Disaster against CrazyStone

holocen · #1

Please analyse this loose of mine against CrazyStone level 2, it is very hard to me to beat this engine even on low level. Thank you.

lobotommy · #2

How to say it in a polite manner...
1) Play more games.
2) Study tsumego and tesuji problems,
3) but first play more games.
4) Invest in some books for begginers. Read them.
5) Come back in 4 months.

If you are 20 or something kyu there is nothing to analyse, every move is just horrible and you should play more on 9x9 and 13x13 board. It's hard to tell you anything meaningful because you need to improve in every aspect of the game. Reading, local and global positional judgement, fuseki and joseki, "life and death" problems, etc...

If you want to see how the game should look - look at games of players stronger than 15 kyu, then 10 kyu, then stronger ones. Compare them to your way of playing and ask yourself for meaning of the moves. At least you'll gain some visual tips how to develop a game.

schawipp · #3

I tried to put some basic comments within the first 50 moves to give you a first idea. Some important takeaways:
- Look up the shape "empty triangle" and try to avoid it in your games
- Don't let the opponent encircle you easily. Try to extend from your stones along the border or into the corner in order to make a base.
- Just play a couple more games (and don't think too seriously about the moves in the beginning...) - maybe also on 9x9 or 13x13 - to get a first feeling, and don't get discouraged. Even after a few years of experience (~3 years in my case) we all still feel a bit like absolute beginners (don't we?...), that's a special feature of this game ;-)

Edit: Some typos...

Mike Novack · #4

Also important?

Do not use this sort of go engine set weakened to such a low level of play. The MCTS based algorithm becomes erratic if weakened that much and makes gross mistakes. Yes of course, you were making even worse mistakes and so lost but you can't learn much that way.

If you need to be playing against the computer better to use one of the programs that use a go knowledge based AI for the lower levels of play. These are all that we had before the MCTS based programs were developed and the AI programs were up to about 6 kyu at that time. So plenty strong enough for you to learn from.

I suggest this because those programs are weakened (for the lower strenght levels) not by making random bad moves but by "not knowing something" much as a human go student might not have yet learned something.

But especially when using one of these do not weaken it so much that you can play even go against it. That is what people are warning about when they say you would learn bad habits from the machine. You want the program not to be making the mistakes typical of (human) go players at your level. The mistakes typical of players a few stones stronger aren't as much of a problem as you haven't gotten there yet and you can keep the program about 3 stones ahead. Once you have reached the limit for the AI based programs you will be ready for the MCTS based programs at a level of play where they won't be making as stupid mistakes as in your example game (go and look about move 170 ---- at this point you should not have been able to form any live groups in the center)

Understand what I am saying? You can learn something where the computer opponent is making some systematic mistake but you learn little if every so often it throws in a randomly bad move.

Bill Spight · #5

You might try playing against this program on the 9x9. You should also take a handicap from it by placing some Black stones on the board before White makes the first move.

Some comments:

Main focus: Think big.

Also: Avoid the empty triangle.

Good luck!

holocen · #6

@lobotomy
1) I haven't played much games, maybe 50 on 9x9, 7 on 13x13 and 15 on 19x19 board. You are totally right.
2) Tsumego is much easier to learn than tesuji.
3)Sometimes I play tired and in hurry
4) I have about 10 GB of Go books as pdfs
5) I will back when my rank get better

@schawipp
Thank you for the empty triangle advice. Comments are very useful.

@Mike Novack I didnt think of engine that way, it is true.

@Bill Spight Thank you for the comments.

Mike Novack · #7

Look, at your current level the Many Faces of Go free demo version* should be about the right amount stronger than you are (its about 18 kyu and is an AI engine, not MCTS).

I think you will need to have gotten up close to the single digits before any of the MCTS programs will become useful. When you get back to report to us that you are ale to beat the MFOG demo version maybe we can suggest what to try next.

* I do not know if this free version has all the useful tools enabled like being able to ask it "why did you make that move".

jts · #8

holocen wrote:

@lobotomy
1) I haven't played much games, maybe 50 on 9x9, 7 on 13x13 and 15 on 19x19 board. You are totally right.
2) Tsumego is much easier to learn than tesuji.
3)Sometimes I play tired and in hurry
4) I have about 10 GB of Go books as pdfs
5) I will back when my rank get better

@schawipp
Thank you for the empty triangle advice. Comments are very useful.

@Mike Novack I didnt think of engine that way, it is true.

@Bill Spight Thank you for the comments.

Personally I would say that if you need to play Go against a computer, as a beginner, you should find the strongest go ai you can, and play it on the strongest setting, on a 9x9 board, with a high handicap. Seriously, start with nine stones on a nine by nine board and lower it every time you can beat the machine.

But I think playing other human beings on KGS or elsewhere is probably better. It's good to have a wide range of experience with playing styles, different handicaps, and different social attitudes towards go. The biggest problem is that if you play one or two opponents all the time you start to expect them to react in comfortable ways, and so you start playing to get an expected response instead of thinking about the best response. But also humans can review games, introduce you to other go players, and add an extra level of intensity.

