bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Re: from money/dmp to match


From: Joseph Heled
Subject: Re: [Bug-gnubg] Re: from money/dmp to match
Date: Sun, 23 Feb 2003 09:26:58 +1300
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021202


This is what I get from using the FIBS rating formula

wins 71.07% 1ptrs - implied rating difference 780

Fibs formula for D=780

match length
3     5    7    9     11    13    15
17%  11%  8.4%  6.3%  4.8% 3.7%  2.9%


This is a far cry from (say) the self play 15-ptr rate of 11.43

I conclude that the FIBS rating formula is indeed loaded against the stronger players, and this may explain why the BOTS are having "good periods" (when playing more against high rated players) and bad ones.

-Joseph

neilkaz wrote:
I don't know about FIB's ELO formula but I assure you all that GamesGrid's
is not accurate for longer matches between lopsided opps.

To explain further on GG. I feel that if I play one pointers vs much weaker
players I need to win 3/5 or even 5/8 to break even and that actually
doesn't seem to be too unreasonable.

However, vs the same much weaker players I might need to win 85% of my 9
point matches to break even which just isn't possible, IMO. This is also the
reason I rarely play rated matches vs anyone much more than 200 points less
than me. I felt the rating system used by KG's list was more accurate for
matches, but GG changed it to have more points at stake per match.

..neilkaz..
----- Original Message -----
From: "Sonic.net" <address@hidden>
To: "Albert Silver" <address@hidden>; "'neilkaz'"
<address@hidden>; <address@hidden>; "'gnubg'" <address@hidden>
Sent: Saturday, February 22, 2003 1:45 PM
Subject: Re: [Bug-gnubg] Re: from money/dmp to match



For my guess I would just use the FIBS rating formula,
but estimate the rating edge as something between
that implied by 1- and 2-point matches.   Something
like 650 rating-points better for the first pairing,
and 170 rating-points better for the second.

This is obviously not a very precise model.  For
example, I have another pairing where the stronger
player wins 68.13 (cpw, not dmp, so slightly more
than the 67.98) but only wins .52ppg, rather than
your .57.

David





_______________________________________________
Bug-gnubg mailing list
address@hidden
http://mail.gnu.org/mailman/listinfo/bug-gnubg







reply via email to

[Prev in Thread] Current Thread [Next in Thread]