How to review your own games

[Uberdude]

Just out of curiosity, do you focus as heavily on endgame analysis in reviews as you do in your own games?

No, as most games I review are decided long before the endgame

[EdLee]

If you do this for one game each day, then by the end of the week you'll have improved on 21 things.

My pet peeve here is the wording "improved on." I have no problem if it was phrased as "worked on 21 things,"
or, "you'll have gained experience on 21 things." My feeling is that nobody, not the student herself,
and not even a pro, can say after having just reviewed 1 mistake, that she has "improved on it." --
Whether the experience (the mistake + review) translates to any improvement at all remains to be seen.
This is not something that can be ascertained in the next game, the next few games, or maybe not even
the next 20 games. If, after some period of time, and many games later, the occurrence of this particular mistake
drops significantly, then and only then can we say she has "improved," on this particular thing.

I can only speak from my experience: I make the same mistakes over and over again,
year after year, even with pro reviews. If after every review of such a mistake I say to myself,
"I have improved on it," then I'm only fooling myself -- maybe I have, but maybe I have not --
the jury is still out. I simply don't know yet. All I can say is I've worked on it.
Until the day comes when this mistake drops significantly from my games, I cannot say I've improved
on it, or that I understand it. Conversely, it also means if this day never happens --
no significant drop of this mistake -- then I have not improved on it,
at all, ever, no matter how many times I've reviewed it.

In other words, actual improvement (or, "better understanding") must be based on empirical data --
We look at my games after some time, many games perhaps, and see that a particular mistake
has dropped significantly, then we can say I have improved on this one thing.

I cannot declare improvement based on reviews alone, a priori, without any empirical evidence.

[rhubarb]

My two cents, probably worth less than $.02 on account of my rank being about yours:

When I review one of my own games, on my own, I do two kinds of thing.

1. Go through complicated sequences (usually fights) to work on my reading. Also, where I identified a mistake, look for better moves (sometimes results in deciding the "mistake" wasn't one after all).

2. More importantly, I think: I try to take a global perspective on the game, much like what I think magicwand is suggesting. During the game, it's easy to get caught up in what's going on locally--am I succeeding in my current aim? am I thwarting my opponent's current aim? did I get the better of this exchange?--and lose sight of the big picture. I like to try to see how the opening and early middle game influence the later parts--"I gave up too much of this corner to be happy locally, but the wall I got worked with my stones over here & helped me win this fight," "I left this unsettled too long, and so the direction of fighting over here let her attack my weak group & bully me," etc.

I try not to spend too long on number 1. I figure problems are better for practicing one's reading, but getting a feel for how the early game affects the later game feels like making progress toward understanding direction of play, whole board thinking, or something else good. And it's something I feel like I can productively think about at my (our) level, unlike, say, joseki. (I can memorize variations reasonably well, but really understanding joseki is still tough--esp. which to choose when. My first two moves are usually pretty arbitrary, for this reason.)

Maybe a way of putting it is this: I spend some time seeing how better to achieve my aims, and I try to spend more time figuring out what aims I should have. (& likewise for the opponent's aims, of course.)

[Splatted]

I was going to write individual replies but it was becoming a novel so I deleted it all. XD

I didn't phrase question 2 very well but a lot of people inadvertantly answered it while answering question 3, so that's fine. ^^


Your replies have all been very helpful and I now feel I know enough to go and experiment until I find what works for me. In particular, I think I was too focused on finding wrong moves and correcting them. A common theme seems to be that all of you prefer to focus on considering and exploring different moves and situations.

Correct me if I'm wrong, and thanks for all the help.

[EdLee]

I was too focused on finding wrong moves and correcting them.

Finding wrong moves and correcting them is good. The problem, as Bill said, is how can we see our own blind spots.
How can we spot moves that are 3 or 4 stones better.

[peppernut]

I like Bill's suggestion of trying to find the losing move. You can take an even more basic step than that though, and look for the general reason that you lost.

The thing is that at the kyu level most games are not that close. Either somthing died or someone made a much bigger moyo and turned it into territory. That means that the reason the losing side lost is probably related to either a life and death mistake or an invasion mistake. Were you too jealous and invaded too early or often? Were you too scared to invade? Did you neglect a move and cause a weak group to come under attack? These sorts of things are what I look for to try to find the losing move. I find it's often earlier than you think it is, and barring a l&d mistake, it's usually more of a strategic error than a single clear losing blunder.

[EdLee]

The thing is that at the kyu level most games are not that close.

Where is the data to back up that statement.
My feeling is the statistics for kyu game results are no different from dan game results: it's a bell curve of sorts.
I don't have the data, either. Curious to see the curves.

Either (1) somthing died or (2) someone made a much bigger moyo and turned it into territory.
That means that the reason the losing side lost is probably related to either
(3) a life and death mistake or (4) an invasion mistake. ...and barring a l&d mistake,
(5) it's usually more of a strategic error than a single clear losing blunder.

(Added enumeration.) No to all five.
Kyu levels, by definition, are still very shaky on the basics.
This means most kyu games are littered with basics mistakes -- mistakes in
shapes, tesuji, life-and-death, direction, counting, cap race, ko, joseki, contact fights,
invasion too deep, invasion too shallow, invasion too early, invasion too late, invasion completely the wrong idea,
reduction not enough, overplays, underplays (too slow, too small, etc.), misreads, ... etc., etc.
The list goes on and on. It's practically endless, especially at kyu levels.
At kyu levels, both players make tons of these mistakes. Even if there is a very clear
blunder, all the other moves are still full of these mistakes.
How a game is won or lost depends on each individual game, and the two individual players.
It is the outcome of the myriad of mistakes (and of course, also the good moves) by both sides.

