Hi,

I am new here in this forum. I am currently 12kyu on OGS (dropped from 9kyu, probably am still around 10kyu theoretically).
I want to know how to review my own games so I can improve. My goal is shodan. I don't need any harsh time limit (like the one year plan often aspired) but I really want to see some progress in my games/thinking.

So far:
I play mostly against bots on OGS, specifically Spectral-10k. I usually look at my games with AI and look for the moves that turned the tables. I don't understand most of the moves the AI recommends, so I don't delve into the variations.
Then I try to conclude what the problem was and try to remember the strategy that I should've followed (that I think is suggested by the AI).
If it is Life and Death, I blame it on my shape. If it is somewhere in the Corners in the beginning, I blame it on my reading (basically Joseki knowledge but I don't memorize Joseki as most pros say you don't need them anyway on my rank). If it is not Joseki/Life and Death related and still in the opening where my percentage goes down, I blame it on my direction of play. Maybe you can already guess I oriented myself around this read here: https://www.reddit.com/r/baduk/comments ... advice_to/

This is very vague though and I want to try a different approach. I think I can reach better results when I first review my games by myself and reconsider what I was thinking and maybe should have thought. After some variations I would then turn to the AI and verify (or rather falsify) my thoughtprocess.
This way I imagine to get more out of my games because my thoughts are much more clear/refined.

Thus the question arises:
How should I review?

Should I play out variations?
What phases of the game are there (except endgame because often I resign before I reach it) that I should focus on?
I often find myself in situations where I simply don't know what (!) to play. I think I lack having a gameplan.

Some help would be much appreciated.

A few ideas:

• Stop playing "mostly against bots". To improve faster, you should play "mostly against humans".
• If you can extract one useful idea out of a game review, that's already good. You don't need to understand every move suggested by the AI.
• At (y)our level, most mistakes are elementary. Just looking at the shape, or at variations 1-3 moves deep, is usually enough to understand your mistake.
• If you want to understand the AI's suggestion, try to refute it with your own moves. For instance the AI plays , you don't understand that move because you think can punish, so you play it out, the AI continues with , you think is a good refutation, the bot responds with and White is dead.
• Playing out the AI's variations from time to time can be interesting. Even if we can't play like that, I think it opens up our mind and shows that our thinking is too limited.

What you have been doing sounds good, as well as your ideas about how to proceed. So I say bravo!

But if I may, why not try asking questions of the AI? It says to move there. OK, what if I, as its opponent, reply here? Or, instead of its play, what about another move? What, if any thing, is wrong with it? IOW, play around with it. And I do mean play. None of this should be a chore.



Excellent!



Of course. In the spirit of play. With or without AI.



Don't worry about that.



Join the club. It wasn't until I reached shodan that I almost always had some idea where to play.



Coming up with a game plan is ambitious, but a good ambition. For starters, you can formulate goals during the game. For example, make life here, kill stones there. Or sacrifice a few stones here to build a wall. Try to take sente as a general rule. And when you do have the move figure out something to accomplish with it. OC, your opponent may thwart you ideas, but you can learn from that, too. Formulating little goals during the game is a good way to learn, even if you don't win, and even if you don't review. It also gives you focus when you do review. Ask the AI if you could have accomplished a goal that failed, or if there was a better plan, for instance.

Good luck!

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Since you asked "how to review", I'll answer that, while the general question may be "how to improve".

Most players don't review, which is a mistake. Those who do, like you (and me) can make the other mistake: put too much time in reviewing:

1. inspect every move
2. playing out enormous variations
3. aimlessly

A good review should be helpful towards the future. These days, I use AI to find turning points in the game, especially where I had a big lead and still lost, or where the game fluctuated. My latest game review showed:

- think twice before peeping at (what can become) a bamboo joint
- if the opponent has claimed 3 line territory, consider forcing him into it, building influence
- even major influence in the centre can use an extension to become really thick

That's three things already to try implementing. More would be overkill.

