[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: computer-go: Go4++ IGS ranking
A problem for computers on IGS is that most people challenge stronger
players, not weaker, and they often take fewer handicaps than the
ranks would indicate. So the program has some upsets with weaker players
at the wrong handicap, which drags down the handicap, but no corresponding
upsets with stronger players. If the programs challenged players 1 or 2 ranks
stronger it would make a big difference I think.
david
At 10:05 AM 3/15/2002 +0000, Michael Reiss wrote:
Following on from the recent discussions about rankings on IGS
I decided to analise my most recent 115 games played there.
The discussions about rank have often been in relation to
the first game so what I have done is to only include one game
from each opponent - the one played earliest (which still may
not be their first ever game, because they might have played
before the 115 I have analysed). I have only monitored results
in the range 14k* to 18k*. I have ingored any games lost on
time by either side because it is too difficult to work out
if this was because of escaping or network problems or problems with
my IGS connection code etc. I have also assumed komi is 5.5 for
all games even if it was played with komi at 0.5 which IGS
sometimes sets (I know this is a bit dodgey but at this level
I don't beleive that so many players are modifying their play
according to such accurate counting). Ther results are as folows:
14k* no games played :-(
15k* 2/4 = 50%
16k* 4/5 = 80%
17k* 11/14 = 79%
18k* 7/9 = 77%
Here are the results in more detail:
IGS name kyu result
tiredboy 15 -
Fumiko 15 +
cslater 15 +
osX 15 -
yyyikes 16 +
JUN 16 +
nonny 16 -
rogluz 16 +
yehqimi 16 +
MoMo 17 -
GENMYO 17 -
imoti 17 +
jerk 17 +
ZINZIN 17 -
mtgo 17 +
Marron 17 +
yehqimi 17 +
jlf 17 +
diter 17 +
jaylee 17 +
blue 17 +
luchao 17 +
tatsuhiro 17 +
HCS 18 +
ink295 18 +
blowy 18 +
ZixZixMa 18 +
mn2763 18 +
jefflin 18 +
charbeneau 18 -
mat21 18 +
kogi 18 -
Mick.
David Fotland