It is currently Fri Jan 22, 2021 2:40 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 1 post ] 
Author Message
Offline
 Post subject: Endgame insight from Ichiriki
Post #1 Posted: Tue May 26, 2020 2:38 am 
Oza

Posts: 2817
Liked others: 16
Was liked: 3929
I recently published The First Teenage Meijin (see Amazon), which was obviously about rising star Shibano Toramaru but I was mainly stimulated into doing it by my interest in go journalism. In particular, I was becoming intrigued by the extent to which AI analysis was being introduced into professional commentaries. I have little interest in knowing whether move A is X decimal points better than move B, but I am interested in how pros (who surely are interested in winning ratios) alter their thinking based on what they believe AI is telling them.

Or don't alter their thinking! There was a particularly good example of that, apparently, in the latest Judan game.



In the position above, where Black has captured 5 stones and White has captured 4, White (the Judan, Murakawa Daisuke) played 170 at the triangled point.

The game had already become a very close boundary-play (yose) battle as early as move 135. Challenger Shibano Toramaru now played the hanetsugi starting at A. Shibano has proven himself very well attuned to AI play (see FTM), but commentator Ichiriki Ryo made the comment that this was a "human-looking boundary play". AI indicated that either B or C was better, and that Shibano's move lost about 1 point according to the bot (so, we can perhaps infer a point-counting bot is preferred by Ichiriki at least, but he didn't say which one was being used), and the game would then be extremely close.

The commentary does say Black A added some fukumi nuances to the position (see different thread somewhere on this), but to explain his remark about this difference between human thinking and AI knowing, Ichiriki specifically said Shibano was worried about the bad aji in the corner at D. He gives an example where Black plays B, White D and then White uses that aji stone to seal off a nice territory on the right side. I am pretty sure, though, that Shibano saw many more moves than that. All top pros must suffer from the curse of seeing too many moves. In short, as a human with too much imagination and limited reading range, he chose to play safe. Actually, I suspect there is a little more to it than that: he is not just eliminating aji (negative); he is being positive at the same time because the hanetsugi adds fukumi for him (i.e. he ends in sente because White doesn't defend there after Black connects, but turns to B, and so Black gets possible follow-ups on the right).

A somewhat similar consideration arose later in the same area, proving this area was indeed something of a hotspot.



Shibano played 201 at the triangle although A was technically bigger. But he got caught out by a tesuji that reduced his territory - I infer he saw it but misevaluated it, because Ichiriki added the comment "In difficult endgames there are no easy conclusions."

The game went on for 310 moves, and a further interesting comment from Ichiriki was that in the course of the boundary plays from 207 on (the "small yose", move 207 being the last "big point"), if they were played out by amateurs, the lead would change hands two or three times. Ichiriki thought at that stage White was ahead by 0.5 or 1.5 points, and being a pro White maintained that lead. Meanwhile, Shibano, being a human, lost an extra point and lost the game by 2.5 points.


This post by John Fairbairn was liked by: hyperpape
Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 1 post ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Cyrion 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