1998 Ing Computer Goe Cup

Stories

This page gives various information which may be of interest to the press who are covering the event.

Million-dollar prize on offer

The first prize for the computer tournament part of the Ing Computer Goe Cup is NT $200,000 (New Taiwanese dollars). At current rates of exchange, this is about US$ 5,000, or 3,000.

While this may seem an ample prize, it is small compared to what is potentially on offer, thanks to the generosity of the late Mr. Ing. There are much larger prizes on offer to the winner, if it is of a high enough standard to beat human players: see details. The largest prize available is NT $40,000,000, worth over US$ 1,000,000.

The way this works is that the winner of the computer tournament gets to play against three "inseis". These are recently qualified professional Goe players, often children, with a similar playing strength to the strongest European players.

Last year, the winner, HandTalk, was able to beat the inseis while receiving 11 handicap stones. So this year, the winner will get to challenge them receiving only ten handicap stones. If it wins two or three of the three challenge games at that handicap, it wins a prize, and challenges them again receiving only nine handicap stones. This continues until the winning program fails to win a best-of-three match against the inseis. In principle, it can continue until it has won a match with no handicap, for the $1,000,000 prize.

In fact, there is little prospect of a computer program achieving this strength for at least ten years.

The Year of the Team program

From the first Computer Go Championship until this year, all computer Go programs have been essentially the work of individual authors. Most of these authors must have received help and advice with their programs, but have been totally responsible for their programs, and have received the full credit for it. See e.g. the results of the 3rd FOST Cup in 1997.

In 1998 programs which are the joint work of more than one author have appeared for the first time. They have been extremely successful. The results of the 4th FOST Cup in 1998 show team entries in 1st, 2nd, and 5th place. The first placed program is the joint work of four North Koreans; the second placed, of a Russian and two South Koreans; and the fifth placed, of four Chinese authors as described below.

First International Conference on Computers and Games

The Ing Computer Goe Cup was originally scheduled to be on November 14th and 15th. But it was rescheduled to avoid a clash with the First International Conference on Computers and Games (CG'98) in Tsukuba, Japan, on November 11th and 12th.

We have heard that almost half the papers submitted to that conference are about Go, so this would have been really bad timing for the Ing cup.

Unfortunately, none of the participants at this event in Japan will have had time to improve their programs using what they may have learned there.

HandTalk beats 3-dan

Professor Chen's program HandTalk played an exhibition game against a 10 year old 3 Dan girl without a handicap, and won.

HandTalk was the winner of the 1997 Ing Computer Goe Cup. It has now be renamed as Goemate.

Professor Chen writes:

The girl, named Saori Ishihara, is a 1-dan given by Nihon Kiin (Japan Go Association). Her 3-dan was given by a club. More correctly, she is a 1-dan.

My Goemate, a modified version of Handtalk, is now weaker than Handtalk. It played against Handtalk with only 1/3 winning in over 50 recent games before the FOST Cup. The results show that the modification is incomplete and has many shortcomings.

The girl was 5 kyu in 1995-96 and 2 kyu in 1997. Although she is now 1 dan, I see that her level does not appreciately elevate as compared with last year. She lost to Handtalk last year by 6.5 points. That game was shown by Martin Mueller on our Mailing List as a puzzle in May this year and opened on May 31. We can see that the girl was bad in reading life&death. In the game played in this year with Gomate, her weakness in life&death is also shown. She lost the game by 8.5 points.

Generally, a boy or a girl progresses in goe level faster than goe-software does. Sauri Ishihara may be a girl with slow progression. She does not exceed the top level goe programs in these years.

Chen Zhixing

Wulu developed by team

Professor Chen writes:

Ms. Chen Guobao is my daughter. She was the main writer of Wulu before the 1996 Ing Cup. Meanwhile, Miss Lei Xiuyu was a professional goe player and spent some spare time in writing Wulu. She became a full-time programmer from December 1996 and then became the main writer of Wulu. As my daughter is a poor Goe player, she cannot be a good programmer of goe software. Mr. Lu Jinqiang jointed the team in October 1997 and wrote the interface part of Wulu. He is also a good player of goe. Mr. Li Zhihua jointed the team in September 1997. He is a player of goe, even better than Miss Lei and Mr. Lu. But he is bad at programming, so he can only do some auxiliary jobs including the joseki library, etc.

Chen Zhixing

GoeMate and its author

Professor Chen writes:

Goemate is a new generation of the old champion, Handtalk. About one third of its engine is new, and the other two thirds are retained from Handtalk. Most of those remaining two thirds will be modified or rewritten over the next several years. However, the new part is incomplete, so its strength does not exceed that of Handtalk, or is even lower. Its winning ratio versus Handtalk was only 1/3 before the 4th FOST Cup (September 1998), but seems better recently. The main changes from Hamdtalk to Goemate are the management of patterns and enlargement of knowledge. Goemate and Handtalk are both written in assembly. Goemate will become a complex of C++ and assembly in near future.

