It is currently Thu Oct 31, 2024 4:30 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 26 posts ]  Go to page 1, 2  Next
Author Message
Offline
 Post subject: The new standard of online play
Post #1 Posted: Wed Mar 16, 2022 3:40 am 
Judan

Posts: 6230
Liked others: 0
Was liked: 793
In earlier years, byoyomi-only online dan games were as much about strategy as tactics. In recent years, play has become more aggressive. Maybe this is so with more impact of Chinese amateurs and copying AI style. Now, a player is likely to lose unless he does the following at all and correctly on each move, usually within a few seconds:

1. verify life and death, or connection, statuses

2. positional judgement

3. do not miss moves of large value during opening and middle game

Time is too fast for always correct endgame but at least play almost correct endgame. Strategy, however, is almost immaterial; you can play what you want as long as it is not obviously bad and you apply the aforementioned principles.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #2 Posted: Wed Mar 16, 2022 4:23 am 
Lives in sente

Posts: 967
Liked others: 24
Was liked: 173
Rank: panda 5 dan
IGS: kvasir
I don't play much byoyomi only games except Canadian byoyomi and I don't think that is what you mean. So I can't say I have the same experience.

Is strategy and style something that can easily be distinguished? In my experience many strong Japanese and Chinese players choose to play a style that empathizes basic fighting principles, life and death, and in general not being too flashy. That is not what everyone does but it sounds similar to what you describe and I don't know if it is better described as a "style" or a "strategy". There are certainly many players that consistently hone the same strategy to the extend that it becomes their own style.

That three point process for every move seem pretty good. Maybe it is pessimistic to say someone is likely to lose if they fail to do it? Maybe the player that can do this more consistently than his opponent is much more likely to win.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #3 Posted: Wed Mar 16, 2022 4:48 am 
Judan

Posts: 6230
Liked others: 0
Was liked: 793
Here is an example.



Attachments:
LD.sgf [3.8 KiB]
Downloaded 403 times
Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #4 Posted: Wed Mar 16, 2022 9:00 am 
Lives in sente

Posts: 1223
Liked others: 738
Was liked: 239
Rank: OGS 2d
KGS: illluck
Tygem: Trickprey
OGS: illluck
I feel like that's always been the case for the faster time settings. I honestly feel like dan games, say, 8-9 years back are similar to dan games now (though my reading has sadly deteriorated haha).

Edit: Can only speak for low dan games, most of which are decided by point 1.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #5 Posted: Wed Mar 16, 2022 10:49 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
What most people probably realize by doing AI reviews is that the opening and the end game don't really matter much and all major swings happen in the middle game, mostly by either losing sente with a slow local move or losing a group by playing a big move elsewhere too hastily. There are more heuristics, for which I refer to my mistakes series both here and on SL.

So the main difference I see is perhaps not so much more fighting but more awareness of sente. Long drawn out joseki are out of fashion because most of the time there's a way to take sente in exchange for a minor local loss. In the early days post AI we saw a lot of early 3-3 invasions and a board full of those but nowadays I see more 3-4 points with 3-5 approaches and even approaches to 4-4, mixed up with those invasions. We even see pincers to the approach, if the pincer cooperates with friendly influence at the other end.

We have also learnt that taking central influence in exchange for side territory is only viable if there's influence at the other side too, forming a framework of sorts. Just taking influence for future use is not so valuable, as the opponent can more easily reduce it than one can remove any territory. Lone boshis are rarer, we attack more sideways to make territory while doing so.

When "the stones go walking", it's a sharper race than before, struggling for getting ahead and/or take sente. Amateur games look more professional there now - there's no room for complacency. Before, amateur games looked "professional" in the opening, only to fall apart quickly after.

Human pros have always been playing away earlier than human amateurs, shifting back and forth between positions. We see AI do more of that but amateurs don't emulate that behavior so easily.

In most of my games I see no significant mistakes in the endgame, except for some outright blunders. Most of us seem to know the shapes, understand what's big or small and how to keep sente in the engdame.

