I heard about a pro comment that the Great wall and similar openings are quite
feasible, but also quite easily thwarted. If you opponent plays it, you just
prevent him from completing it by taking the last point yourself. This should
give him an inferior position. If you let him complete it, he
Don Daily wrote:
so for instance a professional player with a relatively low professional
ranking does not need as many stones
as indicated by his opponents ranking to beat another professional WHO IS
SEVERAL RANKS HIGHER.
For centuries pro ranks were defined by the handicap required against
I am 3k on KGS and I'd say that I win nearly every game that I
give handicap and lose over half the games that I receive handicap. That would
seem to imply that more handicap should be given near my rank.
That is because the KGS rating system is not based on handicap. It is
Hi Phillip,
Thank you for your help and explanations :)
Dave
Van: computer-go-boun...@computer-go.org namens Philipp Hennig
Verzonden: za 7-3-2009 23:50
Aan: computer-go@computer-go.org
Onderwerp: Re: [computer-go] re-CGT approximating the sum of subgames
Hi Philipp,
Thank you for replying. I've seen your article before, but I read it only
superficially then. By the way, I couldn't find any online source about the
Thomson Heuristic. Do you know if ther are any?
Fistly, I want to use ownership and local correlation statistics from light
Thank you for helping, but there seem to be many Thompson heuristics.
But I didn't find the one that Philip is referring to:
W.R. Thompson: On the likelihood that one unknown probability exceeds another
in view of two samples. Biometrika, 25:275-294, 1933.
vs
Clark Thompson: A Greedy
Ok, I'll look into your temperature discovery article.
I noticed that Amazons is mentioned a lot in CGT, but I am not familiar with
the game.
I want to use montecarlo to discover subgames. Then I want to use CGT
techniques to evaluate and sum subgames to get a full board evaluation.
If this is
Are you saying that
- calculating the exact sum is too expensive?
- calculating the exact sum does not give the the best move?
- the existence of Ko in go makes it impossible to split the board in
independent subgames?
A subgame may not be a number in many cases, but using canonicalforms
Or is it it a problem to extract the best move when the sum of subgames is
known but not equal to a number (which will often be the case)?
Van: computer-go-boun...@computer-go.org namens dave.de...@planet.nl
Verzonden: wo 18-2-2009 20:04
Aan: computer-go
A simple alfabeta searcher will only get a few plies deep on 19x19, so it won't
be very useful (unless your static evaluation function is so good that it
doesn't really need an alfabeta searcher)
Dave
Van: computer-go-boun...@computer-go.org namens George
I agree with you, but I wouldn't qualify MC evaluation with MCTS as a static
evaluation function on top of a pure alfabeta search to a fixed depth (I have
the impression that this is what George Dahl is talking about).
Dave
Van:
I've been looking into CGT lately and I stumbled on some articles about
approximating strategies for determining the sum of subgames (Thermostrat,
MixedStrat, HotStrat etc.)
It is not clear to me why approximating strategies are needed. What is the
problem? Is Ko the problem? Is an exact
I think this estimate is a reasonable educated guess. The uncertainties are
quite big.
I would say your estimate has a total margin of error of at least 50% (it will
probably take between 15 years and 50 years) but I don't think it's possible to
estimate much more accurate at this stage.
Still, in the unlimited class of a rally from Bagdad to Beijing, it is probably
not the dragracer nor the solar car that wins, but the landrover with good road
maps and a GPS.
I agree with most posters that in the end, you have to find the best thing to
do the job. Hardware could play a major
Just a remark, no advice: the suicide rule, the ko rule and the score
depend on the rules under which the game is played.
Dave
Van: computer-go-boun...@computer-go.org namens Ernest Galbrun
Verzonden: za 10-1-2009 22:01
Aan: computer-go
Onderwerp: Re:
Michael:
Let's say that active Pros should have 2800+, though players
with 2750+
might still be professional strength.
I think by that definition there would be many players with a professional rank
who wouldn't have professional
What you are saying is that many professionals are overrated or underrated
(sometimes by as much as two stones). The same goes for amateur ranks too.
So a rank estimate from a series of 7 stone games against a 4p will still have
a error margin of one or perhaps two stones.
I agree with that.
I think a 7 stone handicap against a 4p would be normal for an EGF 1d, not for
a japanese 1d.
A japanese 1d is about 3k EGF. He would require more than 9 stones.
