How to Beat KataGo at 3 stones

 This is playing against the version of KataGo that one can download, on a desktop computer, with the initial settings, meaning it's a much weaker version than those running on network computers. However, it's still very strong, and figuring out a way to beat it was quite challenging. The main strategy is to take advantage of certain corner situations which come up that it handles oddly.


The odd moves of 2 and 4 are to set up situations where the W corner can be killed later. To get to these points, one must tenuki when W plays 3. How to handle the Upper Left (UL), given one has handed the advantage over to the computer, is important. One way to do so is shown.


B has made a strong shape, and now it is expected that W will invade in the UR and LL corners.


Here W has invaded the corners, and left them, playing 29. B will set up aji to help the UL group, and start the process of killing the corners.


When B plays 34, W will typically answer at 35, but only if 2 is in place. Without 2, it would answer by playing under 20, and the corner could not be killed. This is why moves 2 and 4 were played as soon as possible, so set up the situation where W will leave the corners to die.

In the above, note that 40 is threatening the UR corner. Also note that B has set up aji to help the UL group, so that if attacked, the 3 stones (3,5,7) can be captured.


Note that W ignored the threat against the UR to play 43. B responded by threatening the LL group. In retrospect, killing the UR group immediately may have been simpler. Let's see what happens.


W unexpectedly play 45. When B answered by 46, and W played 47, B answered with the odd 48. The reason for this is there is still an opportunity to kill the W corner. If the move were played as an atari at P19, W would answer at R19, and the corner would be alive, so 48 is played instead, with the hope of still being able to kill the corner.


Here 55 is a mistake. It should be played at S17. Interestingly, if allowed to "think" for much longer than 5 seconds, it still prefers S19 to S17. The group can be killed by playing T18, but there's no hurry, because W doesn't see the group is in danger.



Before killing the UR, B invades the LR. R10 instead of 74 may have been better, but the concern is if one waits too long to kill the corners, eventually W wakes up and defends them.


Of note is the B UR group does not have eyes. However, if one moves out a bit (78 and following), the group easily has more liberties than the W corner.

When W defended at 87, B had the chance to kill the LL corner with 88.


Here the UL group has been secured, and the LL corner killed. Black has a substantial lead here, but, interestingly, KataGo evaluates it as only being 4 points behind (Blue Spot shows -4.9), with a 41% chance of winning. This is because of how it is treating the UR and LL corners.


This is the final screenshot. KataGo typically resigns on move 153 if the evaluation has it with a sufficiently low percentage. The Blue Spot shows B ahead by 51.3 points.

To summarize, the strategy is to set up corner situations where KataGo errs. The trick is that one has to tenuki several times to do so, so one needs to be able to handle situations reasonable well where KataGo has the advantage, due to having tenukied. Discovering the approach used in the UL helped out a lot.



Comments

Post a Comment

Popular posts from this blog

Versailles Unveiled: The Power Dynamics of Louis XVI's Salons

Image
  It was the best of times, it was the worst of times. In the heart of Paris, during the best and worst of times, sat Louis XVI, ensconced in the grandeur of Versailles. The salons of the era, with their opulent design and luxurious materials, reflected an age of excesses. Yet, these spaces were more than mere aesthetic marvels; they were potent symbols of the nobility's power. As revolution's whispers grew louder and social disparity starker, these salons served as bastions of the old order, standing in stark contrast to the tumult outside. They were not just about style; they were a response, a bold declaration to a world on the brink of transformation. Amid societal upheaval, within these gilded walls, the monarchy's grandeur endeavored to assert its relevance.
Read more

The Secret of Happiness

 This post will discuss the secret of happiness, the subject of many books, and studies, but actually very simple, which is to be grateful, appreciative, and thankful. I understand there has been a fair amount of research into this, but I'm not looked into this, but learned this through my study of narcissism. Narcissism (I'm referring to especially in the context of narcissistic personality disorder) is characterized by a number of things, including excessively seeking attention, excessive dreaming of something better (a better job, a better mate, etc.), perceiving oneself as special and that only other special people perceive this, feelings of grandiosity, an exaggerated sense of entitlement, manipulating others, a lack of compassion for others, arrogance, and excessive jealousy.  I was interested in which of these symptoms is the horse, and which the care, and discovered that the horse is the exaggerated sense of entitlement, whereas "arrogance" is essentially a sy...
Read more

How to Beat KataGo at 4 stones - Another example where W starts out with San San

Image
  This example is similar to a previous one in that KataGo starts out by invading the Lower Left (LL) corner, but a different approach is used by B, where moves 2 and 4 are played and then 6, rather than 6 immediately. Here KataGo has invaded in the UL adn LR corners, and played 31. The strategy B is implementing involves 4 items: 1. Kill 2 of W's corners. 2. Make a territory somewhere else besides the 2 killed corners. 3. Don't allow W a huge territory. 4. Keep the B groups safe. The most challenging item is the second one, as KataGo is very good at sniffing out prospective territories and limiting them, especially if there is a weak group to attack, or some problem with the shape. Before 31 was played, there were potential areas to develop in the upper side, right side, and bottom side, although with 13 sticking out, the top side doesn't look very promising.  Once 31 is played, the bottom side looks to be most promising, especially after played 47 (in the next diagram). O...
Read more