Life In 19x19
http://www.lifein19x19.com/

Using computer go engine for game review
http://www.lifein19x19.com/viewtopic.php?f=18&t=12527		 Page 1 of 1
Author:  Cartier [ Thu Dec 17, 2015 10:12 pm ]
Post subject:  Using computer go engine for game review
Hi all,

Is there any way I can use a gtp go engine with a suitable interface to run game analysis through my own games for blunder check or comparison with what the computer would have played if given my turn in the game?

That way I can more selectively put up games here for review and comments.

Regards,
Cartier
Author:  Mike Novack [ Fri Dec 18, 2015 7:02 am ]
Post subject:  Re: Using computer go engine for game review
Cartier wrote:
Hi all,

Is there any way I can use a gtp go engine with a suitable interface to run game analysis through my own games for blunder check or comparison with what the computer would have played if given my turn in the game?
That way I can more selectively put up games here for review and comments.
Regards,
Cartier


The answers depend on level.
a) machine review --- at least one of the programs can be asked to "show reasons for move", but that's only at the levels below that using MCTS. But MFOG 12 isn't free and this will be of no ues to you once you get to perhaps 6 kyu.

b) Force machine to show you its next move from a given position --- you probably could force most of the programs to do that but might need to be running to instances (reverse colors).
Author:  yoyoma [ Fri Dec 18, 2015 8:42 am ]
Post subject:  Re: Using computer go engine for game review
I don't think the GTP protocol itself has that sort of functionality. I know CrazyStone can show you how much worse it thinks your move is versus what it thinks is best. Probably some of the other commercial programs do as well.
Author:  xed_over [ Fri Dec 18, 2015 11:32 am ]
Post subject:  Re: Using computer go engine for game review
I used to use GnuGo to analyze my games (until I got stronger than GnuGo)
Author:  Cartier [ Fri Dec 18, 2015 9:26 pm ]
Post subject:  Re: Using computer go engine for game review
So it seems that crazystone is the way to go...
Author:  Mike Novack [ Sat Dec 19, 2015 7:32 am ]
Post subject:  Re: Using computer go engine for game review
None of the MCTS programs can do a good job of analyzing moves in the sense that you need. You need MORE than "move A is better than move B because if you make move A you will win a higher percentage of games than if you make move B". Knowing how much better (the percentages) not much use to you. You need to know WHY better in go terms, not the abstract "more likely to win the game".

At 10k (your current reported strength) there are programs that could do what you ask, so I answered "ye"s. If you had said you were a 3k I would have answered "no".
Author:  oren [ Sat Dec 19, 2015 9:39 am ]
Post subject:  Re: Using computer go engine for game review
Cartier wrote:
So it seems that crazystone is the way to go...


It is a good program for doing what you want to do.
Author:  gowan [ Sat Dec 19, 2015 10:06 am ]
Post subject:  Re: Using computer go engine for game review
As said before, Crazystone doesn't really know how to play go. Because it chooses its move based on its estimated probability of a winning game resulting from subsequent moves, it can't explain in go terms why its move is better than some other one. However this might be more effective for learning than being told why moves are good. We learn best by figuring things out for ourselves. If you are told that a certain move is good without being given a useful reason, it becomes a pointer for you to find out the reason for yourself. In school, being told the answer to a problem often is a big boost in finding the complete solution.
Author:  Cassandra [ Sat Dec 19, 2015 10:25 am ]
Post subject:  Re: Using computer go engine for game review
gowan wrote:
As said before, Crazystone doesn't really know how to play go. Because it chooses its move based on its estimated probability of a winning game resulting from subsequent moves, it can't explain in go terms why its move is better than some other one. However this might be more effective for learning than being told why moves are good. We learn best by figuring things out for ourselves. If you are told that a certain move is good without being given a useful reason, it becomes a pointer for you to find out the reason for yourself. In school, being told the answer to a problem often is a big boost in finding the complete solution.

It will help you a lot to check the first few moves in the window wherein the "suggested" / "considered" moves are listed.

At least, you will get some ideas which alternate moves to think of.

This is similar to what I did in the "old" days of my research in "Igo Hatsuyôron 120", using Many Faces of Go at that time (CrazyStone was not available then). These programs do not get "exhausted", and therefore do not "overlook" a "vaild" move in one of the numerous sub-variations that was usually considered in all the previous ones. Some of the program's suggestions will make you think "Hmm, why not ?"
Author:  Gohst [ Sat Dec 19, 2015 11:45 am ]
Post subject:  Re: Using computer go engine for game review
I will say that I hesitate to being this up, but there is a program that does exactly what you are asking (UPDATE: but it doesn't have a GTP interface, AFAIK... you have to manually enter the moves on the board). It's called MoyoGo (http://www.moyogo.com/), and it costs $27 US. I have it, and it is kind of amazing. However, by bringing it up, I am going to precipitate a shitstorm from a lot of the other posters here, based on past years of threads on this same program. Here's a link to one such stormy thread that was, thankfully, locked by the mods (viewtopic.php?f=18&t=5838&hilit=moyogo). Maybe this thread can avoid the same fate, please?