As an end note, yesterday I played a 2d on OGS who didn't seem to know the L-group. We played it out till the 3-point eyeshape with my stone in the middle remained. It's only one game but it's not exactly a sign of greater tactical awareness.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #6 Posted: Wed Mar 16, 2022 11:07 am 
Lives in sente

Posts: 902
Location: Fort Collins, CO
Liked others: 319
Was liked: 287
Rank: AGA 3k
Universal go server handle: jeromie
RobertJasiek wrote:
Here is an example.



If your comment is primarily pointing out that this doesn't adhere to traditional fuseki, this is a fine example, but it seems to me that there were plenty of strategic choices in the opening of this game. Beginning with move 15, black aimed to carve out the base of the white group with the aim of making an attack on the outside beginning with move 31. Black was able to end in sente and take a big point on the side with 41, which isn't too odd even if making a corner enclosure or approach move first would be more typical. That's a 14 move sequence (by black) that dictates what type of game will be played. Isn't that fairly strategic?

There's a lot of fighting after the opening, but that seems expected when one player uses the early part of the game to build up influence. I see the highly tactical middle game as a consequence of black's early strategic decisions (white seemed happy enough to go along, too). In what way would you say this game is lacking in strategy, as opposed to simply not following traditional fuseki?

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #7 Posted: Wed Mar 16, 2022 11:53 am 
Oza

Posts: 3698
Liked others: 20
Was liked: 4660
Quote:
So the main difference I see is perhaps not so much more fighting but more awareness of sente.


I think this is on the right track, but needs more precision for amateurs.

There is a big difference between sente (as it is most commonly used) and the initiative. An obsession with the former leads to ignoring the more important latter. I suspect the roots of the problem lie in two places. One is, specifically, an obsession with boundary plays and counting, which leads to not seeing the overall picture (or if you think you are seeing it, are you seeing it as clearly as you could?). The other is testosterone. Pushing the opponent around, or invading, feels good because it is sente, sente, sente. But it's just like maxing out on your credit cards. There comes a time when the bill is due AND you have to pay interest. Yet you can have the initiative without having sente, and it actually pays YOU dividends!

Sente, sente, sente is just a step up from atari, atari, atari. No half-decent amateur would play a-a-a, so why play s-s-s?

But, that said, the initiative is hard to gain and hard to handle once you have it. Historically, we could say the Japanese went overboard in investing in thickness (in all its senses) and avoiding risk by relying on no komi or small komis, and it took the Koreans to show, in very recent times, how risk-based strategies can pay off (I'd say they were aided in part by shorter time limits, though). Now, AI is apparently telling us even more about the initiative (and next to nothing about sente, incidentally). My own impression is that we (as amateurs) are hampered in getting to grips with that because we lack a framework terminology for it. Pros make up for the lack of terminology by dint of sheer and intense study, so that their intuition is trained to map all the factors inside their brains. You don't need words when you have intuition.

For amateurs, who haven't got the time to build that level of intuition, words are close to essential, as is a tailored course of study. One I would suggest is that, before you embark on AI study, you study Chinese master games of the past. They were not burdened with the deceptive word sente, and instead paid very much to the initiative. The difference with old Japanese games is stark.

Then, to bring this closer to the modern age, this could be followed up by studying what I like to call ley lines. Just as Stone Age man was apparently able to see complex patterns at ground level that we need satellites to see, AI operates with ley lines that connect all the significant parts of the board together. One reason I see value in studying old Chinese games is that they were masters of reading ley lines - they had to be, because group tax meant they had to worry all the time about keeping connections intact, even over great distances. Necessity was the mother of invention. They invented their own relevant terms, too. One was zhaoying (call & response), but they also emphasised "crowding" or pressurising, as opposed to attacking, which kept the INITIATIVE (and so hindered opposing connections, while promoting one's own).

