Is the computer teaching me bad habits?

Inkwolf · Post by **Inkwolf** » Mon Aug 15, 2011 8:24 pm

A lot of my practice is done against the computer. (Goban for Mac, using Gnu Go 3.7.11.) I've come to suspect that some of my moves are odd, because they are played in anticipation of the computer's playing style, rather than necessarily being good plays in general.

For example, I've started using the large knight's move for corner enclosures. Even though I know it's weaker (and have had it demonstrated to me in games vs stronger humans) the computer seems to jump all over my corner when I play the normal knight's move. The large nabs more territory with less conflict, so I play accordingly.

Move 10 and 12 as well--playing those felt wrong somehow, but if the 3-3 point has all 4 liberties, the computer will nearly always glom on to it within the first 20 or 30 moves. Were those actually decent moves? Would kicking with the diagonal be a bad choice? Or a one-point extension?

Do you see anything else that seems strange, and/or which the computer (white) reacts oddly to? Just wondering how many bad habits I might be picking up (apart from freely using the Undo button.)

Also, since you're looking, how should I have played in the upper left when it was invaded? What I did seemed like kind of a mess. Should I have played to the edge before trying to stop white extending?

Any other comments are welcome, though I am already aware of some of the boneheaded moves, such as leaving Q6 alone too long, 20-22 being mall and slow, the mindless chaos in the 80's and 90's, and failing to notice I could have killed the whole lower left corner.

Thanks!

daniel_the_smith · Post by **daniel_the_smith** » Mon Aug 15, 2011 9:25 pm

Re: the upper left: don't tenuki with 57 and expect to still make points in the corner!

Overall, yes, the computer has taught you some bad stuff. I'd recommend you at least find a monte carlo engine to play against. That will also probably teach you bad stuff but at least it will be the opposite sort of bad stuff. I'd call black in this game unenterprising and too territorial. Both players are over defensive. All the stones seem overconcentrated. Neither player ever really goes on the offensive. White gives up like 20 points starting at move 154 for no apparant reason.

There's nothing wrong with the large knight's enclosure, btw. It's just an older style.

Edit: You couldn't (shouldn't, anyway) have killed the lower left (as long as he has two external liberties it's not even ko), and what was 204 for?

moonrabbit · Post by **moonrabbit** » Mon Aug 15, 2011 10:58 pm

Just a few comments.

10: This is slow but playable, as is the kick at R15. I probably would have pincered at R12 (to force White into the corner and take influence on the outside.)

12: I would already be thinking about helping Q6 by playing at R9, attacking White's two stones while making a base. Taking territory in the upper left is slow. If you're going to play in the corner, I slightly prefer B16 (later, if you invade at C12, you have a stronger threat to connect underneath C14.)

26: K16 is slow. Notice that this move is single-purpose; all it does is claim territory. It's not even that much territory, for that matter...White will have a hard time invading even if Black doesn't play K16. Instead, consider moves like E14 (expand Black's area of influence while limiting White's framework) or a severe invasion like C12.

174 & 176: How many points were these two moves worth?

Inkwolf wrote:For example, I've started using the large knight's move for corner enclosures. Even though I know it's weaker (and have had it demonstrated to me in games vs stronger humans) the computer seems to jump all over my corner when I play the normal knight's move. The large nabs more territory with less conflict, so I play accordingly.

Although both moves are fine, neither the small nor the large knight's extension makes any territory without another move in the corner. White can still invade at 3-3, but you get thickness as compensation. Especially early in the game, this exchange can be good for you; it's not something you need to try to avoid.

Even more importantly, the whole idea of making territory while avoiding conflict isn't really healthy. It's actually very hard to make territory efficiently without fighting...

Bill Spight · Post by **Bill Spight** » Mon Aug 15, 2011 11:12 pm

Well, you have a major bad habit that I doubt that you have learned from the computer. You seem to think that go is a game of making territory and then adding to it incrementally. I had the same bad habit when I started out. That way of playing is too slow, so that you quickly start to fall behind.

Dusk Eagle · Post by **Dusk Eagle** » Tue Aug 16, 2011 12:34 am

Daniel, moonrabbit, and Bill all described the problem pretty well with this game. Your sole goal seems to be to expand your existing territory with moves like :b14:

,

, and

. Bill's alternative suggestion for :b26:

is really good and you should try to understand why. This is a natural followup from your C16 stone (though you can play it even without C16) and it invades white's territory and takes makes your own. Don't be afraid to start fights!

Anyway, I would say that using the large-knight's move simply because the computer reacts differently to isn't a very good reason. It's not actually urgent for you to turn the corner into territory. This is one of the things beginners have a tough time understanding about the 4-4 - it doesn't aim directly for the corner territory, so you shouldn't mind if white jumps in and takes it - if you play correctly, your influence will balance it out. I actually think beginners should use the 3-4 a lot more than they do, because its primary goal of corner territory is far easier to understand than the 4-4. Anyway, perhaps you should show us what you mean when you say the computer "jumps all over" the more normal knight's move, and we could tell you whether you're handling it correctly or not.

Mike Novack · Post by **Mike Novack** » Tue Aug 16, 2011 4:52 am

daniel_the_smith wrote:......
Overall, yes, the computer has taught you some bad stuff. I'd recommend you at least find a monte carlo engine to play against. That will also probably teach you bad stuff but at least it will be the opposite sort of bad stuff. ......

Unfortunately he can't find a "monte carlo" engine to play against at low handicap as none are suffiently weak (even at their weakest MC setting). Fortunately if he still needs 3 stones against gnugo he won't learn too many bad habits from the computer. He'll just learn good stuff in a different order than he would if playing against humans and will not learn some of the bad habits one can learn when playing against humans of one's own strength.