Playing only bots will, after one or two hundred games, teach you to play bots. Play hoomans. Find a teacher. They are out there.

You may have misunderstood the advice about joseki; it is not that you need to memorize patterns, rather you need to have a fundamental understanding of the relationships between the stones, which is different. My teacher showed me a common joseki and then removed the white stones. The black stones left a shape that was easy to see as stable and the relationships between them were,shapesmi already knew. It’s the cuts and sacrifices that mess up joseki for beginners.

Once you grasp why a series of plays was made, the series itself becomes much easier to play out without memorization.

Good shape is always good to know. Bots don’t play for shape, they play far into the future. That’s why bot reviews are difficult or useless for beginners—anyone under, say, 5k.

Life and death issues do not arise if your shape is fundamentally good. After your groups stop dying, you can start reaching further And taking bigger risks like reductions and invasions, but a bot will always be five or ten moves ahead of you.

If you set the bot too low, the responses are weirdly randomized and don’t help you learn anything.

Go is unlike any game you’ve ever played and getting better requires that you unlearn some basic western concepts such as kill, win, take, steal, and dominate. Go is not like the mafia. Go instead is more like running a legitimate business; it is all about influence, patience, and efficiency. Force does not enter into it.

_________________
David Bogie, Boise ID
I play go, I ride a recumbent, of course I use Macintosh.

If you MUST play only/mainly against bots do NOT play against bots supposedly your own strength. As already noted, this would teach you only to play against bots (of your own strength). Instead:

a) Play against bots set a good 3-4 stones stronger than you taking a stone less handicap than indicated or otherwsie adjust so that you are losing 2/3-3/4 of the time. When you have improved to where winning 1/2-2/3 of the time, increase the strength setting of the bot.

b) If possible, do this with more than one bot. In this case, more than one means bots using different algorithms. At your current level there are plenty of bots that can play stronger enough than you that you could learn from even of the pre MCTS sort (to say nothing of neural nets). In fact, that's what you probably want as a MCTS bot set that low would be erratic.

Thank you for this. Very poetic and very true

First of all: Thank you all for your replies. I think some are very helpful.

I want to elaborate more on playing bots. Bots are playing instantly and you can play them whenever you want, that means you have a way higher game throughput per hour as opposed to playing against humans, which in my opinion is the most valuable measure.

But how about playing handicap against Katago?


I will summarize some suggestions that I personally will (try to) implement:


What I'll take away from this is to think broader first and then deeper.


This sounds hard to me as I want to see the bigger picture but I guess your right.


This seems obvious now.


I'll stick to one thing, as I think is more appropriate for my level. But how do I come up with "lessons" like this in the first place?


You said this is bad (just for the record) and I agree. Ill stop that.


This is insanely helpful, thanks.


At last I can give some advice back:


This is actually blatantly false and easily to be proven wrong. These are not "western concepts" but rather human concepts. Personally I don't care but some people might get offended about this. You really should think about what you are posting.

The reason why playing humans is better is that human play is much more varied.

If you play against weak bots (=around your level), you will learn to punish their mistakes, but the moves that you will find don't neessarily work against stronger players. Moreover, you won't be confronted to a variety of board positions. For instance the bot will always play the same josekis, but when you play against a human who deviates from the joseki, you will be at loss about how to respond.

This doesn't mean you shouldn't play against bots. A bot which is a few stones stronger than you will teach you better moves.

If you are around 10k, Katago is way too strong, you won't be able to beat it with 9 handicap stones. But you can try.

Hi. From a fellow kyu, I would like to share what I do.
I, for the time being, focus on just two aspects. That is, to be mindful of my and my opponent's weak stones, and to prevent a split shape (as described in sensei's library. btw split shape also includes ripped one space jump). I can win/lose a game by other reasons, but these two are what I focus on during my review.
When I think I have (sort of) mastered them, I move on, and find other aspects that I would like to improve.
There are too many aspects in the game to look after, and strong players probably have built them into their intuition. But for us kyus we probably should focus on a few things, when playing and during reviews.