Dave
Van: computer-go-boun...@computer-go.org namens Darren Cook
Verzonden: wo 17-12-2008 2:48
When the japanese audience stated that CrazyStone was playing like a 4d or 5d
they were talking about japanese ranks.
This suggests that it played like a 1d EGF or 2d EGF according to the audience.
Dave
Van: computer-go-boun...@computer-go.org namens
I was in Japan for a year in 1990. I was 1k - 1d EGF at that time and i was 4d
in Japan. I think the grade difference between EGF and Japan is more like 3
grades than 2 grades.
I also participated in a Japan-Netherlands friendship match by the Japanese
Embassy in Holland. We played against
And 6-7 every now and then (humans imitating MC bots?).
Do you play go competitively, Tony?
Dave
Van: [EMAIL PROTECTED] namens Heikki Levanto
Verzonden: wo 10-12-2008 21:50
Aan: computer-go
Onderwerp: Re: [computer-go] MC Opening stage
On Wed, Dec 10, 2008
It might be a good idea then to look at some games of competitive players to
get an idea of how a game develops from the opening to the middle game and the
endgame.
There are some online game collections, but you could also register an account
on an online go server to watch some games while
I think that would not be enough, because that would only fix one point.
EGF ratings are not pure Elo ratings. EGF ratings are weighted to fit 100
points for one handicap stone, which happens to match about 65% winning
percentage in even games for medium level players (around 3k).
Also, I am
That is wonderful!
When I look at this table it seems to support my claim that USCF master (2200
USCF Elo) compares to about 4d (2400 EGF) and that 2d compares to a strong USCF
class A player.
So this table too, suggests that computer-go has not yet been reached amateur
master level in the
Let me rephrase a bit: USCF master level is about 3% away from the human top,
but 7-8 handicap stones (an estimate of mogo's distance from the top) is about
13% away from the human top.
A program would have to win about 50% of its games with a 5 stone handicap
against a strong pro before
I doubt that much can be learned from comparing the overall length of rating
scales.
1: The large draw margin in chess compresses the high end of the chess Elo
range compared to go.
It takes a fairly large difference in skill for one very strong chess player to
win 65% against another very
Erratum: Such a scale would probably stretch the range at the top and
compress it at the bottom ... Such a scale would probably compress the
range at the top and stretch it at the bottom ...
It requires many more Elo points to compensate for a chess handicap at the high
end than at the low
It depends very much on what exactly you mean by amateur master level. Is it
a level that compares to amateur master level in chess?
And what is amateur master level in chess? USCF master, FIDE master or
international master?
Some time ago I participated in a discussion about comparing chess
Hello Tony,
I'm just speaking for myself here, but I suspect you may not get the answers
you are looking for, because you're not asking the right questions.
I get the impression that the situation is similar to this scenario:
A young guy with a surf board and a hawai shirt
You are absolutely right. Tony did not imply he wanted to go for the summit
today. Also, Tony's qualifications are surely a lot better than a hawaiian
shirt and a surfboard.
My analogy was an exaggeration with the intention of explaining to Tony that he
has more catching up to do than he
Just a quickstart here, and I'm not one of the real experts, so it may contain
inaccuracies. Perhaps you already know all this, perhaps not.
The basis of many game-playing programs (like chess programs) is the minimax
(or negamax) algorithm.
It is a simple algorithm by itself and you can find
Van: tony tang [mailto:[EMAIL PROTECTED]
Verzonden: za 15-11-2008 21:22
Aan: [EMAIL PROTECTED]
Onderwerp: [RE:computer-go] Monte carlo play?
Hello Dave,
Thank you for a thorough introduction of the theory, and i sincerely hope I am
not wasting your time with amateur questions.
Because it was
Van: [EMAIL PROTECTED]
Verzonden: za 15-11-2008 22:44
Aan: tony tang
Onderwerp: RE: computer-go] Monte carlo play?
Hello Tony,
Ok, so you know about the minimax algorithm and the like. My impression was
wrong and I'm very sorry for my analogy.
I'm no expert on montecarlo, so I can only say
Van: tony tang [mailto:[EMAIL PROTECTED]
Verzonden: za 15-11-2008 23:08
Aan: [EMAIL PROTECTED]
Onderwerp: RE: computer-go] Monte carlo play?
Thanks Dave, that was incredible helpful, hopefully this new hobby of mine will
materialise into a decent project.
All the best
Tony
What if the playout uses the AGA rule of paying 1 point for a pass and
requiring white to pass last (so the game does not end by two passes if black
plays the second pass).