He could try one of the MC programs at high handicap.

Inkwolf · Post by **Inkwolf** » Tue Aug 16, 2011 5:11 am

Okay, thanks for the tips. Apparently what I need to learn next is how to attack.

I have to say, though, that every time I get advice on Go, it seems to conflict with half of the previous advice.

perceval · Post by **perceval** » Tue Aug 16, 2011 6:00 am

Inkwolf wrote:Okay, thanks for the tips. Apparently what I need to learn next is how to attack.

I have to say, though, that every time I get advice on Go, it seems to conflict with half of the previous advice.

if it makes you feel any better i feel the same.

Often those conflicts are only on hte surface.

For example you probably recieved the advice to not get weak groups to avoid getting killed:
Very good advice :mrgreen:

and indeed it worked here
now you are overdoing it and playing too tight.
you have to find a balance.

Inkwolf · Post by **Inkwolf** » Tue Aug 16, 2011 6:22 am

Thanks, Perceval. ^^ (Also, 'Don't play so aggressively, concentrate on making territory in the corners and sides.')

So, is this better? As far as i can see, I have weak groups all over ripe for killing, have a narrow strip of the top and a variable portion of the lower left, depending on whether white can finish cutting off my O1 wall, but the computer resigned at 130, so who am I to argue?

jts · Post by **jts** » Tue Aug 16, 2011 6:35 am

Inkwolf -

If you are wondering whether you're learning bad habits from the computer, that's enough reason to start playing humans, in my book.

If you are playing moves specifically because you know how GnuGo responds and feel comfortable with that response, then you really need to start playing humans.

Is there a reason you prefer to play against GnuGo?

emeraldemon · Post by **emeraldemon** » Tue Aug 16, 2011 6:44 am

I will just add that I've been using Fuego more (see the games in the malkovich forum) and I think it's both stronger and less predictable than GnuGo, so you might try that instead.

amnal · Post by **amnal** » Tue Aug 16, 2011 6:58 am

Inkwolf wrote:Thanks, Perceval. ^^ (Also, 'Don't play so aggressively, concentrate on making territory in the corners and sides.')

So, is this better? As far as i can see, I have weak groups all over ripe for killing, have a narrow strip of the top and a variable portion of the lower left, depending on whether white can finish cutting off my O1 wall, but the computer resigned at 130, so who am I to argue?

http://www.lifein19x19.com/forum/download/file.php?id=1093

If you try to get better by continuing to play mostly gnugo, I think you will fail. The problem is, gnugo will never punish certain mistakes. Not ever. You might get better at beating it (previous threads have shown this to be trivial with some very simple tactics, which you'll learn), but this isn't getting better at go - it's getting better at beating a single opponent who never learns. Gnugo is fun to play, and there are certainly good things to learn, but you're really only playing a subset of go because the rest never comes up in your game.

Another problem (which I thought of while looking at your game) is that gnugo never improvises, even badly. That means that not only do you not get challenged to think of new ideas, you never see your opponent play things you didn't expect.

Regarding the game you posted:
2: Generally a bad, thank-you move. This asks white to play Q3 and make a strong shimari for free - but black just gets a high weak stone. Not very enticing!

4: As others have said, you're playing this because it works, not because it's necessarily the strongest move or because you have considered anything else. It isn't actively bad here, but O16 is more common to not back down and keep an eye on attacking white later.

10: A heavy move - if you wanted to play here, why not play O3 in the first place? Now your two stones are inefficient, and a white move at L4 leaves them heavy with no base or territory.

14: The same heavy shape again. This is probably an example of never having seen the better possible shapes, and so sticking with what seems to work. E9 is a better shape that prevents the connection, and B8 (although less common) splits white's stones strongly without making such a heavy shape.

15: Should be at C16, but gnugo will never teach you this because it doesn't seem to like playing this way and will never learn to.

I'll stop here. I just wanted to give examples of how gnugo isn't a great learning tool unless you play lots of other players as well.

Bill Spight · Post by **Bill Spight** » Tue Aug 16, 2011 7:06 am

Inkwolf wrote:(Also, 'Don't play so aggressively, concentrate on making territory in the corners and sides.')

I would be very surprised if you got that advice from a dan player.

Once I got over my incremental territory bad habit, I attacked like crazy. I am glad that I did. It may have been a bad habit, but it was a good bad habit for learning. Gradually I learned how to prepare attacks and how to make dual purpose moves.

hyperpape · Post by **hyperpape** » Tue Aug 16, 2011 7:12 am

Bill Spight wrote:
Inkwolf wrote:(Also, 'Don't play so aggressively, concentrate on making territory in the corners and sides.')

I would be very surprised if you got that advice from a dan player.

Once I got over my incremental territory bad habit, I attacked like crazy. I am glad that I did. It may have been a bad habit, but it was a good bad habit for learning. Gradually I learned how to prepare attacks and how to make dual purpose moves.

Many double digit kyus will describe a play as aggressive or as an attack when it is no such thing. If you are a double digit kyu and you play your first five moves on the sixth line, the right advice is "start in the corners". Perhaps you hear that as "don't play so aggressively" but really it's "play moves with a clear purpose."

Inkwolf · Post by **Inkwolf** » Tue Aug 16, 2011 7:15 am

I do play other humans, too. Playing the computer allows for more experimentation, because I can back up and try alternate moves whenever I like. (It's also less stressful than playing real games.)

I will try some of those other programs, if I find they have a Mac app available.

Amnal, thanks for the advice.

Bill Spight wrote:I would be very surprised if you got that advice from a dan player.

At my level, I don't play many dans. ^^

Life In 19x19

Is the computer teaching me bad habits?

Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?

Re: Is the computer teaching me bad habits?