Reviewing Lee Sedol's commented games with KataGo

[vier]

My thinking is that even if AI was 100% correct and never erred, that does not mean that an AI selected move is the correct move for a human player who cannot play at the same level as the AI.

Indeed.

With this argumentation there would be nothing to learn for amateurs by looking at professional games.
If a stronger player can play a better move because of their bigger skillset, then the weaker player should acquire that skillset.

A mountaineer can learn from an experienced colleague. If he follows the path taken by a mountain goat, he will fail, fall and die.

[kvasir]

One question I have is if Lee Sedol actually wrote any of this. Ghostwriting takes many forms and it is sadly rather common.

[kvasir]

Regarding the first game:

Giving ponuki at move 56 results in 4-5 point loss assuming white is playing thickly at move 96 to prepare for the ko. This seems to me like the kind of active but bad plan that loses games but black doesn't fight the ko and therefore gives some ground back. Is it bad technique or a special plan? If I did the same it would be bad technique but I can't speak for Lee Sedol and he is well know for playing aggressively and make moves that are hard to counter in practice.

When looking at the KataGo analysis it is hard to understand why 107 is called a losing move, I can only assume there are variations in the book. This is one difficulty with using the computer, it sometimes makes it look easy. In this case 107 is one of a few moves KataGo thinks reduces black enough but this is a very hard call in practice. However, there is a human tangible here and that is black didn't fight back when white ate the stone. The mistake is to play 107 and give up the stone in a way that is really bad, maybe it is not useless to put the "losing move" at the point when black made the mistake in his mind.

After move 126 my KataGo says W+0.4. I think that raises the question if the *final* reason black lost was endgame because black loses ground for the next 30 or so moves until it is W+5.6. It must be hard to impossible for anyone to beat Lee Sedol in endgame when behind by so much but less than a half point is not a lead. Somehow I take KataGo's estimates for granted in the previous sentence without trying to estimate myself, it is another pitfall of this kind of analysis.

[CDavis7M]

The AI suggested moves are the correct moves for the AI. Sometimes they are also the correct moves for a human. Sometimes not.

With this argumentation there would be nothing to learn for amateurs by looking at professional games.

If a stronger player can play a better move because of their bigger skillset, then the weaker player should acquire that skillset - if they want to become better. There's nothing magical about AI. It's the same stones, the same rules, ... just better moves.

I argued that Lee Sedol knew he needed to play a different move and you argued that he should study how his actual move was potentially the better move all along. Of course it makes sense to study moves and see how they work better or not. But like I said, Lee Sedol knows himself best and after review of his own game he thought that he did not play the best move for himself,

He's not wrong.

[CDavis7M]

What commentaries do provide, though, can nevertheless be very useful. I think that broadly they offer two things. One is a collection of "oh, I didn't know you could do that" moves. This seems to be the major contribution of AI, in fact. Even modern pros thought "oh, I didn't know you could do that" when they saw what bots were doing with Direct 3-3s, early shoulder hits and outrageous contact plays. We amateurs get the same experience even when we just look at life & death problems and joseki books. I think this is all truly, truly valuable, even when we don't "understand" the new move we've just seen.

The other facet of commentaries that I believe is useful is, rather than promoting understanding, in reducing uncertainty.

I might be completely mistaken, but I thought the main thing that Go game commentaries provided (like all game commentaries) was entertainment. I'm not saying that some people don't study game commentaries, but I believe that the teachings are intended to deepen appreciation and enjoyment of the game. Not to improve skill.

In fact, I'd be reasonably confident in asserting that at least 90% of commentary readers end up no stronger than they were before. They may sometimes "understand" more, but then think that that understanding exempts them from the hard work of reading. As a result, they may even end up weaker.

Right, 90% of people are enjoying entertainment. Not studying.

By the way, I find your books to be entertaining. Maybe I've gotten weaker but I had fun doing it.

[John Fairbairn]

I might be completely mistaken, but I thought the main thing that Go game commentaries provided (like all game commentaries) was entertainment. I'm not saying that some people don't study game commentaries, but I believe that the teachings are intended to deepen appreciation and enjoyment of the game. Not to improve skill.

