Eighty-first KGS Computer Go Tournament

Sunday April 1st, 2012

These results also appear on an official KGS page.


format12-round Swiss
board size13×13
time19 minutes plus 10/30s


The first round started at 08:00 UTC.

Result table

Zen13 pachi stv ManyF AyaMC oakfo Orego
B12R W17R B112R W16R B010½ W13R B15R W18R W11R B111T B14R 117365Winner
2pachi2 W02R B07R W012R
W14R B111R W15R B18R W13R B110R B19R W11R 98249
3stv B06R W110½ B04R W011R
W19R W01R B112R W13R B17R B15R W18R 87338
4ManyFaces1 B03R B05R W08R B09R
B12R W07R B14R W112F W16R B111R 76320
5AyaMC W05R B08R B03R W010R B11R W012R W02R B17R
W16R B19R 67722
6oakfoam B01R W011T W09R B03R W07R W04R B012F B06R
B12R W110R 4736
7Orego12 W04R B01R W05R B08R B06R W011R W09R W02R B010R

In the table above,
   0 is a loss
   1 is a win
   J is jigo
   left superscript is the player's colour
   right superscript is the round in which the game was played
   a subscript shows how the result was determined:
      R for resignation
      T for time
      F for forfeit
      a number for the points difference after counting.
All the 0s, 1s and Js are links to the game record.

"April Fools" day on KGS

As soon as I connected to KGS, about 40 minutes before play in this tournament was due to was due to start, I realised that it was "April Fools" day. In the past, KGS has celebrated this by changing its graphics, using pink and purple stones instead of black and white. That was amusing and popular – and did nothing to interfere with its normal operation as a Go server.

This year, there were two changes to KGS. The font which users use to chat had been changed, to one which made text in the Latin alphabet hard, though not impossible, to read. And many room names had been changed in a jokey way, in particular the "Computer Go" room had become "Confuser Go".

I did not consider either of these amusing. The essence of an "April Fools" joke is that some people are fooled: they don't understand what is happening, and are puzzled, until they realise that it is deliberate. But with two (rather poor) jokes, it was immediately obvious that it was deliberate. I did not notice any users who were fooled, though admittedly, with the changed font, I did not make much effort to read what they were typing.

The consequences for the Computer Go tournament were severe. Two bots had been left running in what had been the "Computer Go" room when they had connected to KGS, and was now "Confuser Go". A helpful user connected after the name change, found that there was no "Computer Go" room, and created one; and some bots using their normal .ini files had joined this. I think one of these bots was also unattended, so I had no way to explain to its operator what was happening. A KGS admin who had realised there was a problem (perhaps alerted by my appearance early on a Sunday morning) tried to rename "Confuser Go" back to "Computer Go"; and failed, because a "Computer Go" room already existed. Although several admins realised that this was an urgent problem, the font change made it hard for us to chat to each other about how to fix things.

Eventually we renamed "Computer Go" to "not Computer Go", left a message in its room info (where the font was legible) telling bot operators to go to the real Computer Go room, and then renamed the "Confuser Go" room back to "Computer Go". But this still left three bots in the wrong room, one of them with no active operator. I kicked these three bots, and was pleased to see them reappear in the correct room.

Meanwhile, Erik van der Werf asked me (after the deadline for registrations) whether stv (his program Steenvreter) would be welcome to play. Normally I would have refused, as this was after the 30-minute deadline before the start of play, and would result in an odd number of registered players. But at the time I could not be sure how many of the registered players could be got into the right room anyway, so I accepted.

I apologise, on behalf of KGS, to all bot operators for the difficulties caused by the "April Fool" changes. I apologise particularly to those who do not normally use the roman alphabet: you must have been inconvenienced even more than I was by having to read a distorted form of it. I thank you all for your tolerance.

Partly because of the difficulty of reading users' comments, I followed this tournament much less closely than I usually do. Therefore I have little to say about any of the games, and my report is short.


oakfoam vs stv
Move 82.

In round 7, oakfoam played move 82 against stv as shown to the right. This seems strange. It threatens to start a ko to win the semeai with Black's lower left group. It seems that playing at a5 instead must be better (though still pointless): a5 threatens to win the same semeai outright.
       But there may be a reason to play at a6 rather than a5. I recall having a similar move played against me in a tournament, when I was 5-kyu. My opponent had realised that I would notice and answer a direct threat, and hoped that I might fail to notice a threat which only worked in ko. Is it possible that MC-based move generation can generate such a move, with White realising that a5 will always be answered and will therefore achieve nothing, but finding that a6 is ignored in some roll-outs, giving it a chance to win?
       Stv was not fooled. It responded at a4, winning the semeai.

In round 10, stv beat Zen13 by half a point. This was Zen13's only loss.

In round 11 oakfoam stopped moving in its game with Zen13, and lost on time. I later learned from its operator that it lost its internet connection, and was unable to reconnect to KGS for the rest of the tournament.

In round 12 oakfoam did not join its game with ManyFaces1, and lost on time.

Annual points

Players receive points for the 2012 Annual KGS Bot Championship as follows:

Many Faces of Go2

Details of processor numbers, power, etc.

Aya, running on 6 cores of an i980X, at 3.3GHz.
Many Faces of Go, running 4 cores (8 threads) of an i7-2600.
Oakfoam, probably running on an Intel Xeon E5645 @ 2.40 GHz
Orego, probably running on one of the five nodes of a custom Linux cluster built by PSSC Labs: the node has two AMD Six Core Dual Opteron 2427 2.2 GHz (12 cores total), 8 GB RAM, Centos Linux.
pachi, running on 64 platforms, each x86 64 bits, 32 GB ram, using 22 cores of each (total 1408 cores), giving about 1500 playouts/s/core at the beginning of a 19x19 game.
Steenvreter, running 46 threads each at 2.2 GHz, on a system whose use was generously provided by the Maastricht games and AI group.
Zen, running on a Mac Pro 8 core, Xeon 2.26GHz.