I'm hoping that most of the people who comment on MoyoGo will exercise some discretion. It isn't my intent to provoke yet another shouting match. Consider the points below (and the ones that others will inevitably raise), and make up your own mind. As a studying tool, I believe the program has few peers, but you will hear strong opinions about it, in the replies this post gets. I don't have strong opinions, and believe the amount of money involved is small enough to be inconsequential.

MoyoGo has a bit of a history, going back years to GoDiscussions.com. Here's a brief, dispassionate summary:

1. When it was released, there was a fair bit of controversy over its use of game data bases "taken" from other products. Lots of back-and-forth over whether a historical game record (sans commentary) could be copyrighted (I don't believe it can), and whether it was ethical to profit from other's work in compiling the data base of games in the first place (seems borderline unethical to me, but I still bought the program). I can't remember whose data base was appropriated, but wait 5 minutes and someone will fill in the blanks.

2. The author of the program is a very bright, but unstable, man who is prone to alarming outbursts. He has no filters on what he says, and therefore he seems to wind up being banned from any public forum he participates in, including GoDiscussions and (I believe) this board as well. No doubt his personality puts off many people with whom he interacts, and hampers sales of his program. In my minimal dealings with him, he was honest and helpful. YMMV.

Anyway, check it out and make up your own mind.

If I have offended any of those folks with strong feelings on this matter, I apologize. Despite its "pariah" status among some in the community, I think it is a valuable program.
Author:  Cartier [ Sun Dec 20, 2015 8:42 am ]
Post subject:  Re: Using computer go engine for game review
It's a pity I can't find it online any more...
Author:  Gohst [ Sun Dec 20, 2015 8:14 pm ]
Post subject:  Re: Using computer go engine for game review
Not sure if you are replying to me. I put a link to the MoyoGo page (www.moyogo.com) in the previous post. The page is still there, and looks to be enabled to take your money, in an E-commerce sort of way. So I think it is still available.
Author:  Javaness2 [ Mon Dec 21, 2015 1:14 am ]
Post subject:  Re: Using computer go engine for game review
There are a few database programs that will offer you a % played here figure for various moves. Don't think they are that useful beyond he opening stages of the game.
Crazystone's %win figure might be a bit more useful. Isn't that the kind of thing they have in the chess computers?
Overall, I think it is a gap in the market. There is no really slick tuition software out there, although JosekiFarm was a pretty fun idea. It would have been nice to see that make it out of Alpha.
Author:  Cartier [ Mon Dec 21, 2015 7:59 am ]
Post subject:  Re: Using computer go engine for game review
Yes indeed computer chess engines can suggest stronger moves at every turn, and even without the ensuing moves it broadens the mind's eye and helps as a blunder check. I'm looking for something that can help me weed out DDK moves from my games, to a certain extent.
Author:  Mike Novack [ Mon Dec 21, 2015 3:52 pm ]
Post subject:  Re: Using computer go engine for game review
Cartier wrote:
Yes indeed computer chess engines can suggest stronger moves at every turn, and even without the ensuing moves it broadens the mind's eye and helps as a blunder check. I'm looking for something that can help me weed out DDK moves from my games, to a certain extent.


Try going back and rereading all the prior posts.

MANY of the available engines can "suggest stronger moves". That isn't the problem. Look, at your current level SOMETIMES just seeing what the stronger move is might be enough. But most of the time, no. You would need to see the REASON behind the stronger move.

Look, suppose you are following the game of strong players, and all of a sudden, in the middle of a fight (say in a joseki) one of them makes a move across the board. Are you going to be able to see:
1) That's a ladder breaker (for the current fight, a fwe moves ahead)
2) If not responded to. more will be gained there than lost by accepting less than equal in the local fight.

You NEED the comment "ladder breaker" at least. Not just "this move is better".

Those chess programs. Are they telling you WHY the suggested better move is better? Moving a rook to an open file, that's obvious. But moving a rook to a file to restrain its being opened, that's more subtle.
Author:  xed_over [ Mon Dec 21, 2015 5:39 pm ]
Post subject:  Re: Using computer go engine for game review
Mike Novack wrote:
MANY of the available engines can "suggest stronger moves". That isn't the problem. Look, at your current level SOMETIMES just seeing what the stronger move is might be enough. But most of the time, no. You would need to see the REASON behind the stronger move.

I have to disagree.

I quit using a computer in this way when I became about the same level or stronger as the computer.

But until then, it was quite sufficient for self study.


One of the main ways people usually suggest to get stronger, is self-review.
Computer-aided self-review is not much different -- except you have a computer recommending alternative (perhaps better) moves for some reason. You still have to either work out your own reasons, or ask a stronger person for their opinion. The same as I would for my own non-computer-aided self-study.

I say, use what you have available to you, until you can get something better.
Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/