Ok, first off - I'd like to announce that from now on, I'm going to be writing this blog in English. A very good friend pointed out that I can reach a much wider audience this way, and really, there is no real reason to limit this blog to a Serbian audience just because a large majority of my friends are Serbian. Indeed, probably any of my friends that would even find it worthwhile to read this blog already use English well enough to not be hindered by my (lack of experience in) use of English.
Anyway, let's look at a game.
Yesterday I played a game against SmartGo on my iPod touch. Right now, I'd like to take a moment and talk about the AI that I've encountered so far, and my experience with it. Probably the most popular go AI out there is GnuGo. You can find GnuGo in many PC go clients (such as GoPanda, for instance). This is because GnuGo is only an "engine" for playing go. The brain, basically. Obviously, there are many different GUI applications built around GnuGo, but the logic is always the same. GnuGo is a pretty powerful AI, as well. I have won against it only a few times so far, and that was with undoing some stupid mistakes I still tend to make in the heat of battle. If you can, I suggest playing against a weaker AI when you are only starting out. Bear in mind that GnuGo is an ideal partner for working on your battle technique. Here I mean small, local battles, and not the "war" that goes on throughout the entire board.
The only other AI I encountered by now is SmartGo. This one is weaker than GnuGo, but can still be a challenge for a novice such as myself at times. SmartGo is a very neat set of applications for the iPhone (and iPod touch). It includes "GoBooks" (an application in which you can buy and read popular books on go, adapted for the iDevice), "SmartGo" and "SmartGo Pro". The "Pro" application has some neat things such as a very large database of professional games for you to look at and analyze, as well as a bunch of games annotated by professional players. The regular version of the app is more than enough for simply playing the game and reading .sgf files. The AI built into SmartGo has a skill level, as well as a time limit. Before you start the game, you get to pick how much time (at most) the AI can spend on each individual move. The game I played yesteday was against a 4s AI. Here it is.
2012-02-20-Median-SmartGo4s.sgf
Notice how strange his opening is. He's spread too far apart, so he has to fill up gaps while I play meaningful moves in the lower right. Now, one thing you get to learn when playing against SmartGo is working on your goals. SmartGo seems to have a very short attention span, and will tend to jump around the board like a monkey on speed. I'm assuming it always scans the whole board, or a very large part of it, and looks for the optimal move in the current situation, without going into much depth. This means that SmartGo is going to try your patience and focus A LOT. You must not give in, however. As long as you have a clear idea of what you're trying to accomplish, SmartGo's craziness shouldn't be a problem.
One good thing you can learn with SmartGo is when you can tenuki. For example, look at moves 31 and 32. There is no immediate need for me to play any more in the upper right. He can't destroy me in one move.
The play in the lower right had me intrigued during the game, and now as well. I'm reffering to moves 58-62. Did I do as well as I could there? I'm pretty sure I could have done better, but I can't see it. At least I should have been able to have a similar situation and sente.
I deeply enjoyed the attack on the upper edge (starting at 64), and that is all the more reason I handed myself an impressive facepalm at 101. I was sure for a solid number of moves that I was safe there (and I wasn't - that attack of his could have happened even earlier), and never thought that he could take away two eyes from me on the upper side. In stead of using undo (a reflex I have when playing against the AI), I decided to play this out. I couldn't count on him goofing up and letting me live (like when he let me attack in the first place), because when the AI goes for the kill, it starts reading deeper into the local situation, and it gets pretty tricky to gain life.
Luckily, his position downwards was weak enough that I had just enough time to surround one of his groups. Move 151 confused me. I'm pretty sure he should have connected there, instead of sacrificing 6 stones. Putting aside the 12 points I gained, connecting would mean that the pressure is still on me, instead of him. After the capture, I can easily start an attack on his group on the upper edge. It would probably be an unsuccessful attack in the sense that I wouldn't kill anything, but I would still take away a lot of (otherwise) his territory in the process. And all this for a move such as N4, which I neutralized quickly afterwards. The AI resigned at 196.