[peppernut]

The thing is that at the kyu level most games are not that close.

Where is the data to back up that statement.
My feeling is the statistics for kyu game results are no different from dan game results: it's a bell curve of sorts.
I don't have the data, either. Curious to see the curves.

I don't have data either, but it is logical. The less skill of the players, the bigger the errors. Errors have a negative value measurable in points. So I agree that it probably will look like a bell curve, but the curve should be fatter the lower kyu you go. In short, the less skill of the players, the higher variance.

No to all five.
Kyu levels, by definition, are still very shaky on the basics.
This means most kyu games are littered with basics mistakes -- mistakes in
shapes, tesuji, life-and-death, direction, counting, cap race, ko, joseki, contact fights,
invasion too deep, invasion too shallow, invasion too early, invasion too late, invasion completely the wrong idea,
reduction not enough, overplays, underplays (too slow, too small, etc.), misreads, ... etc., etc.
The list goes on and on. It's practically endless, especially at kyu levels.
At kyu levels, both players make tons of these mistakes. Even if there is a very clear
blunder, all the other moves are still full of these mistakes.
How a game is won or lost depends on each individual game, and the two individual players.
It is the outcome of the myriad of mistakes (and of course, also the good moves) by both sides.

I agree that kyu games can be a practically endless series of errors in kyu games, but we're talking about reviewing your own games. How is a kyu player supposed to evaluate these errors when the concepts aren't clear? Some of these mistakes can be identified by a kyu player, but many cannot.

That's why I think when you get a game that's won or lost by a large amount, right there you have a hint as to where there might be a large error or a conceptual mistake.

I have a feeling though that we're just not going to agree on this point. All I can say is that I find starting from the big picture, asking myself "why did I lose this game" and looking it more generally helps me find improtant things to study quickly. I find it easier than going through all the moves and trying to identify every mistake. That's the kind of analysis you want to do with a teacher, imo.

[EdLee]

I don't have data either, but it is logical. The less skill of the players, the bigger the errors. Errors have a negative value measurable in points. So I agree that it probably will look like a bell curve, but the curve should be fatter the lower kyu you go. In short, the less skill of the players, the higher variance.

Not necessarily. Because there are two players. When both players are making these big errors,
the net effect of all these big errors can be... a very small win-loss margin, or a medium margin, or, a large margin. Who knows.

I agree that kyu games can be a practically endless series of errors in kyu games, but we're talking about reviewing your own games. How is a kyu player supposed to evaluate these errors when the concepts aren't clear? Some of these mistakes can be identified by a kyu player, but many cannot.

Agreed. Correct. (1) Many, maybe even most of these mistakes, cannot be identified by the (kyu) player.

That's why I think when you get a game that's won or lost by a large amount, right there you have a hint as to where there might be a large error or a conceptual mistake.

This is where we disagree. Because of (1) above, which we agree, I can only speak from my experience (pro reviews),
and from watching hundreds of other (pro) reviews (of both kyu and dan games),
the win-loss margin does not, in general, tell us where we went wrong.
Sometimes, it is one place (or even one move) -- which may not be easily identified by the player;
other times, it is many mistakes. Either way, it goes back to (1).

I have a feeling though that we're just not going to agree on this point.

That's OK.

[Polama]

The biggest difference between a review and a game is that you can go backwards and forwards in time during a review. Perhaps you got to a point where you've got no choice but to form an empty triangle to keep some stones connected. During the game you were reading "do I really have to do this? Ok, I guess I do". How you got there wasn't relevant. But in a review, it is. Perhaps the local configuration of your stones and your opponents stones aren't great for you. Could you have played something slightly different earlier? What if you had attacked your opponent here before he attacked you?

Reviews of games of weaker players often include a series of moves where the reviewer says "excellent! White can basically win the game by playing here. Oh, he didn't. Well, now black can stop that by playing there. Nope, he didn't. Ok, they're still playing yose moves...no, don't need to defend that, it's alive. Nope. Nope. Ah, white finally saw it and he's now way ahead." Once you spot a move where somebody got ahead, maybe cutting a group or walling off the center, go back and ask yourself "could the other player have prevented this the turn before?" "Was the attack available the turn before that?" each turn until you think you've found when that area should have first been played. Then just look at that board position for a while, trying to connect it with the move, trying to figure out what reading would have lead you to see this.

In a commented professional game there will often be a 20 move sequence demonstrated that ends with one player capturing or clearly ahead, to explain why they didn't take an obvious move. Knowing the sequence, I try to recreate the thought process of reading that out. Not just reciting the shown sequence, but trying to figure out how they saw the tesuji involved, which other paths they pruned out and why, that sort of thing. I'll get 3 moves in, and there will be 5 reasonable looking moves, and I know that for some reason 4 of these were pruned early, so I try to figure out what about the right one makes it better. Can I read the other 4 far enough that it starts looking bad? Can I use how the sequence shown went to figure out where the other choices are likely to have gone wrong? Is there something with # of liberties going on, or good shape, or a tesuji I should know? You can do the same thing with capturing races or other fights in your game. Whether you won the fight or lost it, can you go back to the start and, now knowing how it ends, read it out?

[Bill Spight]

These sorts of things are what I look for to try to find the losing move. I find it's often earlier than you think it is, and barring a l&d mistake, it's usually more of a strategic error than a single clear losing blunder.

Sensei's Library has a very nice feature, the Big Question Mark (BQM), in which players often ask how to play from a certain position of the whole board. Very often the real answer is, you already made a mistake.

[EdLee]

Bill, wow! Big Question Mark