[computer-go] A chess programmer at the Go-Olympiad

Dear Go programmers

Besides playing with me FPGA-chess programm Brutus at the WC I was operating together with Stefan Mertin GoAhead at the Computer-Olympiad. The Go-Olympiad was definitely more fun. Not only because GA became 2nd in 19x19 and Brutus "only" 4th in chess. The computer chess atmosphere is - at least under the top 4 programmes - extremly compitetive. GA did not support the Modem-Protocoll. I aggreed with all other operators, that the internal clock of each programm is the valid time. This takes away much stress, but it would be completly impossible in a chess tournament. Everybody would suspect that the other side cheats on his clock. In the game against NeuroGo GA crashed. I did not know how to setup the position again. Markus Enzesberger gave me the position on a floppy, but the computer provided by Siemens had no floppy drive. The memory stick did not work with Markus Linux. I started to enter the 200 moves made so far by hand. Having only little Go-experience, this was a hard job. Martin Mueller saw my problems and took the job over (enpassant he commented the game and especially NeuroGO´s somewhat unfortunate moves). Many thanks to Martin for saving the game.

One of the strangest experience was the game against Jimmy. Chess programs react immediatly, if they are in the book (or have guessed the move correctly). At first I found the immediate Jimmy-reply natural. After all there should be also a Go book. But after 30 moves I got suspicious and found the real reason. Jimmy (or its autor) does not know what to do with time. This is for a chess programmer a shocking experience.

Like every chess programmer I am a believer in search and I do not really understand why Go programms do not use it (Martin gave me some good reasons, but the believe is too deep that one can be converted while drinking one beer). I was therefore deeply impressed by Aya. It seems to be one of the first real searchers. GA had not much trouble with Aya on 19x19, but it was helpless against the 7 Ply search in 9x9. The same holds with NeuroGO. On 19x19 it was a clear-cut win of GA. In 19x19 NeuroGo searches 1 Ply. In 9x9 the 3 Ply searching NeuroGo was clearly better than GA.

The node count of Go-programs is unbelievable slow. I also thought that the number 620 on Ayas screen means 620 Kilo-Nodes. But Hiroshi explained me, its nodes. My impression from looking at the Gnu-Go and the GA code is. It is not only the complexity of the game. Go programmers obviously do not spend months to save a few nano-seconds in their time critical parts (maybe because there is no time critical part). E.g. GA is written in Basic, it also does not use the time of the opponent for its own calculations. I do not know of any attempts to parallelize a Go-Programm. It seems to be complicated enough to get everything right.

The 50 Liter beer bet:

There are some plans that I write a 9x9 FPGA Go programm. This is yet not fixed, it depends on the result of the negotiations with a potential investor (I can not finance such a project from my own pocket. Especially I would not be able to produce a few thousand Go-cards).

In case I get the project money, I have made with Martin Müller and Markus Enzesberger the following bet:

I can write a 1-Dan 9x9 Go programm within 1 year.

There is a similar bet with Ingo Althoeffer. I suggest that Ingo organizes a tournament with 11 1-Dan players.

I will inform everybody when the contract is signed - or when it has become clear, that the project does not start due to lack of money. In case of signing the contract I will accept additional offers to join the bet (as long as the deposit is reasonable).

If you think this bet is a very save way to get a few bottle of beer, wine, whisky.. etc. Keep in mind, that this programm will be able to search in the Mega-Positions/Second range. I do not know if this is enough for playing at 1-Dan (I do not even know, what it really means to have 1-Dan). But maybe Hiroshi could make a guess how strong an 1000 times faster Aya would be.

Chrilly Donninger