If you were SmartGo, would you perform the attack on the large white group in another way? Have you noticed any big blunders by myself or the AI?
Anyway, let's look at a game.
Yesterday I played a game against SmartGo on my iPod touch. Right now, I'd like to take a moment and talk about the AI that I've encountered so far, and my experience with it. Probably the most popular go AI out there is GnuGo. You can find GnuGo in many PC go clients (such as GoPanda, for instance). This is because GnuGo is only an "engine" for playing go. The brain, basically. Obviously, there are many different GUI applications built around GnuGo, but the logic is always the same. GnuGo is a pretty powerful AI, as well. I have won against it only a few times so far, and that was with undoing some stupid mistakes I still tend to make in the heat of battle. If you can, I suggest playing against a weaker AI when you are only starting out. Bear in mind that GnuGo is an ideal partner for working on your battle technique. Here I mean small, local battles, and not the "war" that goes on throughout the entire board.
The only other AI I encountered by now is SmartGo. This one is weaker than GnuGo, but can still be a challenge for a novice such as myself at times. SmartGo is a very neat set of applications for the iPhone (and iPod touch). It includes "GoBooks" (an application in which you can buy and read popular books on go, adapted for the iDevice), "SmartGo" and "SmartGo Pro". The "Pro" application has some neat things such as a very large database of professional games for you to look at and analyze, as well as a bunch of games annotated by professional players. The regular version of the app is more than enough for simply playing the game and reading .sgf files. The AI built into SmartGo has a skill level, as well as a time limit. Before you start the game, you get to pick how much time (at most) the AI can spend on each individual move. The game I played yesteday was against a 4s AI. Here it is.
2012-02-20-Median-SmartGo4s.sgf
Notice how strange his opening is. He's spread too far apart, so he has to fill up gaps while I play meaningful moves in the lower right. Now, one thing you get to learn when playing against SmartGo is working on your goals. SmartGo seems to have a very short attention span, and will tend to jump around the board like a monkey on speed. I'm assuming it always scans the whole board, or a very large part of it, and looks for the optimal move in the current situation, without going into much depth. This means that SmartGo is going to try your patience and focus A LOT. You must not give in, however. As long as you have a clear idea of what you're trying to accomplish, SmartGo's craziness shouldn't be a problem.
One good thing you can learn with SmartGo is when you can tenuki. For example, look at moves 31 and 32. There is no immediate need for me to play any more in the upper right. He can't destroy me in one move.
The play in the lower right had me intrigued during the game, and now as well. I'm reffering to moves 58-62. Did I do as well as I could there? I'm pretty sure I could have done better, but I can't see it. At least I should have been able to have a similar situation and sente.
I deeply enjoyed the attack on the upper edge (starting at 64), and that is all the more reason I handed myself an impressive facepalm at 101. I was sure for a solid number of moves that I was safe there (and I wasn't - that attack of his could have happened even earlier), and never thought that he could take away two eyes from me on the upper side. In stead of using undo (a reflex I have when playing against the AI), I decided to play this out. I couldn't count on him goofing up and letting me live (like when he let me attack in the first place), because when the AI goes for the kill, it starts reading deeper into the local situation, and it gets pretty tricky to gain life.
Luckily, his position downwards was weak enough that I had just enough time to surround one of his groups. Move 151 confused me. I'm pretty sure he should have connected there, instead of sacrificing 6 stones. Putting aside the 12 points I gained, connecting would mean that the pressure is still on me, instead of him. After the capture, I can easily start an attack on his group on the upper edge. It would probably be an unsuccessful attack in the sense that I wouldn't kill anything, but I would still take away a lot of (otherwise) his territory in the process. And all this for a move such as N4, which I neutralized quickly afterwards. The AI resigned at 196.
If you were SmartGo, would you perform the attack on the large white group in another way? Have you noticed any big blunders by myself or the AI?
No comments:
Post a Comment