Knotwilg has usefully emphasised several times his realisation that slow connections are bad. As far as I can recall, the main lesson he has taken from that is simply to try to avoid them. A good start. But potentially passive? Maybe a better appreciation of the initiative (and not sente), as per AI play, would teach us not to be so passive. Perhaps we should be thinking more about FAST connections (e.g. via call & response), and connections that lead somewhere, even if we don't understand quite where - like ley lines!

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #8 Posted: Wed Mar 16, 2022 12:03 pm 
Judan

Posts: 6230
Liked others: 0
Was liked: 793
As to the example game, I do not say that there would be no strategy. Rather that choosing a particular strategy in a particular position is not so important in such games of constant L+D reading.

Knotwilg, "no significant mistakes in the endgame": Endgame mistakes tend to be relatively small compared to important mistakes during opening and middle game. However, lots of small endgame mistakes can accumulate and go unnoticed even in dan games. When both avoid the big mistakes and make relatively few endgame mistakes, they still do decide the game.


This post by RobertJasiek was liked by: Kirby
Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #9 Posted: Wed Mar 16, 2022 12:43 pm 
Honinbo

Posts: 9550
Liked others: 1602
Was liked: 1712
KGS: Kirby
Tygem: 커비라고해
John Fairbairn wrote:
Historically, we could say the Japanese went overboard in investing in thickness (in all its senses) and avoiding risk by relying on no komi or small komis, and it took the Koreans to show, in very recent times, how risk-based strategies can pay off (I'd say they were aided in part by shorter time limits, though).


Going "overboard in investing in thickness" is a "risk-based strategy", because it means you are playing too slowly, thereby increasing the chances of losing the game. If it's an issue of komi, strategy should adapt to play well according to the rules of the game. On the flip side, if a particular strategy increases the overall win rate of a given pro, there is evidence that that strategy was less risk-based. Professionals are in the business of consistently winning games, and a strategy that is consistent with that is not a risky one.

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #10 Posted: Wed Mar 16, 2022 12:45 pm 
Honinbo

Posts: 9550
Liked others: 1602
Was liked: 1712
KGS: Kirby
Tygem: 커비라고해
RobertJasiek wrote:
As to the example game, I do not say that there would be no strategy. Rather that choosing a particular strategy in a particular position is not so important in such games of constant L+D reading.

Knotwilg, "no significant mistakes in the endgame": Endgame mistakes tend to be relatively small compared to important mistakes during opening and middle game. However, lots of small endgame mistakes can accumulate and go unnoticed even in dan games. When both avoid the big mistakes and make relatively few endgame mistakes, they still do decide the game.


I have seen this happen in an AI review of one of my own games. The AI suggested that I had something like a 98% win rate, but the overall difference in score was very small. I made an endgame mistake at nearly the end of the game, and suddenly, the win rate dropped to something like 1%. This is rather rare, since as you state, most of the big point swings in my games happen earlier :-)

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #11 Posted: Wed Mar 16, 2022 2:03 pm 
Lives in sente

Posts: 967
Liked others: 24
Was liked: 173
Rank: panda 5 dan
IGS: kvasir
Claiming "opening doesn't matter" or "endgame mistakes are small", besides not being true, doesn't change that opening and endgame need to be played well and preferably better than the opponent.

Endgame starts early in many games and it is hard to get a good handle on the big endgame, it is also often more than half the game. Opening is just the start of the game, if played badly then you could equally give handicap and play a little bit better. What I notice, when reviewing with computer, is that when any opponent plays the opening or endgame especially well it is like I never had more than a slim chance in the game.

The opening is how you get to a middle game that you can handle and endgame is how you finish it off. It has to be important :scratch:

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #12 Posted: Thu Mar 17, 2022 1:11 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
kvasir wrote:
Claiming "opening doesn't matter" or "endgame mistakes are small", besides not being true, doesn't change that opening and endgame need to be played well and preferably better than the opponent.