It's infotainment. When you listen to a teacher in class, that's like a commentary. You typically learn very little there, but you are put on the right path: true learning only comes when you do your homework. But if she does her job well, you will be inspired and so pay extra attention to your homework, try harder, and even maybe do some follow-up of your own with a visit to the library. You will also be guided by being alerted to tricky technical details, and shown how to think. That's the info part. Appreciation and enjoyment are bonuses which lead to wanting more info, and so a virtuous circle is created.

I don't think there's much of a market for a book that begins "Guy walks into a bar and the bartender is a gorilla," although my Go Companion might contradict that. In any event, it's a much more limited market.

[John Fairbairn]

Contrary to John's experience to train intuition, replacing subconscious thinking by reasoning has enabled me to become stronger.

I've read a few books on the brain which (combined with my own experience) have led me form some opinions, but I freely admit I can't directly make those opinions stand up.

That said, my current understanding is that there is no such thing as subconscious thinking. It's more an intricate system of information retrieval which is triggered by various stimuli, which can include conscious thinking but also extraneous events such as an itch in your armpit or a fly buzzing round your head. Furthermore, what you get when you retrieve information triggered by conscious thinking is not necessarily what you expect. Various associations (such as that fly) may trigger extra data, either unwanted or actually rather interesting.

But because what is inside your subconscious brain is based on probabilities, you can control to some degree what comes out. If you work on life & death problems with intent focus, and then see an L shape in a game, the mantra "the L shape is dead" is likely to flash into your mind. But if you work on L&D on the train home every day while eating messy snacks to stave off the rumbling tum, chatting to your neighbour, checking your emails and keeping an eye out for your stop, what might pop into your mind in the game is not the "L shape is dead" but something like "Oh, I wouldn't mind a cheeseburger right now."

In your case, I suspect that you have the ability to work intently with great focus, and that leads to the information being stored in your subconscious having a high probability of being go related, and apart from that being new information, it will also reinforce whatever relevant information is already there. Whether that leads to becoming stronger on the board can only really be tested (for amateurs) by achieving a higher dan grade, but I would expect that it would make you feel more comfortable at the board when dealing with new situations. Your uncertainty would be reduced. Since "understanding" is a loaded word in a game that is rife with uncertainty, I would argue that it's better to avoid that word. After all, we don't talk of AI bots having any need to understand. Any feeling you may have that your rationality is controlling everything is a delusion. It's just a factor that helps determine what goes into or comes out of your subconscious.

Numbers in positional judgement, endgame or semeais reduce (and sometimes eliminate) uncertainty.

I don't accept that for positional judgement, at least not in the way you do. I had a good real-life example last week of the problems with numbers. A teacher, who happens to be a mathematician, was taking a dance class for some beginners practising for a ball. There was an intricate move that was causing some problems because of the speed at which the move had to be done - as dictated by the music on his iPad. People were bumping into each other. So the teacher went to the iPad, twiddled some virtual knobs, and then came back and announced he had reduced the speed of the music to 96%. He instantly noticed the whirring eyeballs and blank gazes. Uncertainty has risen by a factor of 1.7358. And so sheepishly added, "Well, we've paid for the software so we have to get our value out of it." All he had to say when he came back from his iPad was, "OK, I've reduced the speed of the music a wee bit" (or say nothing) and the uncertainty among the beginners (which would be partly psychological, of course - nothing to do with numbers) then have been massively reduced.

As it happens, at the ball a few days later, with a live band, a glitch occurred in that movement in one set. The live accordionist spotted it from the stage, and instantly played an extra bar in the 8-bar measure. Just enough to solve the traffic problem. An iPad couldn't do that. Furthermore, the other musicians took their cues from that and all the experienced dancers in the hall also coped seamlessly with the extra bar, without having forewarning or any time to think what was happening. They just followed the flow of the music and relied on intuition to adjust their steps. I would guess many of them weren't even aware of what they did. In other words, they relied on intuition. Rational questioning would have led to collapse of the dance.

[gowan]

There are various factors other than "objective" rational thinking that affect our choice of moves. When we "think" something can be done in a position we are more likely to make self-serving moves. When we are under pressure, such as being in a kadoban, we might pull in and become too passive, or maybe we'd play unreasonably. Perhaps many of us have lost games where we had a large lead by playing "safe", allowing our opponent to catch up. We have a sense of risk when playing, perhaps connected to uncertainty, and playing to lessen risk is understandable. We are given advice to play moves we understand, which reduces uncertainty, but will we always understand the moves our opponent plays? So uncertainty is unavoidable.