Chen Zhixing, the programmer of Goemate/Handtalk, born in 1931, was a professor of chemistry in Zhongshan University, Guangzhou, China, and retired in 1991. He founded a company in 1997, named Zhixing Computer Goe Company, to develop goe software and to popularise goe. He is the general manager of the company, but almost only in name. The business of the company is handled by his secretary. He spends most of his time in programming Goemate. In a minor part of programming he was assisted by two assistants in their spare time. He spends a little of his time instructing a team of the company to write another goe program named Wulu.

There is an interesting point to discuss: the techniques and necessity to retain the results of calculation for earlier positions. Similar discussions appeared recently in the Computer Go Mailing List. In Goemate/Handtalk, there are some fundamental data of the last position retained and partially renewed on the change of the position. Other data, such as values of moving at each point, are retained and partially renewed.

A related point is to think during the opponent's thinking. This seems easy to do, but may not be important. Goemate/Handtalk waits without any calculation during the thinking of its opponent.

Chen Zhixing

Deatails of some other entrants

Park-YongGoo writes:

"FunGo" was developed by one person and use C++. FunGo uses some results (may be less than 10%) which was made before. FunGo also think during its opponent's time but little (may be less than 5% ). I don't have my employment and my age is 34.

Park-YongGoo

Michael Reiss writes:

Is your program all the work of one person, or of a team? Programming wise, just me, although a large number (20?) of people have contributed with advice.

What is your age? 35

What qualifications have you? Degree in Physics, PhD in neural networks.

What is your employment? Go programmer.

What language is your program written in? C (not C++)

Does it assess the position anew after each move, or does it retain the results of calculations which it made earlier? It maintains about 30% of previously calculated info from move to move.

Does it think during its opponent's time? No.

Michael Reiss

Arnoud van der Loeff writes:

TurboGo is entirely my work, although many users have helped me to improve TurboGo with many suggestions, comments and criticisms.

I'm 30 years old, and I studied tax law after finishing high school. I'm currently working as an Application Developer for KLM (Royal Dutch Airlines).

My program written in Borland Delphi, the most popular Pascal compiler for Windows in the world.

It retains certain results of previous calculations, if the opponent made a move that TurboGo expected.

No, it doesn't think during its opponent's time. If there is one thing that is not a problem for TurboGo, it's time. TurboGo plays rather fast (hence part of its name), so there is no need to think in its opponent's time.

Arnoud van der Loeff

Martin Müller writes:

Explorer is a loose collaboration between two people: Anders Kierulf has written the core engine called Smart Go Board. I have written the Go-playing program called Explorer which uses Smart Go Board for the low-level functionality.

I am 33.

I have a Ph.D. in Computer Science from ETH Zurich, Dipl.Ing. in technical Mathematics from Graz University of Technology.

Currently, I'm an STA fellow of the Japanese Science and Technology Agency at ETL in Tsukuba, Japan.

Explorer is written in C++.

Information on blocks and liberties is updated incrementally. In addition, it caches information from pattern matching and tactical analysis. This saves about 93% of pattern and 75% of tactical calculations.

It does not think during its opponent's time.

Martin Müller

Frolov Arkadiy writes:

Alpha-1.1 is the work of a team of four. I am the project manager, the others are programmers.

I am 47. My strength at Go is 3 kyu.

Alpha-1.1 is written in Borland C++. It uses results of calculations which it made earlier, but does not think during its opponent's time.

Frolov Arkadiy

Bruno Bouzy writes:

I am the sole author of Indigo. I am 36. I am an assistant in computer science at the university.

Indigo is written in C++. It assesses some data anew after each move but it also retains some results which it made earlier. It does not think during its opponent's time.

Bruno Bouzy

Tristan Cazenave writes:

I am the sole author of Gogol.

My age is 30. I have a Ph.D. in Artificial Intelligence. I am employed as an Associate Professor.

Gogol is written in C++ and First Order Logic. It reevaluates the position anew after each move, and does not think during its opponent's time.

Tristan Cazenave

Yan-Shi, Jim writes:

I am the sole author of "Jimmy". I am 30.

I am a PHD student at National Taiwan University.

"Jimmy" is written in C. It assesses the position anew after each move, and does not think during its opponent's time.

Yan-Shi, Jim

 

 

 

 


This page is for the use of Press covering the Ing Computer Goe Congress.

Last updated: 1998-10-28.


This page is a historical copy of a page originally hosted on the web site of the British Go Association, at its former domain www.britgo.demon.co.uk. Any links that it contains are likely to be broken, and will not be fixed. For current information in the 1998 Ing Cup and other past computer Go tournaments, see www.computer-go.info.