Endgame starts early in many games and it is hard to get a good handle on the big endgame, it is also often more than half the game. Opening is just the start of the game, if played badly then you could equally give handicap and play a little bit better. What I notice, when reviewing with computer, is that when any opponent plays the opening or endgame especially well it is like I never had more than a slim chance in the game.

The opening is how you get to a middle game that you can handle and endgame is how you finish it off. It has to be important :scratch:


At the low-mid dan level, opening and endgame are fairly well understood and it's comparatively hard to make mistakes in these stages that make a difference to the end result. Among the 150 significant mistakes I analyzed in 2021 there was one endgame move and one opening move. All 148 others were middle game moves and bad choices in that conceptual frame. Compound effects do exist but when compounding larger mistakes the effect will be even bigger. Perhaps that just pertains to my own game but I reckon it's representative.

The "big endgame, which starts early", in my frame possibly belongs to the middle game. For me the endgame starts when all groups are settled, all areas are contested and boundary plays remain. Occasionally such boundary play is played before all groups are settled but I consider that a high level probe.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #13 Posted: Thu Mar 17, 2022 1:36 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
John Fairbairn wrote:

Knotwilg has usefully emphasised several times his realisation that slow connections are bad. As far as I can recall, the main lesson he has taken from that is simply to try to avoid them. A good start. But potentially passive?


In most of the cases I analyzed there's an active way to avoid the slow connection:

* When in the dominant position, the more active alternative to the slow connection is a surrounding move or an eye space reducing move, in any case a harassing move that forces the opponent to reply and doesn't give any time to execute the cut which the slow connection defends against. The cut would be heavy anyway ("heavy cut" is another observed pattern).

* In the defensive position the active alternative to the slow connection is to live with points or escape into open space.

http://senseis.xmp.net/?DietersMistakesIn2021/Solution4
http://senseis.xmp.net/?DietersMistakes ... Solution31
http://senseis.xmp.net/?DietersMistakes ... Solution46
http://senseis.xmp.net/?DietersMistakes ... Solution47
http://senseis.xmp.net/?DietersMistakes ... Solution48
http://senseis.xmp.net/?DietersMistakes ... Solution73
http://senseis.xmp.net/?DietersMistakes ... Solution80
http://senseis.xmp.net/?DietersMistakes ... olution107
http://senseis.xmp.net/?DietersMistakes ... olution137
http://senseis.xmp.net/?DietersMistakes ... olution140
http://senseis.xmp.net/?DietersMistakes ... olution141
http://senseis.xmp.net/?DietersMistakes ... olution146

After having reworked the lot into separate problem/solution pages I'm now working on the aggregate view, listing solution pages from pattern descriptions.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #14 Posted: Thu Mar 17, 2022 2:28 am 
Lives in sente

Posts: 967
Liked others: 24
Was liked: 173
Rank: panda 5 dan
IGS: kvasir
148 out of 150 mistakes in middle game. Is this not grounds to suspect a bias in your analysis? Alternatively you could be lumping more under "middle game" than is warranted.

It can be hard to find endgame errors and fully evaluate them. I am not sure about your criteria for "significant" errors but I think you mean something that can affect the expected outcome of the game. I'll point out that studying is not only about the size of the errors is also about better technique and improving ones overall resilience as a Go player.

There are many skills that may appear insignificant. Opening doesn't seem to matter if the opponent plays it equally badly, still many games become difficult for one or the other players before move 50. Endgame and border play in general may seem insignificant if you don't occasionally play someone that can outplay you in this part. Not everyone makes it through the opening or middle game when playing stronger players but if you do you may find that you don't have the skill or actually the time on the clock to hold a small lead in endgame. Other skills are similarly limited, for example ten-thousand year ko and carpenter square never occur if you and your opponents don't understand them and the same can be said about many other shapes.

The list of unimportant skills in Go is long. So very long that it probably makes up the greater part of the skills that we lack; that is my conjecture.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #15 Posted: Thu Mar 17, 2022 2:48 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
kvasir wrote:
148 out of 150 mistakes in middle game. Is this not grounds to suspect a bias in your analysis?


