Error ratings of World Class backgammon players...

[Karl]

Hi i have been training myself with gnu for the last year and have now have reached the point where i am often rated at "expert" or occasionally above. I would like to know how this compares to the best players in the world but i cant seem to find any analysis of there games. Does anyone know a website or other method to find this information.

Of course i could just manually input the moves in gnu and see how they are rated but this will take a long time and i am sure someone has already done this :)


--------------= Posted using GrabIt =----------------
------= Binary Usenet downloading made easy =---------
-= Get GrabIt for free from http://www.shemes.com/ =-

[wintom]

Forget about the categories "expert", "worldclass" etc. Due to the fact
that you can manually set up the threshold values for these (and people
do this because the original thresholds are rather frustrating ;-) and
because the labels differ from the ones used by snowie) they are
difficult to compare. I would try the snowie error rate. Play a
25pointer against 0-ply (no tutor mode!!), analyse it with 3-ply and if
you stay below 5 you can call yourself a worldclass player. In a real
match over the board you might do a few more mistakes and end up at 7.
There are also a few people getting below 3 in tournaments (Tardieu,
Falafel et al). In order to get below 2 you would have to learn to even
make the mistakes the robots do, so better don't ;-)

I analysed only a few matches, so please correct me if I am mistaken...

Thomas Koch

[Unknown Entity]

I don't know of any sites where these are talked about in detail. You
can check places like Gammon Village where they have games with a
Snowie analysis. I remember seeing somewhere that Fran�ois Tardieu
said he had a Snowie error rating of around 3.2 if my memory serves me
right. I've also seen games with top players where the error rating is
lower but that's "just for one match".

Before going further, what type of games are you playing? That can
have a big impact what skill level shown at the end. For instnce, it's
fairly easy to get a low error rate if you're playing 3pt matches as
cube decisions are often "no brainers". If you're playing that type of
game, try 5pt or more. One other point, if you want to compare
yourself, is to make sure you don't show the pip count as that can also
make cube decisions easier. Sorry, just thought of another - At what
level does GNU play when you play it? I find that I play better the
higher the level settings. This is because my errors are punished
harder leading to more forced or almost forced moves with an eventual
loss. I feel (I have no basis for this!) that my error rate is
typically higher in games that I win.

Sorry, I don't mean to be pouring cold water on your achievements of
which you can be proud - well done, keep improving and keep having
fun!!

Post by Karl
Hi i have been training myself with gnu for the last year and have now have reached the point where i am often rated at "expert" or occasionally above. I would like to know how this compares to the best players in the world but i cant seem to find any analysis of there games. Does anyone know a website or other method to find this information.
Of course i could just manually input the moves in gnu and see how they are rated but this will take a long time and i am sure someone has already done this :)
--------------= Posted using GrabIt =----------------
------= Binary Usenet downloading made easy =---------
-= Get GrabIt for free from http://www.shemes.com/ =-

[Raccoon]

I remember seeing somewhere that François Tardieu
said he had a Snowie error rating of around 3.2 if my memory serves me
right. I've also seen games with top players where the error rating is
lower but that's "just for one match".

François Tardieu - "I recently sent recorded matches from the last
three years of live play to the Karsten Nielsen and Iancho Hristov who
are surveying seasonal players and listing their average results. They
have computed my average at 3.16 using Snowie 4 Pro, based on 50
matches."

François Tardieu - "There is no point in trying to play at a zero
error rate for a number of reasons - first you need to adapt your
cube decision according to your opponent's skill and tendencies, and
then, bots still make some errors that humans don't make - therefore
some plays bots tag as error can be considered negligible. I think it
possible for a hard-working human, under perfect conditions, to play
somewhere around an average of 2.2 and consider this the best one can
achieve on a consistent basis. This might take three to four years of
regular analysis and some hundred of matches played."

-- http://www.gammonlife.com/interviews/2006tardieu1.htm

[Robi]

interesting interview, thanks for the link.