[RobertJasiek]

Numbers in positional judgement, endgame or semeais reduce (and sometimes eliminate) uncertainty.

I don't accept that for positional judgement

To start with the obvious: the current territory count of the territory and territory-surrounding intersections is a number used in positional judgement reducing uncertainty by
assessing the almost settled part of the board and leaving only the rest of the board significantly more uncertain.

[CDavis7M]

One question I have is if Lee Sedol actually wrote any of this. Ghostwriting takes many forms and it is sadly rather common.

My impression from reading Vol. 2 is that he wrote it. He points out his appreciation for help preparing the book. There is surely a professional writer and editor like with many books, but this book seems to genuinely be his. Here is a snip from the beginning.

There were some tough periods during my recent six month leave of absence, but I had the opportunity to try many new things in life. It also allowed me to reminisce about my journey, and appreciate my surroundings with serenity
...
Baduk was always on my mind during my leave, even though I didn't play. Fortunately, preparing these books filled the void admirably.
...
Unlike common game commentaries... this book was designed to proivde more detailed explanations, as well as comments on my feelings and emotions during the actual game.
...
I want to express my gratitude to my family for assisting me in so many ways, Mr. Hong Taesun for gladly donating a camcorder to record my comments, [etc].

May 2010, Lee Sedol.
....

I added in various anecdotes about my brother. These stores are either things my brother told me from his perspective or things I observed
...
Lee Sena

----------

By the way, my impression is that the NHK Books from the Go Focus lectures are actually ghost written. But that is actually a good thing isn't it? The Go player has already prepared the lectures and given instruction. The only thing the writer does is write the main points and prepare the diagrams.

[kvasir]

My impression from reading Vol. 2 is that he wrote it. He points out his appreciation for help preparing the book. There is surely a professional writer and editor like with many books, but this book seems to genuinely be his. Here is a snip from the beginning.

I had some cynical thoughts when reading "professional writer" but I'll let it be. The sniped is in Lee Sedol's name and certainly does appear somewhat personal. I was more interested if the commentary is Lee Sedol's or if there were helping hands. Maybe that is a ridiculous question if you have one of the books open in your hands.

Regarding NHK lecture series. I don't have them either but I don't see a problem with a second author or editor adapting a lecture into a book. Like you say it can be beneficial or the lecturer simply isn't interested in doing the work. It is too bad if this is not acknowledged properly, I take it that it is not clear.

[jeromie]

AI analysis is not, repeat: NOT, objective. It may be helpful and provide a new view but this does not make it objective. AI can err and has erred, e.g., judged two very similar positions very differently or life and death status wrongly.

In this context, I merely meant that the AI is without emotional bias, not that it was the “objectively right” move. I realize that the word is used that way when discussing game evaluation, so it was a poor choice. Sorry for the confusion.

I fully agree that a bot, even a strong one, can have an algorithmic bias. And I agree with others in this thread that it’s possible for a move to be good for the bot, but not good for a human. But I am in some doubt that a move chosen by a player of Lee Sedol’s caliber during a game that is validated by AI analysis later was actually bad when he chose it. It could be, I suppose, that he couldn’t actually play the follow-up, or that his ideas of why it was useful was wrong and therefore he shouldn’t have played it to begin with. But I think it’s just as likely that it was a fine move in the flow of the game, but an emotional response (or incorrect re-evaluation) in later analysis caused him to incorrectly decide it was poor.

[Knotwilg]

I was more interested if the commentary is Lee Sedol's or if there were helping hands. Maybe that is a ridiculous question if you have one of the books open in your hands.

It's one of the most personal books by a pro I have ever seen. There will surely have been some help in editing but I genuinely believe the commentaries are his. The commentaries are also spiced up with stories by himself and his sister.

This is also why I find the comments so interesting, because (I believe) they show the emotional bias with which certain moves were played or even discussed in retrospect, as exposed by the more objective (I believe) AI.

I'm also trying to take Lee's strong claims more like his own positional judgment. And when you're the strongest player in the world, you're entitled to leave out the nuances of "in my humble opinion". As said, the exercise of reviewing it with KG is valuable to me.

