Re: [computer-go] Statistical Significance

["Robin Kramer" <robin@xxxxxxxx>]

I think there is a reason why the masters of a game only ever win by a
couple points, and it is not "promoting the game".

I don't know about you, but when I play with a particular group of people
and have a successful fuseki, I have to worry about playing against that
fuseki the very next game.  So, avoiding tournament play to improve, then
every once and a while playing in a tournament, I can imrpove faster
relative to the general tournament audience, which is what counts in
tournaments.  The problems of course being self assesment, and trade
secrets.

http://www.apa.org/journals/psp/psp7761121.html  <- this is an interesting
article mostly unrelated to go.

Sincerely,

Robin Kramer (alias KoDream)

>
> Yes,  what I  have  seen  many times  is  that a  player  has a  great
> tournament and  is afraid to play  again, knowing that  he is probably
> overated.
>
> But people  who are  rating conscious are  very often afraid  to play.
> There is the constant fear that your rating might go down!
>
> ELO  ratings   are  extremely  fine   grained.   1  rating   point  is
> immeasurable  and even  100 points  isn't much.   That's why  I always
> thought it would be interesting if players had 2 ratings.  One of them
> would be internal to the rating organization and would be based on ELO
> but  wouldn't be  the published  rating of  the player  (and  would be
> basically hidden.)  The published rating would be based on the highest
> ELO rating ever achieved and  organized into classes.  Sort of like in
> ELO chess where 2200 or  above is considered "master."  There might be
> classes defined by ELO 100 rating point differences.
>
> The  idea would be  that you  would always  be trying  to move  into a
> higher class  but would never  have to fear  dropping down to  a lower
> one.  You would get to keep your achievements.  I think there are lots
> of games/sports  where you  achieve rank based  on an  accumulation of
> accomplishments and can't lose it  once you have achieved it.  This is a
> bit more  compassionate and human.  Of course  players sometimes get
> weaker but that doesn't apply so much in intellectual games as it does
> in  sports where  youth is  a  much bigger  advantage.  Older  players
> sometimes  get significantly  weaker over  time  but to  me the  older
> players who have presumably contributed much more to the game over the
> years deserve  the dignity and  status of keeping whatever  title they
> have earned.
>
> If such a  system were used in Chess, there would  probably need to be
> some adjustments to the basic formula.
>
> How is this in Go?  Can your rank go down in Go, or do you get to hold
> on to ranks you have achieved?
>
> - Don
>
>
>    >Last time I checked (which is many years ago), the Danish Go
> Association used something like the ELO system to calculate Go
> ratings. They had it calibrated so that the numbers corresponded to
> kyus and dans pretty directly. I think higher dan rankings were
> still awarded by a committee, based on the calculated numbers...
>    >
>    >No idea how it performed over a longer time, nor indeed if they
> still use it.
>    >
>    One serious disadvantage of the Elo-System is, that "before the
> Elo-system, people played chess, now the play for getting Elo
> points". If I ask a good friend of mine: "How was the last
> tournament, his answer is: "I gained/lost X Elo".
>    It has become a fetish. But this is general accordance with the
> development in society. "How is your work?". "I earn XY Euro/Dollar".
>    >From this perspecitive the Dan-Ranking has something
> oldfashioned/charming.
>
>    Chrilly
>
>
>    _______________________________________________
>    computer-go mailing list
>    computer-go@xxxxxxxxxxxxxxxxx
>    http://www.computer-go.org/mailman/listinfo/computer-go/
>
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx
> http://www.computer-go.org/mailman/listinfo/computer-go/



_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/