Computer Go

Discuss anything you like about chess related matters in this forum.
John McKenna

Re: Computer Go

Post by John McKenna » Wed Mar 09, 2016 1:08 am

Paolo Casaschi wrote:
Christopher Kreuzer wrote:This bit may also be relevant: "Throughout the whole game, no time is deducted if the player plays in less than one minute. The Japanese expression for this is literally "no time" (in katakana). So when a player is said to have played in no time, it actually means within fifty-nine seconds rather than instantly."
In other words, I only start using the "initial" 2h time allotment after I exceed three times the 60s per move limit. Correct?

Couple of questions for personal curiosity:

- how many moves is a typical go game?
- is zeitnot and losing on time more or less of a problem than in chess?
See the table below for some approximate stats on International chess, Chinese chess, Japanese chess and Go -

Code: Select all

Board. Size  /	 Pieces  / Diff. Pieces  / Legal Posns. / Poss. Games /Av. moves 
Chess    64   /   32	    /   6        /   10 to 47th  /   10 to 123rd  / 80 
Xiangqi 90   /   32	    /   7        /    10 to 48th /   10 to 150th  / 95 
Shogi    81   /   40	    /  8       /    10 to 71st /    10 to 226th  / 110 
Go      361   / Up to 360 /  1       /    10 to 171st /   10 to 360th /   150
NB: Apologies for the wonky table and non-std. exponents.

Time is allocated as follows -
In normal games of Go the principal thinking time allowed is 1.5 to 2 hours, usually combined with a byo-yomi (秒-読み = second-count) of between 20 to 30 seconds.

After the principal time has been used up the clock jumps to byo-yomi (second-count) time. Each time a player completes a move the clock jumps back. If the player does not complete the move before the clock reaches 0 seconds a flag is displayed.

For top matches the principal thinking time is 9 hours followed by 5 byo-yomi periods of 60 seconds each.

At the end of the 9 hours principal time the clock jumps to the 1st byo-yomi period of 5 minutes. Then if at any time a player completes a move BEFORE 4 mins. are exceeded the clock reverts to the 1st byo-yomi period of 5 mins. If a move is completed AFTER 4 minutes is exceeded the clock jumps to the 2nd byo-yomi period of 4 mins.

Subsequently the clock will revert to the 2nd, 3rd or 4th, or jump to the 3rd, 4th and 5th byo-yomi periods of 3, 2 and 1 minute(s) respectively in the same manner.

In the 5th and final byo-yomi period if the clock reaches 0 seconds a flag is displayed to indicate the player's time is all used up. (DGT 2000)


So the average number of moves in a game of Go is about 150.

In the match that begins in Seoul later today the time limit is for shorter "normal" games not much longer "top" ones.
Therefore there will be much more potential for time trouble to occur. Also, note that the rules of Go were slightly different in China and Japan in the 1970s. Perhaps the rules have been standardised since then in the three main centres - the other being S. Korea. No idea if Go is popular in N. Korea and whether their players would be allowed to compete internationally.

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Wed Mar 09, 2016 8:33 am


MartinCarpenter
Posts: 3048
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Wed Mar 09, 2016 8:40 am

Oh well :) Bet they've really scaled the hardware its running on up since the European champ, and maybe it benefited more from that than the computer engines did.

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Wed Mar 09, 2016 8:52 am

In a report on the Radio 4 Today programme this morning, they said that experts described one of the computer's moves as "either brilliant or stupid".

I suppose the conclusion has to be that it was brilliant! :lol:

User avatar
Paolo Casaschi
Posts: 1187
Joined: Thu Jan 08, 2009 6:46 am

Re: Computer Go

Post by Paolo Casaschi » Wed Mar 09, 2016 9:47 am

Clive Blackburn wrote:In a report on the Radio 4 Today programme this morning, they said that experts described one of the computer's moves as "either brilliant or stupid".
In a related report, after hearing the comment on Radio 4, the alphago computer allegedly said the exact same about the supposed experts ;-)

stevencarr

Re: Computer Go

Post by stevencarr » Wed Mar 09, 2016 2:18 pm

This is just astonishing. I never thought it would happen.