[John Fairbairn]

the exercise of reviewing it with KG is valuable to me.

If I may, a question for both you and Robert, one I ask because I haven't a clue what the answer is. I don't study, but you both do, intensely, and so may have some insight I've missed.

Neither of you seem to have reported any significant increase in grade. At most, I imagine it would be one amateur dan. You may feel stronger, but that's more about appreciating the game as a fan, rather than as a competitor.

But at the same time little Sumire is winning game after game and has already bounded up to 2-dan in a very short space of time (I'm assuming a pro grade promotion is harder to achieve than a low amateur dan). I pick on her because she has to spend a lot of time on schoolwork and sleep longer than adults, and so in some sense is comparable to amateurs as regards how much time she can spend on go.

Have you got any sense of what she's doing that you are not doing, or advantages that she has that you don't? I'd be reluctant to put much emphasis on DNA, or even on the nebulous concept of talent. Starting young must help, I suppose, but have you any sense of specifically how this might help?

I have not found many reliable clues in the literature (including chess). I have looked, because my next book (currently being proof-read) is entitled "Ogawa Doteki, go prodigy." I have incorporated there a brief survey of many other go prodigies. One that does seem to percolate through most biographies of top players is that enormous emphasis is put on ultra-basics. By that I mean something like tactics as pure tactics. Adult amateurs seem to bring strategy into the mix far too early and justify tactics in terms of strategy (or efficiency or positional evaluation). My feeling is that they are trying to run before they can walk. I'm guessing that youngsters don't find pure tactics as boring as adults do. In go there is also a great emphasis on even talented youngsters playing handicap games (where the handicap stones take out some of the strategic complications and allow tactics to flourish unencumbered. Western amateurs are, in contrast, notorious for avoiding handicap games, and many even try to insist on playing even games with stronger players. They seem to be in love with fuseki and out of love with tactics. But is it not likely to be true that you can't appreciate strategy properly until you have a really thorough stranglehold on tactics?

Outside of go, in talent shows, for example, young people described as "talented" are usually (?always) just perfect mimics of adult stars. Can we view mastery of "tactics as tactics" as go's equivalent of such mimicry?

But I am guessing. Hence the question. I'd like to put the same question (what have Sumire and her ilk got/done that you haven't got/done?) to any other amateur who studied very hard, with or without AI, and didn't get to the elevated dannage they'd hoped for.

[Knotwilg]

the exercise of reviewing it with KG is valuable to me.

If I may, a question for both you and Robert, one I ask because I haven't a clue what the answer is. I don't study, but you both do, intensely, and so may have some insight I've missed.

Neither of you seem to have reported any significant increase in grade. At most, I imagine it would be one amateur dan. You may feel stronger, but that's more about appreciating the game as a fan, rather than as a competitor.

But at the same time little Sumire is winning game after game and has already bounded up to 2-dan in a very short space of time (I'm assuming a pro grade promotion is harder to achieve than a low amateur dan). I pick on her because she has to spend a lot of time on schoolwork and sleep longer than adults, and so in some sense is comparable to amateurs as regards how much time she can spend on go.

Have you got any sense of what she's doing that you are not doing, or advantages that she has that you don't? I'd be reluctant to put much emphasis on DNA, or even on the nebulous concept of talent. Starting young must help, I suppose, but have you any sense of specifically how this might help?

Well, it's not entirely true: I do feel a change, backed up by regaining a solid 2d rank on both OGS and KGS. But indeed I have not gone beyond that, into the 3d-4d realm, whether or not I study a lot. And I think that's where my ceiling is really, if not 2d already, due to a mixture of talent, time spent/available and age.

- Talent: this is the hardest element to measure. I've seen people with a better capability of move selection, remembering tesuji, keeping their head cool ...
- Effort spent: these days it's about a couple of hours per week. Say 4 games, of an hour each and about the same time analyzing.
- Age: I believe both Robert and I are born in 1971.

Anyway I would never compare myself to a young Asian professional. That's just another game, in all meanings of the word.

My feeling is that they are trying to run before they can walk.

On this part, I don't think I have ever shied away from tactics and have not spent too much time on opening/strategy. But I do feel that my tactics are not solid. As my mistakes series shows, there are still too many blunders and a laziness/incapability to read.