Fifty-ninth KGS Computer Go Tournament

Sunday May 2th 2010

These results also appear on an official KGS page which links to the records of all the games.


format18-round Swiss
board size9×9
time4 minutes plus 25/20s


The first round started at 16:00 UTC.

Result table


The numbers in this table may not add up
as you might expect. This is because some
entrants played no games, and games
against absent opponents treated as wins.


Eighteen players registered. We welcomed three players new to these events: coldmilk, by ChengWei Chou and ShiJim Yen, cyVersion1, by Chun yi Chen, and MCark, by Nobuo Araki. They all use Monte-Carlo methods, and are all written in C++.

break registered, but failed to appear. It had been intending to play as 'break9', but I failed to understand this from its registration email, and registered it as 'break'. Break is in fact an account belonging to a human user of KGS, so the program of that name was unfortunately unable to in the tournament.

I would like to avoid mistakes like this in future, so I have this advice for all entrants. When you register a bot for one of these events, you should receive a reply from me, saying something like "Thank you for entering. xyzbot is now registered". Please read it. If the name 'xyzbot' is not the name of the bot which you intended to register, something has gone wrong. Maybe I didn't read your registration email carfully, maybe I mistyped the name in my reply, maybe you made a mistake. In any case, you should reply, and try to get the problem fixed. This will of course be easier if you don't leave it until the day of the tournament to register, as seven people did yesterday.

Aloril entered his three bots IdiotBot, SimpleBot, and WeakBot50k. WeakBot missed rounds 1 and 2. IdiotBot missed rounds 5, 6, 10, 13, 14, 15 and 17. SimpleBot did not join any of its games.

I could have removed break and SimpleBot from the tournament, once it was clear that they weren't playing. I chose not to, reasoning that it would give an unfair advantage to those opponents which had played them in the early rounds; and that, in an event with 18 fast rounds, have a missing opponent for a few of the rounds is not particularly annoying.

In round 1, coldmilk and ManyFaces1 were both late for their game SGF. Coldmilk was black, and joined the game after losing one and a half minutes; but ManyFaces1, which had had a problem with its script, was even later, and had only 30 seconds left when it made its first move. With only enough time for 24,000 playouts per move, ManyFaces1 played poorly, and lost.

In round 2, kurtBot and ManyFaces1 disagreed about the status of kurtBot's groups (which were in fact all dead) SGF. After ManyFaces1 played one stone in the resumption, kurtBot resigned.

In round 5, Zen9 beat Fuego SGF to become the only undefeated player.

In round 7, Manyfaces beat Zen9 in an exciting fighting game SGF. This was the first of Zen9's two losses.

Fuego vs stv
Position after move 57.

Also in round 7, Fuego and stv reached the position shown to the right SGF. I assume it did this with the intention of saving its group. But in fact the black group has no way to live. Play continues, with Black making some odd moves, until White's eye at b8 became solid and obvious, when Black resigned.

kiseki vs MCark
Position after move 82.

In round 8, kiseki and MCark reached the position shown to the left SGF. White already has a won game. Play continued with Black at a, White captures at b. Now Black tried to recapture, which was forbidden by the superko rule, but it did not understand this and timed out.

cyVersion1 vs WeakBot50k
Position after consecutive passes.

In round 10, cyVersion1 and WeakBot50k both passsed in the position shown to the right SGF. They then disagreed about the status of the dead groups, triangled. But rather than a resumption starting, the game somehow stuck with nothing happening. I observed that White should have won, and assigned the win to White. But WeakBot50k failed to leave the game, even after the next round had started. So I used by admin power to "kill" the game and free up WeakBot50k.

Consequently, WeakBot50k was a minute late for its round 11 game with CzechBot. But WeakBot50k plays fast, so I trust it was not inconvenienced.

MCark vs Fuego
After the end of the resumption.

In round 12, MCark and Fuego both passed, with Black 9½ points ahead SGF. They then disagreed on the status of some dead stones, and resumed play. The resumption ended with one dead black stone still on the board, as shown to the left, and Black just half a point ahead. Thus Fuego achieved its MC objective, of winning by the minumum possible margin.

Zen9 vs stv
After White's resignation.

In round 14, Zen9 lost its second game of the tournament, to stv SGF. The game was interesting, but I am certainly not qualified to comment on it. Even in the position in which White resigned, shown to the right, it is not easy to be sure of the status of the groups.

Zen9 vs ManyFaces1
Move 65.

In round 15, Zen9 and ManyFaces1 played a game which was all fighting SGF. Black played move 65 as shown to the left. A move one point further left looks better, though not good enough to achieve anything.

Zen9 vs stv
Move 19.

In round 18, Zen9 and stv played their third game SGF. If stv had won this game it would have achieved 15 wins, the same as Zen9, though Zen9 would still have won the tournament on the SOS tie-breaker. Black played move 19 as shown to the right. It looks to me as if this stone will be swallowed up and killed; and this is indeed what happened to it. But I may be wrong.

Comment – Assignment of Colours

As there were many rounds, and as the KGS tournament draw program thinks (rightly, in my view) that matching opponents of the same strength is more important than avoiding two opponents playing each other more than once, this tournament had many repeat pairings. For example, Zen9 and stv were paired three times, in rounds 6, 14 and 18. In all three of these games, Zen9 played as white. In fact, over all 18 games, Zen9 played only three as black. Several players commented on this oddity of the draw program. They said that when two opponents are matched for the second time, it should be with reversed colours. I agree.

I hope that the pairing algorithm can be changed, in this minor way, and I shall be writing to KGS administrator 'wms' about it. I assume that colours are at present allocated randomly. It is usual practice to try to arrange that each player plays as black and as white equally often. But this may be difficult to implement. All I will suggest is that whenever two players are matched other than for the first time in a tournament, they be assigned opposite colours from the previous time. This change should be easy to implement, because it will not interact with anything else.

Details of processor numbers, power, etc.

coldmilk, running on 4 cores, core i7 920 4.09Ghz.
cyVersion1, unspecified platform.
MoGo, running on double-core AMD Athlon(tm) 64 X2 Dual Core Processor 4800+ (2.5GHz).
Fuego, running on an 8-core Intel Xeon 2.5Ghz, 8GB of memory.
Gomorra, running with 8 threads on two Core i7 CPUs (not using hyperthreading).
running on Linux, 2GiB RAM, Intel(R) Celeron(R) M CPU 530 @ 1.73GHz
kiseki, running on a Windows PC with a core2quad 2.4GHz.
kurtBot, running on an intel i7 quadcore @ 2.67GHz (currently only using one core)
Many Faces of Go, running on 16 cores: four Core2 Quad 2.3 GHz.
MCark, running on a Core2Quad 2.4GHz (Only use 1 core)
pachi, running on an 8-core i7 machine;
GNU Go, unspecified platform
Steenvreter, running on an Intel Core2, Q6600 (4 cores)
valkyria, running on an Intel i7-860 processor using 4 Cores
running on Linux, 2GiB RAM, Intel(R) Celeron(R) M CPU 530 @ 1.73GHz
running on a Mac Pro 8 core, Xeon 2.26GHz.