Human-Computer Go Challenges

This page lists Go matches between humans and Go-playing programs.

The programs appearing in this list are generally the strongest, or among the strongest, of their time. In many cases the program is one which has just won a competition among programs.

The humans are mostly strong players, many of professional or near-professional rank. Their ranks, where known, are listed.

Thus this page gives some indication of the increasing strength of Go programs.

Warning

Caution is needed in drawing conclusions from the data presented here. In particular, it is possible for a human to observe, and learn to take advantage of, the weaknesses in the play of a program. Someone who persistently plays the same program can improve his results against it greatly; by nine stones or even more. For example, in the table below we see that in 1995, two of three inseis (about amateur 6d) who had played HandTalk only once before, lost to it giving 13-stone handicaps; but in 1998 Jean-loup Gailly, amateur 5-kyu, was able to give it 20 stones and beat it. He was playing the commercially-available version, which may have been weaker than the version that competed for the Ing prize; but it is probably more significant that he had spent many games investigating and learning to exploit its weaknesses.

This warning has been rather less pertinent from 2008, with all leading programs using Monte Carlo techniques. It is more difficult to learn to take advantage of the weaknesses of a program such as MoGo or CrazyStone. The play of these programs has a significant random element, so you can never be sure that "in this position it will make this blunder" and arrange to set up such a position.

List of games, most imminent/recent at top

The Ing Prize

From 1985 through 2000, Ing Foundation (properly The Ing Chang-Ki Wei-Chi Education Foundation) sponsored an annual computer Go tournament, preferring to hold it on the eleventh day of the eleventh month, for numerological reasons favoured by Mr. Ing. This was originally called the World Computer Go Championship, but came to be known as the Ing Cup.

The Ing Foundation arranged generous prizes for these events. At least for the later events, the first-placed program receieved 200,000 Taiwanese dollars, with 40,000 for second place, and 20,000 for third place. (Between 1992 and 2000 the Taiwanese dollar fell in value from 26 to the US dollar, to 31 to the US dollar, so the first prize was worth about US$ 7,000.)

As well as providing this prize money, the Ing Foundation also arranged for the winning program to play a challenege match, of a game against each of a set of inseis. These were trainee professional Go players from Taiwan, typically in their early teens, with strengths corresponding to about amateur 6-dan European. If the program won a majority of its games in this match, it won a further prize. There was one prize (to be won once only, then the next prize in the list would be competed for) at each of a series of handicaps, as listed:

There has been some confusion about the handicaps involved. I believe that the above table is correct, and that some of the other versions which have been published are mistaken. In this table "17 stones" means, as is usual in the West, that Black places the first 17 stones, and then White places a stone. This differs from the usage of the Ing Foundation, which describes this as a 16-move handicap, as 16 stones are placed before the alternation is begun by Black. This terminology may be the cause of some of the confusion.

If you have ideas for other information that could usefully be here, please email the webmaster, Nick Wedd.

www.computer-go.info
Computer Go article on Wikipedia