Lee Sedol vs Alpha go

[Cassandra]

For game #2:

Here is what two electronical "seniors" thought about the winning probability of their youthful buddy.

Please be careful while interpreting the diagrams.
I suppose that "EQUAL" is ABOVE 50 %, maybe something about 55 %, at least.

CS2012.jpg (125.41 KiB) Viewed 11658 times

MFGo12.jpg (97.97 KiB) Viewed 11658 times

[RobertJasiek]

Before the match, I suggested that Lee must play for long-term, global interaction. Instead, Lee focuses on the contrary while Alphago opts for long-term and global interaction. I expected top human players being able to surpass Alphago there but Lee seems to be the wrong opponent for the task of beating Alphago. We do not need the top human players but, among those with deep LD reading, players very strong at positional judgement of the dynamic aspects incl. center influence and attacking potential. Back to the 60s and 70s of Japanese top professional play! Lee misjudges aji, influence, potential etc. With his focus on reading and a mostly territorial judgement, he continues to lose. Alphago is a holistic player but Lee is not.

[longshanks]

Apparently Lee said at the post-game press conference “As I was playing today’s game, I kind of felt that things would get difficult after the mid-game, reaching closer to the end-game. So I feel that for the next game I would have to focus quite hard on the beginning part of the game.”

On a related note hasn't Lee also said in the past that the opening is the weakest part of his game? That might just mean he likes to fight though..

[MinjaeKim]

Before the match, I suggested that Lee must play for long-term, global interaction. Instead, Lee focuses on the contrary while Alphago opts for long-term and global interaction. I expected top human players being able to surpass Alphago there but Lee seems to be the wrong opponent for the task of beating Alphago. We do not need the top human players but, among those with deep LD reading, players very strong at positional judgement of the dynamic aspects incl. center influence and attacking potential. Back to the 60s and 70s of Japanese top professional play! Lee misjudges aji, influence, potential etc. With his focus on reading and a mostly territorial judgement, he continues to lose. Alphago is a holistic player but Lee is not.

Please don't try to judge Lee Sedol's go strength. Even the top professional's won't do that.

[Fadedsun]

On a related note hasn't Lee also said in the past that the opening is the weakest part of his game? That might just mean he likes to fight though..

He has said that and they discussed that in the commentary during the stream. It's as you said, he believes his opening is weak because he likes to get into fights too early.

[Cassandra]

Please don't try to judge Lee Sedol's go strength. Even the top professional's won't do that.

I am very sure that the evaluation of the skill level of any of AlphaGo's opponents must be done with extreme caution.

In my opinion, AlphaGo has a very "adaptive" style (if is has any).

Didn't Michael Redmond mention that AlphaGo did not let Lee Sedol do what he wanted to do ?
Replace "Lee Sedol" with any other name, and Michael's impression will remain true.

+ + + + +

Another evidence might be the also "highlighted" fact that there were no "grave" ko fights, neither in AlphaGo's games against Fan Hui, nor against Lee Sedol.

Let us assume for a moment that "complex ko fights" would be really one of AlphaGo's Achilles' heels (we have not seen any, so cannot be sure at all).

Would it be so unlikely that AlphaGo was aware of this "weakness" ?
And would want to avoid as much as possible to become forced into such a fight ?

[Charlie]

The next match is on Saturday, yes? I truly hope so because these 04h30 (Germany) starts (leaving time for tea-making and breakfast-finding) are quite tiring - a rest day would be nice.

EDIT: I truly hope that Lee comes back, after the break, and wins on Saturday and Sunday. I really want this match to stretch to five games - I haven't had this much fun on YouTube since YouTube was a Thing. Also, I can't help but feel that Google have had their victory and don't need the straight-sets win. They can leave some honour and mystique for the humans!

[Charles Matthews]

Please don't try to judge Lee Sedol's go strength. Even the top professional's won't do that.

I agree. But it probably makes some sense to see "Korean go" as on trial here. AlphaGo is playing "new kid on the block", and the pre-match comment was rather dismissive of whether pros would have anything to learn from it. Recall the comments "doesn't know" this or that.

[Uberdude]

EDIT: I truly hope that Lee comes back, after the break, and wins on Saturday and Sunday. I really want this match to stretch to five games

Even if Lee loses games 3 and 4 there will be 5 games. It's a 5 game match, not best of 5.

[RobertJasiek]

Please don't try to judge Lee Sedol's go strength. Even the top professional's won't do that.

It has not been my intention to judge Lee's absolute go strength. Instead, I have commented on his strengths and weaknesses. (Everybody is doing so.) Among those few top professionals I have heard commenting on game 2, Lee Sedol is the only one, whose judgement on who has been ahead (at no time Lee) I consider correct. Judgement of the players' play in this match is at the core of what is happening.

[RobertJasiek]

After game 3, my advice to Lee Sedol is the same as before game 1: instead of fighting locally, he must, from the beginning, go for global, long-term interaction including good positional judgement. (Move 77 was the first such move, but at this moment it was too late; Alphago already was far ahead.) AlphaGo demonstrates that go is a global, long-term game and that winning requires such considerations throughout the game.

The final ko fights must be evaluated whether AlphaGo lost points for better winning probability. If so, it would not comply to standard human strategy (defending the lower territory unconditionally) but nevertheless maintain the win and so not be a mistake from the program's POV.

Cho Hyeyeon several times mentioned a strategy of a White win by 1.5 points (and her English is good enough not to confuse it with 1/2 when saying "1 point and a half") but she has not been aware of the implication of Chinese rules that White could not win by 1.5 points (but only by 0.5 or 2.5 since the number of neutral seki intersections was 0). During that live stream, it was mentioned that superko was being used, but, ALA we do not hear about special tournament rules specifying the long cycle ko rules, the Chinese Rules are more complicated than superko. As interesting Cho Hyeyeon's other comments have been - the rules understanding appears to have been wrong.

[trout]

3rd game;

AlphaGo defeated Lee Sedol by resign.

AlphaGo vs Lee Sedol 3:0

P.S. I cannot seem to get in page 1 or 2.

[Charlie]

After watching game three, I have two main thoughts.

Firstly, Redmond is pretty brazen, saying plainly that AlphaGo "Beat Lee Se-dol at his own game" and following it up by drawing a line from Dosaku to Seigen to AlphaGo. (He did that during the game and at the post-game conference.)

Secondly, I really wish we got to see more live pro. games with English commentary. Watching these games has been so much fun, I wish it happened all the time! In fact, I think it would be even more thrilling if both players were emotional, mortal humans to whom we could relate in some way.

[Cassandra]

For game #3:

Here is what two electronical "seniors" thought about the winning probability of their youthful buddy.

Please be careful while interpreting the diagrams.
I suppose that "EQUAL" is ABOVE 50 %, maybe something about 55 %, at least.

CS2012.jpg (140.11 KiB) Viewed 10788 times

MFGo12.jpg (95.86 KiB) Viewed 10788 times

[mika]

Time setting for these match was announced.

It will be set to 2 hours and 1 minute x 3.

This is what I read everywhere, except in Wired's coverage of the matches, where they say stuff like

One more failure and his time for each move would drop to 30 seconds.

Is there some special rule about overtime used that's not mentioned or are the Wired writers totally oblivious to how overtime works in baduk (which would seem rather odd)?