MartinCarpenter
Posts: 3048
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Wed Mar 09, 2016 2:56 pm

I guess not saying 'never' about such things is the one thing we've learnt in recent decades :)

Go was always going to be a major target with the publicity pay off in China/Japan (and elsewhere of course), so I think the real question was probably always going to be 'when'. This is definitely (considerably) ahead of the schedule most people were anticipating, partially because its so much further ahead of other Go software.

A vastly bigger gap than say Deep Blue was vs chess engines of its time.

User avatar
Christopher Kreuzer
Posts: 8822
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Computer Go

Post by Christopher Kreuzer » Wed Mar 09, 2016 9:10 pm

For anyone wanting to play through the game (no, really!), I was pointed to this link:

https://online-go.com/review/114161

There is also a Wikipedia article on the match:

https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol

Apparently, Lee Sedol tried an offbeat approach in the opening that didn't work out and AlphaGo played well after that in a complex position. The final result was very close.

I am guessing that Lee Sedol will take a more orthodox approach in the next game to win and then stay with that for the rest of match, rather than taking risks as he did in this game. Will be interesting to see what AlphaGo manages to do.

User avatar
Carl Hibbard
Posts: 6028
Joined: Fri Dec 08, 2006 8:05 pm
Location: Evesham

Re: Computer Go

Post by Carl Hibbard » Wed Mar 09, 2016 9:27 pm

It's good publicity but I expect Sedol to adapt and win 4-1 now.
Cheers
Carl Hibbard

Brian Towers
Posts: 1266
Joined: Tue Nov 18, 2014 7:23 pm

Re: Computer Go

Post by Brian Towers » Wed Mar 09, 2016 9:53 pm

Carl Hibbard wrote:It's good publicity but I expect Sedol to adapt and win 4-1 now.
I'm with you on this one, Carl. I think Lee Sedol is killing two birds with one stone (sorry about that). His loss grabs the headlines and gives the computer some dodgy data to lead it astray for the next game.
Ah, but I was so much older then. I'm younger than that now.

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Thu Mar 10, 2016 8:22 am

Christopher Kreuzer wrote: I am guessing that Lee Sedol will take a more orthodox approach in the next game to win and then stay with that for the rest of match, rather than taking risks as he did in this game. Will be interesting to see what AlphaGo manages to do.
This is reminiscent of the Deep Blue match, when Kasparov didn't take the computer seriously at first and took risks, which is precisely the wrong strategy! :?

User avatar
Christopher Kreuzer
Posts: 8822
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Computer Go

Post by Christopher Kreuzer » Thu Mar 10, 2016 9:01 am

Brian Towers wrote:
Carl Hibbard wrote:It's good publicity but I expect Sedol to adapt and win 4-1 now.
I'm with you on this one, Carl. I think Lee Sedol is killing two birds with one stone (sorry about that). His loss grabs the headlines and gives the computer some dodgy data to lead it astray for the next game.
We all turned out to be wrong. Lee Sedol has lost the second game! :shock:

Google AI wins second Go game against world champion (BBC)

stevencarr

Re: Computer Go

Post by stevencarr » Thu Mar 10, 2016 9:05 am

Carl Hibbard wrote:It's good publicity but I expect Sedol to adapt and win 4-1 now.
He lost the second game.

MartinCarpenter
Posts: 3048
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Thu Mar 10, 2016 9:50 am

Can't feed the computer dodgy data - it'll have pretty much every game he's ever played in there.....
(And have analysed them, although it didn't read like they're doing specific anti opponent measures.).

User avatar
Christopher Kreuzer
Posts: 8822
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Computer Go

Post by Christopher Kreuzer » Thu Mar 10, 2016 11:14 am

Apparently, AlphaGo played "an endgame brilliancy" (according to a friend of mine who plays Go). Lee was also under time pressure.

What I'm wondering, is whether, if you postulate some 'perfect' way to play go and a 'perfect' way to play chess, whether humans, before the advent of computer programs that managed to beat them, were closer to playing perfectly in go or chess? I am wondering whether, because go is much more complex, that the (incredibly high) standard of top-level human go is actually further from 'perfection' than top-level human chess?

Once computers started to do better in chess (especially in certain areas), then humans began to be able to play better chess (to a certain extent) because they could understand the computer moves. I wonder whether this will happen in go, and whether it is even possible? To put this another way, can top-level go/chess players explain their moves [or computer moves] to lower-level go/chess players in such a way that they can improve? Seeing how accurate top-level commentary is with or without an engine switched on is another angle.