BGJ 128 Autumn 2002
Reviewer: Nick Wedd
In issue 109 of this Journal, I wrote a review of version 10 of David Fotland’s program The Many Faces of Go. Since then, I have been recommending it as the best all-round Go program for someone who is willing to pay for the best. Version 11 has now been released. Like version 10, it is a program for Windows, and comes on a CD.
Many Faces of Go is primarily a Go-playing program – it is one of the world’s strongest such programs. In addition, it comes with a large amount of other useful material, and many Go players will value it for this, rather than for its playing ability.
Many Faces of Go version 11 (MFoG11) can now play on any sized board from 7x7 to 19x19. It can play against a human, as Black or as White; it can play against itself; or it can act as a board for a game between two humans. It can be set to use American, Chinese, Japanese or SST (Ing) rules. It plays as fast, or as slowly, as you tell it to. Unlike many computer Go programs, it adapts its speed of play to use the time available to it – and unlike some, it plays better if given more time. It can play very fast. I set it to play itself on a full board at its lowest strength level, and the entire game took 10 seconds.
I tried playing against it at the highest of its ten levels of strength, giving it nine stones, and beat it (I am about 1-kyu). This is something which I still cannot do against HandTalk. The reason for this difference is, I think, that Many Faces plays ‘honestly’, whereas HandTalk tries sequences which, if played by a human, would be described as swindles.
I also tried playing it against a recent version of Mick Reiss’s program Go4++. The game was close, but the styles of the two programs are very different. Many Faces plays rather like a human 10-kyu, making influence and then not knowing quite what to do with it. Go4++ behaves as if it knows its own limitations: it contents itself with modest third-line territory, and relies on the fact that its opponent won’t know how to make the most of its outward influence.
Commercially available programs such as MFoG11 are ‘frozen snapshots’ of their programmers’ work, as this is undergoing continuous improvement (or rarely, relapse). MFoG11 was frozen shortly before the recent 21st Century Cup, which it won, defeating all its opponents, including Go4++. It can therefore claim to be the strongest computer Go program that you can buy. I doubt that it is measurably stronger than the latest version of Go4++. But it should be stronger than the currently available version of Go Professional, which is a version of Go4++ frozen about two years ago.
As well as playing Go, MFoG11 includes many features not present in other Go- playing programs. There is an introduction to Go, starting at a very elementary level, explaining liberties, connections, capture, and so on. This is displayed more clearly on a computer screen than it could be in a book. There is plenty of advice on play, and two sample games with very full comments directed at beginners.
There is a Joseki tutor, and two large databases of openings which can be used for studying fuseki and joseki. It can analyse the status of groups. It includes a problem- solver, which is much stronger than the one that was included with MFoG10.
You can use it to record and replay games, and to add variations and comments to the game record. Something which I find particularly useful is that it can record games in either Ishi or SGF format, and can read both formats. It can therefore be used to convert between them. It comes with two large collections of game records: 1200 commented amateur games, from the Go Teaching Ladder, and over 1600 commented professional games.
It also comes with a collection of over 2000 Go problems. It keeps a record of your performance at these, so you always have an incentive to try to get them right and improve your score. When you ask it to set you a problem, it randomises the orientation, so you get rather better value than if they were always presented the same way round. If you solve a problem correctly, it plays the next move to check that you can follow up you solution correctly; and if you give a wrong answer, it shows you a refutation. It has the facility to update its collection of problems directly from some place on the internet – I don’t know how often this place is likely to be updated.
A new feature is that MFoG11 can be used as a ‘Go Client’, to connect to IGS, NNGS, and other compatible Go Servers. It is not the best such client, but it does work, and you may prefer to use it as your Go Client rather than go to the trouble of downloading and familiarising yourself with a different program. If you do use it as a client, you can use it to cheat, by asking its opinion of the status of your groups, the balance of territory, etc., as you play.
You can use MFoG to play go using your modem with a direct telephone connection to anyone who owns any version of MFoG, or Nemesis or certain other programs that support the same protocol (the Go Modem Protocol).
MFoG11 has a feature which I know of in no other program, and which I particularly like. This is a ‘game score graph’. For any game, you can ask it to display a graph of how much White is winning by after each move. In general, this rises when White moves and falls when Black moves; and it rises and falls by more than usual for what MFoG considers to be particularly effective moves.
Many Faces version 11 has in one package almost all of the features available in any other Go program. It is suitable as an introduction to Go for complete beginners, and has plenty to offer to experienced players. I would not hesitate to recommend it to anyone who has a 32-bit Windows system (Windows 95 or better). It is attractively packaged, and would make a suitable present for a games player.