I think my methodology has been fairly well documented so I won't repeat it here. If you don't like the results of my exercise, there's little I can do about it. I was not biased towards the middle game but the data pointed me there. As to what marks the middle game 1) when the distribution of stones into corner and sides takes a turn to involve the center, due to invasions/reductions or other ways of contesting the areas claimed in the opening 2) when all groups have settled and no area is to be contested anymore.

I don't think my definition of the middle game is substantially different from anyone's gut feeling. You can look at my 150 situations posted: there's a clear peak in move number in the 40s to 90s and then a long tail, depending how long the fighting continued.

I'd like to see other exercises in this fashion, i.e. more data research on the impact of good or bad play in certain stages. I don't remember who it was - you? - but someone pointed out that frequent analyses showed that the point gap is usually larger than the sum of individual point gaps, so there's an invisible compound effect going on. That's the best counterargument to my methodology up to date.

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #16 Posted: Thu Mar 17, 2022 3:15 am 
Oza

Posts: 3698
Liked others: 20
Was liked: 4660
Quote:
Going "overboard in investing in thickness" is a "risk-based strategy", because it means you are playing too slowly, thereby increasing the chances of losing the game. If it's an issue of komi, strategy should adapt to play well according to the rules of the game. On the flip side, if a particular strategy increases the overall win rate of a given pro, there is evidence that that strategy was less risk-based. Professionals are in the business of consistently winning games, and a strategy that is consistent with that is not a risky one.


I'm not entirely sure what you are saying here, Kirby. I suspect that may be because you omitted the important word "historically" from the quote, and so may be talking about something different from me.

We can easily surmise that ever since go became popular, it was populated mainly by people who saw it as a form of gambling. There is certainly enough evidence of this in recent centuries. The point is important because in gambling go it is not simply victory that matters, but the size of the victory. The prize is the number of silver taels you collected.

But when it came to prestige as the prize, a pro mindset developed in which simple victory, and then the total number of victories, was all that mattered. Life then resembled the two main styles of go - dour, long-term investment (thickness) with slow rewards over the thrill of spectacularly kills balanced by spectacular losses.

In Japan the dour style became favoured for various reasons. One was that face was important. A go master could make a living by securing patronage, but he had to be sure he didn't upset the patron. Small victories or even small, engineered losses against said patron became the safest mindset. This was reflected in modern times by the well known story of Go Seigen terrifying his fellow pros when he first played warlord Duan Qirui and demolished him. But this was not unique to China. Segoe Kensaku, in his autobiography, mentions the need for prudence in playing with Duan even before the Go Seigen incident, and remarks that this was just like playing the politicians and peers in Japan. He specifically says it was about face.

In the days of Sansa and the shoguns it may have been more about the whole head rather than just face, of course! Shoguns and daimyos avoided loss of face by not having their games recorded, or by avoiding playing the pros - just watching. But this crystallised into an effective prohibition of Japanese pros playing gambling go at all, or anything that looked like it. This wasn't restricted to go. the whole ethos of Edo life was about control, hence avoidance of risk. So the Shogun kept the daimyos clustered round the foot of Edo Castle for much of their lives, where he could keep an eye on them. Art forms such as No acting, linked poetry, flower arranging an the tea ceremony were all straitjacketed into rather rigid forms bounded by rules and made-up traditions. Even in the economy this applied. Merchants - the world's most notorious risk takers - saw fit in Japan to aim for modest wealth, to avoid sumptuary laws and expropriations. The wisest way to make money was to follow the candlestick theory of stock markets and to aim for small, consistent and relatively consistent gains. It is no accident that the favoured style of Japanese pro go uses the stock-market term souba as a description for good go.

This mindset permeated the professional ranks of all walks of life. At the lower, amateur levels, of course, chaos, speculation and just having a good time breaking the rules (e.g. kyogen, festivals, gambling go) all flourished. But what was esteemed, and ultimately led to greater fame and material profit, was professional mastery.