I remember seeing somewhere that François Tardieu
said he had a Snowie error rating of around 3.2 if my memory serves me
right. I've also seen games with top players where the error rating is
lower but that's "just for one match".

François Tardieu - "I recently sent recorded matches from the last
three years of live play to the Karsten Nielsen and Iancho Hristov who
are surveying seasonal players and listing their average results. They
have computed my average at 3.16 using Snowie 4 Pro, based on 50
matches."

François Tardieu - "There is no point in trying to play at a zero
error rate for a number of reasons - first you need to adapt your
cube decision according to your opponent's skill and tendencies, and
then, bots still make some errors that humans don't make - therefore
some plays bots tag as error can be considered negligible. I think it
possible for a hard-working human, under perfect conditions, to play
somewhere around an average of 2.2 and consider this the best one can
achieve on a consistent basis. This might take three to four years of
regular analysis and some hundred of matches played."

-- http://www.gammonlife.com/interviews/2006tardieu1.htm

[b***@gmail.com]

www.gammonluck.com

here is the full rancking:
www.gammonluck.com
Most of the matches i recorded from gamesgrid,after this Karsten help
me,sending me matches of more players.But remmember 50 matches is
really not so much.....I have 800 matches of Falafel fro example in
some of them...last 50,he is under 3....so it is hard to say and to
create realistic rancking.Other some players's matches are from
live,other are from online.The Future of the rancking is not clear,it
is hard to get new matches from the players,they stop playing in
gamesgrid,and i think there is the one place that is possible to
record matches,and other to get matches from live is hard and u can
get no more then 5 matches per year for the player....Soon i will
create my own page,there i will do a small update,inclunding
MalcolmDavis,some other names and correction of some er.rates with new
matches.
Unfortunately i had crash of the Hard drive,so i loose some of the
datebase...so i can't update statistics for some of the names that
dont play online...

Post by Robi
interesting interview, thanks for the link.

I remember seeing somewhere that François Tardieu
said he had a Snowie error rating of around 3.2 if my memory serves me
right. I've also seen games with top players where the error rating is
lower but that's "just for one match".

François Tardieu - "I recently sent recorded matches from the last
three years of live play to the Karsten Nielsen and Iancho Hristov who
are surveying seasonal players and listing their average results. They
have computed my average at 3.16 using Snowie 4 Pro, based on 50
matches."
François Tardieu - "There is no point in trying to play at a zero
error rate for a number of reasons - first you need to adapt your
cube decision according to your opponent's skill and tendencies, and
then, bots still make some errors that humans don't make - therefore
some plays bots tag as error can be considered negligible. I think it
possible for a hard-working human, under perfect conditions, to play
somewhere around an average of 2.2 and consider this the best one can
achieve on a consistent basis. This might take three to four years of
regular analysis and some hundred of matches played."
-- http://www.gammonlife.com/interviews/2006tardieu1.htm

[m***@lycos.com]

Karl:

I am in a similar situation and seek a human player who is very good to
"spar" with once in a while. I'd like to play 7 or 9 point matches,
for free or small stakes - you decide. I play at GE, TMG, XG, and CG,
so if you want, let me know when you will be at a particular site and
what your username is. I think we could help improve each other's
games.

[k***@gmx.de]

There are many matches available in *.mat or *.sgf format. Open /
import them into GnuBG, analyze them... there you go. Very fine
collection of old and new tournament matches. I am sure you will enjoy
it:

http://www.hardyhuebener.de/engl/matches.html

Cheers, Klaus

Post by Karl
Hi i have been training myself with gnu for the last year and have now have reached the point where i am often rated at "expert" or occasionally above. I would like to know how this compares to the best players in the world but i cant seem to find any analysis of there games. Does anyone know a website or other method to find this information.
Of course i could just manually input the moves in gnu and see how they are rated but this will take a long time and i am sure someone has already done this :)
--------------= Posted using GrabIt =----------------
------= Binary Usenet downloading made easy =---------
-= Get GrabIt for free from http://www.shemes.com/ =-