Wouldn't the score then be equivalent to the japanese score?
Dave
Van: [EMAIL
And of course black should pay 1 point for each extra handicap stone.
http://www.britgo.org/rules/compare.html#coun
Dave
Van: [EMAIL PROTECTED] namens [EMAIL PROTECTED]
Verzonden: do 6-11-2008 19:28
Aan: computer-go
Onderwerp: RE: [computer-go] Monte-Carlo and
ERRATUM:
Sorry, I made a small mistake in my example.
The komi should be 3.5 so white wins by 0.5 if 2 passes end the game.
Dave
Van: [EMAIL PROTECTED]
Verzonden: vr 24-10-2008 10:00
Aan: computer-go
Onderwerp: Ending games by two passes
Is it correct to
Please correct me if I'm wrong, but are you saying that white is alive with
TT-rules (=Tromp-Taylor?) or other rulesets with positional superko if black
has not enough eyes left to fill as ko threats?
If that's true, I would be disgusted if positional superko would ever be
accepted as a rule in
I know white is dead, but what matters is that the controller does not know.
The only way for the controller to know that white is dead is by requiring
black to capture white before ending the game.
And when 2 passes end the game, black is unable to do that. So the controller
will have to assume
I'm glad we agree on this :)
Your previous respons suggests that this issue has been debated before on this
list, so I'll probably be able to find references about this issue.
I wouldn't want to restart a debate here about positional superko :)
Thanks,
Dave
In my opinion the goal of a ko rule is to prevent games from not ending.
1: If one player can force a game to an end even when the other player aims at
not ending the game, then the rule is good enough.
In my previous example I would consider it an undesired side effect of a ko
rule that white
After reading up a bit on this issue, I didn't find a clear positive consensus
in this list about a preferred ruleset for computer-go, human-computer-go, real
life go and go servers.
(I did find a negative consensus about the current Japanese rules, though)
I'm curious if there exists a
I (EGF 4d) am probably not strong enough to give well founded comments on 9x9
games, but already move 2 at D3 seems strange from a shape point of view
(whatever that may be worth on 9x9)
The continuation B C3 B4 D5 seems the most natural continuation once D3 is
played, but on 19x19 this is kind
Sorry, I just realized this is about 6x6 go. Please ignore my previous response.
Dave
Van: [EMAIL PROTECTED]
Verzonden: ma 29-9-2008 20:09
Aan: [EMAIL PROTECTED]; computer-go; computer-go
Onderwerp: RE: [computer-go] Analysis of 6x6 Go
I (EGF 4d) am probably
When I teach beginners, I use area scoring on 9x9 until they are advanced
enough to understand territory scoring without disputes (which usually does not
take very long).
Dave
Van: [EMAIL PROTECTED] namens Peter Drake
Verzonden: do 18-9-2008 6:14
Aan:
It seems to be more efficient for humans to count territory instead of area
during the game.
I've heard that even chinese professionals save time by estimating the score
during the game by counting territory japanese style and correcting for stones
captured (you have to remember captures, which
I'm not surprised that the data for games with 90% winning chances is lacking.
The McMahon pairing system is widely used in western go tournaments to prevent
mismatched games,
because most players don't like mismatched games (either as the stronger or the
weaker player).
Rating systems are
There seems to be a discrepancy: 11.5% between 2d and 5d in EGF rating system
versus 2.0% between 2d and 5d in KGS rating system.
I think this can be explained by hidden biases in the EGF statistics:
1: To be a 5d in real tournaments in Europe does not mean your rating is
between 2450 and 2550
The other way around happens too: in 2006 I had a 4 month pause on KGS and my
rank dropped from 4d to 4k.
Van: [EMAIL PROTECTED] namens Jason House
Verzonden: do 31-1-2008 20:33
Aan: computer-go
Onderwerp: Re: [computer-go] Go rating math information
On Jan
It might be possible to automatically compensate for lag by looking up the
geographic location of a bot's ISP. For instance via
http://www.hostip.info/use.html
https://webmail.planet.nl/exchweb/bin/redir.asp?URL=http://www.hostip.info/use.html
.
Dave
Van:
It might be possible to estimate lag by looking up the geographic location of a
bot's ISP. For instance via http://www.hostip.info/use.html .
Dave
Van: [EMAIL PROTECTED] namens Peter Christopher
Verzonden: vr 4-1-2008 5:27
Aan: computer-go@computer-go.org
51 matches
Mail list logo