I hope you don't find Lobotomy's comments too harsh. I think on the internet, without nuance or tone, and especially where multiple people who perhaps don't speak English well are conversing, it can be hard to communicate exactly. The fact is that anyone who has played Go for more than a year knows how bewildering it is to play Go at first and how much you learn from game to game. There's also the wrinkle that it's easiest to sympathize with someone's mistakes when you're at the same level of development as they are, but it's easiest to correct their mistakes when you've been playing much longer. Beginners are in a tough spot. So I agree with Lobotomy's point (minute for minute, you'll improve more and have more fun if you play another game rather than reading another game review), but you shouldn't understand him to mean "leave the forum until you have an awesome rank, I don't want to see you until then". I assume he meant something more like, "you will be absolutely shocked by how much you improve in the next four months, trust me, it's awesome".

By the way, there are both easy and hard tsumego and tesuji problems, as well as books that describe techniques for killing groups and finding tesuji.

lobotommy · #9

holocen wrote:

@lobotomy
5) I will back when my rank get better

As jts wrote - english is my second language and clearly I lack empathy and subtleness while using it. Please, don't take my words as banning you from the forum (I really have not intended something like that...). Take them rather just as an advice that a some amount of time need to pass before your games will turn into something which can be fully analyse and get something meaningful out of them. Please wrote your thoughts about a game, you feelings, ask about problems you encounter etc. but a publishing a full games for reviews - ekhm, well, maybe not yet (you'll laugh just after a month or two when you'll see your todays games).
Play games, enjoy them, and give yourself a time before you start serious study with games reviews. And remember - It's just my opinion, not any kind words of wisdom. So... Have a good game!

leichtloeslich · **#10**

I kind of agree with everything that has been said up to this point:

1. Play on small boards only. (At least until you get the basic tactics down.)
2. If you want to play bots, play strong bots on small boards with handicap. (Fuego is a strong free bot, for example.) I would start with 5 stones on 9x9, though. Anything above that is just ridiculous. (Even 5 stones is almost impossible to lose.)
3. If you want advice from stronger *human* players, ask for a teaching game. On KGS there are usually people willing to play absolute beginners, check the KGS teaching ladder room. You can also make an account on the Dragon Go Server and challenge this user (please read the userinfo carefully before issuing a challenge). Note that Dragon is a correspondence go server where moves can take several days, so it's not the best way to get strong quickly.

Mike Novack · **#11**

Learning style matters. When we give advice like "play a lot of games on small boards" we mean that for most people, especially those who can only learn by hands on doing, that's the best way.

But when somebody tells me that they have obtained lots of books I am going to assume that this is a person who at least thinks "I learn well from books". I was. I never had anybody to play against (and this was before even a computer opponent) and I was at about 10 kyu (judged by when I finally found a go club and had opponents).

OK --- how long have you been playing go?
What books are you trying? (maybe not the right "fundamentals" books)

The suggestion to have an opponent much stronger and taking not quite enough of a handicap is a good one. Vital to the game is "maintaining connection" and at an even more fundamental level "making one own's stones already on the board more useful and the opponent's already placed stones less useful".

tekesta · **#12**

As a beginner, you would benefit from replaying a couple of pro games a day. Just a couple; when you have more experience you can replay more game records per day if you feel like it. You will benefit from replaying kifu in two ways. One, your mind will become accustomed to what happens in a game of Go. One important benefit of this is gaining a feel for what is good shape and bad shape. As well, you will have a feel for which techniques are used at what time; one common mistake I see with novices is the use of monkey jumps against established groups in the late opening and early middle game. The other is developing bit by bit the mental stamina needed to play a game of Go. Mental exhaustion can really mess things up even for talented players, so developing mental stamina is essential for any aspiring Go player. A third benefit is that you will pick up ideas that you can use in actual games.

As for problems, for now concentrate on life & death and tesuji. These are the bread & butter of every Go player. They will allow you to make your groups alive or increase their size and kill or reduce the size of your opponent's groups. After you gain a little more experience and develop a good command of the above two skills, you can learn joseki. Knowing how a joseki works requires good knowledge of tesuji and life & death. As well, whole-board thinking is needed to know how to apply joseki. Which is why it helps to do opening and middle game problems. Alongside L&D and tesuji, do endgame problems since many close games are decided in the last few dozen moves. Endgame techniques such as hane & connect on 1st line and monkey jumps (knight's jumps to the first line from a second line stone) are good to know.

When you replay a kifu, you get to see how it all comes together.

In the beginner stage, you will be taking handicap in most of your games against stronger players. When you play as Black with star-point handicaps, you should not concentrate on getting territory. Rather, you should do two things. One, increase the influence of your stones by using one-point jumps and two-point jumps; this will increase your chances of winning a fight when White invades. Two, always try to keep White's groups divided; since Black's influence is greater across the board, White must secure life for her groups and keep the Black stones divided. Black will gain territory as a result of attacking White's groups. Since playing well on star-point handicaps requires knowing how to use outward influence (towards center of board) in addition to knowing about life & death and tesuji and being able to apply it during play, you many want to replay games by Awaji Shuzo, Takemiya Masaki, and Go Seigen. Awaji is known to play a very aggressive influence-based playing style. Takemiya is known for his "cosmic" style of play and many players study his games when learning about how best to use outward influence.

I cannot guarantee you will go on a winning streak by applying the above advice, but you will gain an increased understanding of Go and how to play it.

billywoods · **#13**

The bot seems much stronger than you currently. (Don't be disheartened by this: bots are very good at calculation, so it's rather difficult to make one that is below 15 kyu unless it plays completely randomly.) Maybe try to find some opponents of around your strength online.

Disaster against CrazyStone

Who is online