What distinguished professional mastery was not eliminating the risk of losing, as such. That is impossible anyway. Once you embark on a game of go, there is close to 100% risk of one side losing. Rather, the go and shogi professionals aimed to reduce the risk of uncertainty (or of disorder in other, similarly complex fields), and thereby reach a state where things became clear enough for a human to understand what is going on. At that point, they believed that with their superior training could kick in and they could "read" their way to victory, irrespective of the margin. In many cases, this amounted to a heavy reliance on boundary-play skills, a reliance which was reinforced in two major ways. One was the lack of komi, which increased the safety margin for those who relied just on winning irrespective of how much (even in the early Edo days there was an understanding that first move was worth around five points). The other was the effective absence of time limits, which allowed deep, deep reading.

Both these conditions (no komi, long time limits) persisted in Japan until modern times. The Oteai (and its equivalent in Korea) enforced no komi until close to the end of the 20th century. Hence we may assume that the traditional style of play was still in favour. Indeed, we can see this in small but insignificant sidelight ways. The self-written potted biographies that used to be included in Kido yearbooks usually included a description of go style. There were some brave souls who described themselves as fighters, but the majority would label their style as "orthodox" and no-one ever saw the need to explain that that meant solid, thickness-based play. Similarly, women's pro go was looked down on because it was mostly fighting go. This contempt was nothing like Nigel Short's contempt for the female chess brain. It was more that female go pros were relatively weak and so had not yet reached the most desirable level of "orthodox" go.

A similar state seems to have obtained in Chinese go, although with an entirely different background. There as no government patronage. Rather, patronage came from civil servants, often corrupt, for example Salt Commissioners diverting the gabelle to their own pockets rather than the Emperor's coffers, and rich merchants allowed to speculate and take risks (foreign trade included, unlike in Japan). In what might be considered the heyday of Chinese go, lavish go parties were held on the Yangzhou painted boats and masters were put up in rich men's villas. Yet even in this ambience, it seems that the best masters evolved a mindset that favoured simple victories over big victories, even though gambling go was rife. Given the absence of formal organisation of go in old China, hard data is lacking, but we can make inferences in various ways. For example, games records often would not give the size of victory (and they did, it was apparently to emphasise the closeness of a result). There are anecdotes about rich patrons trying to tempt masters into spectacular play by offering extra taels for capturing groups. And when masters passed their peaks, they would turn to writing books rather than hustling amateurs.

A safe style of achieving victory was thus favoured in old Chinese pro go as it was in old Japanese pro go, but it didn't have the same external reinforcing tendencies as in Japan, and so an argument can be made that old Chinese go began to fall behind Japan in the late Qing. This was less to do with relative skill and more to do with a deterioration in the economy, but it does seem that the masters who managed to survive (such as Zhou Xiaosong and, later, Wang Yunfeng did sill cleave to the "simple victory not size of victory" style of play.

Professional go did not exist in Korea until very modern times, so bangneki gambling go was almost always the dominant form. Indeed, it became a problem for the authorities more than once in South America when migrant Koreans took their "bad" go habits with them. But within Korea, fighting go was not looked down on. It might even be said to have been favoured. Japanese style go dominated the peaks of pro go in Korea for a long time, because the likes of Kim In, Cho Nam-ch'eol and Cho Hun-hyeon had studied in Japan. Yi Ch'ang-ho studied in Korea, but under Cho Hun-hyeon, and of course there as also the example of Korea's own Cho Chikun (and Ryu Shikun) in Japan. But fighting go, risking uncertainty, was always bubbling just below the surface, and was popular with fans. For these, Seo Pong-su, admired for having acquired his high level without studying in Japan, was the archetypal Korean pro. Yi Se-tol became another glowing example later.

