First batch has been released:

https://deepmind.com/research/alphago/a ... lay-games/

_________________
“The only difference between me and a madman is that I’m not mad.” — Salvador Dali ★ Play a slooooow correspondence game with me on OGS?

The SGF archive is also hosted at Go Eye's web site (And in Go Eye's live broadcast list) http://goeye-app.com/go-resources/ag.tar.gz

Thanks.

But I am only able to play throught game 1. When I try to choose another, a new screen appears with game 1. {shrug}

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Clicking inside the Choose Game text-box activates the dropdown, and the different games populate just fine for me. (using Chrome on a Mac)

Thanks. Maybe I should try Chrome.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

For those who just want direct SGF downloads (and can't open .tar.gz's from the link above), here they are!

_________________
The road to wisdom? Well, it's plain, and simple to express: Err, and err, and err again; but less, and less, and less!

These games seem so strange.

It's Go Seigen and Kitani Minoru all over again! These games look experimental. I wonder if they don't come a little too early. Studying the five games of this tournament should take priority since they involve human players and might be more useful in the short run. One of the pros said he found many new motives. Let's find those motives and write them up for other pros (amateurs?) to use. Fan Hui could also be helpful (as after the Lee Sedol match) in supplying variations and winning percentages.

_________________
Couch Potato - I'm just watchin'!

10 more games released and black only won 3 of them. Maybe we'll see change in komi soon, as many pro already feel the same way as AlphaGo

For some reason, DeepMind seems to have gotten bored with the release schedule and they just posted all 50 games a few hours after posting 11-20. I've scraped them all and uploaded them here: https://s3-us-west-1.amazonaws.com/apet ... lfplay.zip and as usual, here they are (11-50) below as well!

_________________
The road to wisdom? Well, it's plain, and simple to express: Err, and err, and err again; but less, and less, and less!

Amazing, thanks.

Wild looking games. In the first game listed, white attaches or hanes 5 times in the first 10 of its moves.

A later game involves a whole board semeai where the game continues after a 21 stone group is captured.

Also, how do we uniquely refer to these games?

Google numbers them G1 up to G50.

why did they release it early?

they stated multiple times that it would be only ten a day?!

Wouahhh already all games are available . Now I have an addiction. I want to see more games. Let's sign a petition hahaha

they release all hundreds of millions that have been self played.
in compressed format it wouldn't be unmanageable in size. no bigger than say the full bitcoin block chain which as of now is 250 GB

I wonder why these games feel so much more "bot-like" than the mostly-human feeling of the moves AlphaGo played against Ke Jie and Lee Sedol. The way that the players jump from area to area in the first game, for example, is reminiscent of the MCTS bots of yore.

Without commentary, I can't seem to find any interest in these games. I'm probably too weak.

I feel sure that we are going to get some pro commentary before long.

Looking at these games, I can see why Fan Hui talked about AlphaGo's philosophy being different from that of humans. And that is true of the approach strictly in terms of (not well defined) probability of winning instead of territory (points). But human pros do not think strictly in terms of territory, either. As Takagawa said, you make territory through fighting. (You can make territory in a plodding fashion, but then you lose.)

As for jumping around, that may be explained in human terms. In general, adding stones to a region reduces the local temperature (how many points a play gains there). So in general it makes sense to switch to a different region after a few plays. (OC, there are many exceptions to that. ) Around 20 years ago, in response to a request by John Fairbairn made to the go community at large, I coined the "proverb", "Tenuki is always an option." As the AlphaGo self play games show, it is, isn't it?

Eary in the last century Kitani and Go Seigen ushered in the New Fuseki, which stressed center play. Perhaps AlphaGo will usher in a kind of New New Fuseki, or Ultra-Modern Fuseki. Stay tuned.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Any chance we can put the large number of embedded sgfs behind hide tags? This page stops rendering halfway through on my iPhone, and I have a hunch that's the reason. It's slow but functional on my laptop.

_________________
Occupy Babel!

In his commentary on AlphaGo-Ke Jie game 3, Ryan Li said the one proverb that AlphaGo consistently follows is, "If there is no local play that will allow you to gain an advantage, tenuki." (The wording may not be exactly correct - I can't find the line in his video right now.) Some of these games show that in the extreme.

Yes, as I discovered a few years ago, there is at least one actual proverb along those lines. One I can loosely translate as, "If you don't know where to play (locally), play somewhere else."

OC, AlphaGo does not think locally. That may be a potential theoretical disadavantage, but you can't prove that by its games.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.