

and is currently being discussed here:

 

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=67052#67052

 

which contains a a table of Match Winning Chances vs. Elo difference.

 

Note that most of the discussions use Snowie error rate, which is usually about half of the gnubg error rate.

 

There seems to be a consensus that 200 Elo points is about 5.5 Snowie ER, or 1.0 Snowie ER is about 36 Elo points.

 

I assume that you are aware of the Jacobs/Trice MET tables in gnubg, which account for differences in Elo points.

 

I think that amongst all this is the information you need to put together what you were looking for. I hope it helps.

 

-- Ian

 

---

From: address@hidden [mailto:address@hidden On Behalf Of Adi Kadmon
Sent: 16 May 2010 11:37
To: address@hidden
Subject: [Bug-gnubg] Computing winning chances when self and opponent'sexact Error rate mEMG's are known

Hi all,

 

I"d like to have a formula, or a very detailed table, by which I may be able to compute the theoretical winning chances in a single game or in a match against every opponent when both his/her and my exact "Error rate mEMG" are known from the accumulated measures in GNU's database.

 

If such a thing is available, I'd be grateful to receive it or a link to it.

 

Thanks in advance

 

-- Adi