Even without the presence of Yi Ch'ang-ho, it would be rash to say that it was simply the fighting skills of Yi Se-tol and his peers that brought Korean go to the top of the go mountain. It was probably more to do with the go environment they worked in. They had fewer sponsors, fewer games, and there was increasing pressure to lower time limits to make go more marketable for sponsors. We cannot discount the desire to knock Japan and its surrogates in Korean off their pedestals, either. We all know the result. More fighting games, which terrified the Bejasus out of Japanese pros, games ending in resignations more often than in endgames. Possibly for psychological reasons as much as skill factors, Koreans began to prove stronger. But skill factors cannot be discounted. Japanese pros had developed a bad habit of their own. Far too many spent the bulk of their time on the opening (an over-obsession with thick play), and while their reading skills in the middle game were clearly very good, they just as clearly weren't as sharp as those of the Koreans.

The Chinese, meanwhile, were watching form the sidelines, and analysed the Korean success. They concluded that the Japanese souba style had had its day and they copied the Koreans (and favourably re-evaluated the old Chinese masters along the way, looking at then through the prism of the modern Sino-Korean style rather than the Japanese style).

Again, we have all seen the results, but now AI is changing the environment. Indeed, we might argue that we have seen the long evolution of pro go encapsulated into the few years that AI go has been around. At one point, when Monte Carlo algorithms were all the rage, programs would sacrifice points just to get down to the ultimate simple victory of half a point. Programs like katago have moved on beyond that, and while they don't obsess with size of victory (as far as I know) they don't mind the odd big win.

I suppose it would be of some scientific interest to look at gambling go in AI terms, though. If size of victory is the criterion, would go change a lot? I mean how much would it change the style of play if the measure of success was to, say, lose 8 games by 1 point and win only 2 games by 5 points each.

There are quite a few arguable points in what I said above, but I think I have said enough to show that the word "historically" was important. I also think that the word "uncertainty" needs to be borne in mind constantly when we consider the relationship between strategy and tactics.


This post by John Fairbairn was liked by 3 people: Elom0, ez4u, jeromie
Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #17 Posted: Thu Mar 17, 2022 3:44 am 
Oza

Posts: 3698
Liked others: 20
Was liked: 4660
Quote:
I'd like to see other exercises in this fashion, i.e. more data research on the impact of good or bad play in certain stages.


Something of this ilk has been available for some time, in my Go Wisdom books. The GW indexes do not point out mistakes specifically, but do point to move numbers for each occasion where a pro deemed a concept worthy of mention, good or bad.

To give one example (from just one book) of a concept you mention, reduction/erasure, comments occur for the following move numbers:

17, 20, 21, 24, 26, 27, 34, 36, 37, 37, 37, 45, 45, 46, 48, 49, 54, 56, 62, 66, 67, 68, 69, 72, 73, 78, 83, 87, 90, 91, 96, 116, 119, 150, 153, 166

I'll leave you to assess how this marries up with your own findings, though I'd be tempted to say it may argue for more emphasis on the opening than you seem to allow for (e.g. an invasion may occur on move 55 but the pro may tell us it was prepared by means of move 25).

Apart from sheer number of data points, the GW approach covers many games, many styles, many eras. While it doesn't have the bling of AI graphs, it does offer explanations of each data point in pro words we can all understand. Furthermore, it covers many more concepts than you mention (over 100, all fully explained in the index).

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #18 Posted: Thu Mar 17, 2022 4:15 am 
Lives in sente

Posts: 967
Liked others: 24
Was liked: 173
Rank: panda 5 dan
IGS: kvasir
Knotwilg wrote:
I think my methodology has been fairly well documented so I won't repeat it here. If you don't like the results of my exercise, there's little I can do about it. I was not biased towards the middle game but the data pointed me there.


I looked through some of your 2021 problems and there is a number of them that could be directly out of opening problem books. That is my opinion, I have some opening problem books. The same thing for endgame, there are some yose problems and some cases of things going haywire during the endgame. We can of course disagree about what is opening and endgame (it is just too bad if we talk in cross).

