2005.11.02 - 2016.12.31
TYGEM "9D vs 9D" dataset (1,516,031 games).

2003.09.25 - 2011.12.28
TOM "9D vs 9D" dataset (50,956 games).

1940.01.01 - 2017.01.09
Professional dataset(73,522 games).

2015.10.05 - 2017.06.14
AI dataset (5,449 games).

https://github.com/yenw/computer-go-dataset

This should be a goldmine for training neural networks.

The Tygem stuff set unfortunately not in some readable (by Drago, Sabaki etc.) format, probably there's some heavy grep/sed like action necessary in order to make them an sgf collection Beyond my abilities.

How can someone copyright games played by someone else? That doesn't make any sense.

The general theory is that you can copyright a large collection of public domain stuff, even if individual items aren't copyrightable, because there was a creative act in choosing what games to include.

Updated......

Update AI dataset

AlphaGo
..AlphaGo Ke
....AlphaGo vs Ke Jie
......3 games
......2017.05.23 - 2017.05.27
......3W / 0L / 100%
....AlphaGo vs China Team
......1 games
......2017.05.26
......1W / 0L / 100%
....AlphaGo + Lian Xiao vs AlphaGo + Gu Li
......1 games
......2017.05.26
....AlphaGo selfplay
......50 games
..AlphaGo Master
....Master
......60 games
......2016.12.29 - 2017.01.04
......60W / 0L / 100%
..AlphaGo Lee
....AlphaGo vs Lee Sedol
......5 games
......2016.03.09 - 2016.03.15
......4W / 1L / 80%
....AlphaGo selfplay
......3 games
..AlphaGo Fan
....AlphaGo vs Fan Hui
......5 games
......2015.10.05 - 2015.10.09
......5W / 0L / 100%

FineArt
..FineArt
....590 games
....2016.11.01 - 2017.05.12
....454W / 136L / 76.94%
..FineArt UEC
....11 games
....11W / 0L / 100%
..Li Long
....477 games
....2016.10.21 - 2017.06.05
....416W / 61L / 87.21%
..Xing Tian
....45 games
....2016.12.27 - 2017.01.02
....38W / 7L / 84.44%
..Tian Xia Wu Gou
....204 games
....2016.09.28 - 2016.10.31
....159W / 45L / 77.94%
..Sweeper
....55 games
....2016.09.02 - 2016.10.05
....38W / 17L / 69.09%
..Tiger
....63 games
....2016.07.28 - 2016.10.28
....42W / 21L / 66.66%

Zen
..DeepZenGo
....Fox Go Server
....97 games
....2016.11.19 - 2017.06.11
....80W / 17L / 82.47%
..DeepZen(P)
....TYGEM
....2422 games
....2016.12.29 - 2017.05.18
....2007W / 415L / 82.87%
..Zen19L vs Godmoves
....KGS
....3 games
....2016.11.29 - 2016.12.01
....0W / 3L / 0%

Leela
..Leela(P)
....TYGEM
....1354 games
....2017.03.29 - 2017.06.14
....559W / 795L / 41.29%

Are you sure that this is Leela software and not some pro?

And how can I open professional games, they seem to be in txt format?

Thanks!

There seems to be no easy way to download the 1.500.000 games, unzip them and plug them into MasterGo or any other database or is there?

No one can play Go all day without rest, so I guess "Leela(P)" is a robot. ^_^

Doesn't (P) reserved for professional only?

Do you think DeepZen(P) is a professional player?

No.
This is why I asked if someone knows what is criteria for (P)

update: The 1st World AI Go Open 2017. (40 games)

Has this been addressed?

_________________
Occupy Babel!

You can delete GoGoD Winter 2016 games, and then re-collect them in another way, since the game itself has no copyright!

October 2017 Monthly Update. (FineArt, Li Long, CGI etc.)

Update. (fuheyuqi, dolbrarm9, CGI)