Life in 19x19. Go, Weiqi, Baduk... Thats the life.
https://lifein19x19.com/
I don't: I'm more interested in the content.pookpooi wrote:I'm wondering when did DeepMind record [url=https://www.youtube.com/watch?v=CUZxe7bXKxo&list=TLGGh_Yr91zRhmUxNTA5MjAxNw]
Ke Jie beat Park Junghwan in this event by 2.5. He did his now-traditional 3-3 invasion before move 10.xiayun wrote:The direct translation of the event name would be World Summit Match. It's a bi-annual event between 2 of the top players in China and Korea. Here are their past results:Bohdan wrote:What's the tournament Ke Jie play Park?
2003: Cho Hunhyun beat Chang Hao
2005: Chang Hao and Lee Changho played to a 4-ko tie
2007: Luo Xihe beat Lee Sedol
2009: Lee Sedol beat Gu Li
2011: Choi Cheolhan beat Kong Jie
2013: Chen Yaoye beat Park Junghwan
2015: Kim Jiseok beat Tang Weixing
Ke realised his formation was bad: to be better either move high extension to low so there is no slide (AlphaGo would then shoulder hit the extension, they looked at some variations), or move 3-3 to 4-4. He had thought about reinforcing for his previous move (e.g. big knight shimari to lower side) but that was unsatisfactory as white can still easily sabaki on the right side.
was bad, -3% said AlphaGo. It liked to just tenuki and invade at the top (seems to me a you-destroy-my-thing so I-destroy-your-thing balance, as more stones added to area it gets smaller, also if black invaded there before right side extension and white pincered then AG liked black to tenuki and extend on right side, already made top smaller in sente).
was bad, -2%. It preferred 2 space jump to b to grow lower side (Ke never considered), after peeping ladder on left exchange. Its thin shape made Ke uncomfortable.
was sharp, Ke didn't expect it: even normal answer to the kick was bad for him. Apparently some other humans, maybe Gu Li, had suggested it though.
was bad -2%. AlphaGo preferred to just dodge to c. Fan showed a variation in which AlphaGo ended up capturing the stone on the right ("it got more territory than me in my area!" says Ke), but black squeezed the outside to develop the lower side and got sente to extend on upper right side. Damage limitation.
was bad -3%. AlphaGo says hane underneath. The trade of game was good for white.Brief comment aboutUberdude wrote:Something I found particularly notable in the recent DeepMind AlphaGo vs Ke Jie analysis videos is how critical AlphaGo was of a sequence of 4 moves Ke played in game 1, quickly losing 10% win rate (he'd already lost a little and was black so AG went from about 55% up to 65% which is approaching 70-80% game over level -- it's not really a probability). As throughout the whole video Ke accepted the criticism gracefully (much humbler than a year ago!) and showed a tewari analysis (to 3-3 invasion of 4-4 and double hane) saying that he knew his result sucked. For the best human in the world to be told he keeps making mistakes from bad to worse, rather like when a pro reviews my game, was amusing/sobering.
In summary:
- Ke's high extension on the right side had already come in for some criticism (AlphaGo had preferred a low extension, or invading the top at a earlier).
- When AlphaGo invaded at
-
. makes a base with 
and pincers , preventing it from making a base. For almost two millenia top players have cherished making a base, with some exceptions. Making a base is not just a question of making territory, but one of making a viable group. In addition, induces 
, while if White had made a base for 
with an approach at "a", Black would surely have replied. is thus a triple purpose move, and almost a double sente. How can it be a mistake???? 
It just might be one, though, eh?On the critical side. is inducingBill Spight wrote: It seems to me that AlphaGo brings into question one of the oldest ideas in go, so old that it shows up in the oldest game records we have, the idea of a base.
really something you want to do, as it weakens the black stone above? (which has the white shimari looming in the distance, and as it's big and loose you don't want to play forcing moves from the outside to help sabaki like you would happily against a smaller and more solid shimari). And given white's central support the group doesn't feel so weak once it's jumped out. It's not as bad as the kyu player squeezing through a small gap, but topologically speaking it is rather similar. It certainly goes against the Takemiya or O Meien 'connecting over the top' style. Also you could say ends up too close and flat, because of the nice press shape weakness AlphaGo used. The 3-3 already has at least 1 eye in the corner and still has miai to extend to bottom or right side so is healthy enough.Bill, Agree with you about the importance of bases but I'm not quite sure I see how it's relevant here. The 3-3 stone makes a base on its own - that's its whole point. So 1 is not needed, and is even overconcentrated. Also, I can't see that it's a sensible pincer - the White jump is attacking Black above at least as much as Black is attacking White. And even if you do assert Black is attacking, he shouldn't be - at least if we follow Mizokami Tomochika's method of looking at the half-board: White has more stones than Black on the right, so Black should be playing prudently on this side (e.g. tenuki to the upper side), not attacking. Incidentally, my sense is that Mizokami's method seems to marry quite well with AlphaGo's style.It seems to me that AlphaGo brings into question one of the oldest ideas in go, so old that it shows up in the oldest game records we have, the idea of a base.
Don't think so - at least the Chinese version says W+0.25.Park won by 1.5.