I tried going through some of my games from last year in similar fashion as you at one point but it takes a lot of time and I was not happy with the problems I came up with. I only came up with two of them and showed to some friends.

One of the difficulty is that there often isn't a clear best move, it is just that the move played was bad. I don't know how to make that into a problem. Also what to do with cases when the move played is not objectively bad but invites trouble that is beyond what I can solve but there is an alternative that is solid and trouble free? My two problems were of this type.

One way would be to limit it to only very clear cut cases, I guess these are mostly going to occur in very hairy positions were you have to get the one-and-only move. Maybe you can explain your criteria sometime. Maybe you already did, it is hard to read absolutely everything on here.


Knotwilg wrote:
I'd like to see other exercises in this fashion, i.e. more data research on the impact of good or bad play in certain stages. I don't remember who it was - you? - but someone pointed out that frequent analyses showed that the point gap is usually larger than the sum of individual point gaps, so there's an invisible compound effect going on. That's the best counterargument to my methodology up to date.


Yeah, that was me. This kind of research is difficult, one also ends up looking a lot at ones own games. I do agree it can be explored deeper.


This post by kvasir was liked by: Elom0
Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #19 Posted: Thu Mar 17, 2022 4:41 am 
Lives with ko

Posts: 148
Liked others: 134
Was liked: 23
Kirby wrote:

Going "overboard in investing in thickness" is a "risk-based strategy", because it means you are playing too slowly, thereby increasing the chances of losing the game. If it's an issue of komi, strategy should adapt to play well according to the rules of the game. On the flip side, if a particular strategy increases the overall win rate of a given pro, there is evidence that that strategy was less risk-based. Professionals are in the business of consistently winning games, and a strategy that is consistent with that is not a risky one.


I'm not at all sure sure I'm reading this right - but professionals aren't completely rational economic operators in the 'business of winning games' go market. There are fashions and trends and diversity of styles. Especially before AI, new strategies were experimented with, popular for a certain time period and widely adopted and discarded as pros move onto the next trend or explore their own interests. Different strategies suit different players and capabilities. Over time the reputation of the effectiveness of a strategy can jump around like the value of stocks and shares, and risk isn't always neatly "priced in".

Top
 Profile  
 
Offline
 Post subject: Re: The new standard of online play
Post #20 Posted: Wed Mar 23, 2022 2:11 pm 
Beginner

Posts: 1
Liked others: 0
Was liked: 0
OGS: grtna
John Fairbairn wrote:
For amateurs, who haven't got the time to build that level of intuition, words are close to essential, as is a tailored course of study. One I would suggest is that, before you embark on AI study, you study Chinese master games of the past. They were not burdened with the deceptive word sente, and instead paid very much to the initiative. The difference with old Japanese games is stark.

Then, to bring this closer to the modern age, this could be followed up by studying what I like to call ley lines. Just as Stone Age man was apparently able to see complex patterns at ground level that we need satellites to see, AI operates with ley lines that connect all the significant parts of the board together. One reason I see value in studying old Chinese games is that they were masters of reading ley lines - they had to be, because group tax meant they had to worry all the time about keeping connections intact, even over great distances. Necessity was the mother of invention. They invented their own relevant terms, too. One was zhaoying (call & response), but they also emphasised "crowding" or pressurising, as opposed to attacking, which kept the INITIATIVE (and so hindered opposing connections, while promoting one's own).

Knotwilg has usefully emphasised several times his realisation that slow connections are bad. As far as I can recall, the main lesson he has taken from that is simply to try to avoid them. A good start. But potentially passive? Maybe a better appreciation of the initiative (and not sente), as per AI play, would teach us not to be so passive. Perhaps we should be thinking more about FAST connections (e.g. via call & response), and connections that lead somewhere, even if we don't understand quite where - like ley lines!


What would be a good starting point for studying old Chinese master games with good commentaries? I'm looking at your book Evergreen Go Records (1682), but I wanted to know if there are other good choices.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 26 posts ]  Go to page 1, 2  Next

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group