The seventh tournament took place at The University of West Bohemia in PC lab UI 303. There were 11 computers Intel Pentium 4 3.4GHz HT, 1GB RAM in the lab. Drawing lots to decide basic groups was similiar as the last year. Considerable change was playing from openings (watch the image). One opening was chosen by Ales Rybka (gomoku player from www.piskvorky.cz known as Kaprar) , two others by me.
On Thursday evening I had all my stuff prepared and I have found out, that my old program Beta was not able to deal with openings. Because of confusion and hurry before the last tournament I had chaos in sources and versions of old version of Beta. So I decided to try to complete (in any working form) unfinished, entirely new version of my Beta. I had finished at half past five AM, I have been sleeping for three hours and went up to the lab. In the lab I have figured out, that my new Beta is compiled in a bad way and doesn’t work, so I was forced to eliminate it. It took me one hour to set all computers up and draw lots to decide the basic groups.
Programs was sorted according to the last tournament rank. Instead of removed Beta, there was one program (Tito) drawn lots from the group of not qualified and new programs to replaced it. Then one member of each qualification group was drawn lots from every trinity. The rest was drawn lots from the group of new and not qualified programs.
The basic groups was finished very fast (approximately in one hour). New Kanec was really a surprise, because it won the first basic group and defeated even swine, the former champion. Tito was present to be very good as well. The results are in the next three tables.
Then I estimated the duration of the final group and I have chosen the time limits 300s per move and 1200s per match. Final group was really suspensful. HTML interface have been automatically reloading and the placings have been continuously sorting. At first, Kanec’s chances was very good, then Tito and HGarden were catching it up. This three programs have been alternating in the the top of table. Then Goro started to fight through from 4th position. A moment before the end, Tito was still the leader and there were several interesting duels which determined the winner. In the meantime, Pela and Swine overtook Kanec with big effort. The best program in using time limits was Pela 818s per match, the worst QMentat 17s per match. It is interesting that Goro won even if it needed only 33s per match. It is obvious that Goro has still big reserves. There were no errors and no timeouts in the finals.
This tournament brings 4 new versions and one entirely new program. In contrast to the last tournament, there was much more time to think and programs had a lot of time to show off. Futhermore, the influence of presence or absence of the open book was decreased because of the openings. I checked the usage of CPU randomly (via Task manager) in final group. I didn’t catch any program to use more then 50% of CPU, so propably no program took advantage of two processor units.
In the end, I launched a fast game (5s per move, 30s per match), the result are in the table below. Pela stood up for the last victory.
Some programs didn’t respect time limits in the fast game, especially bestmove and brutefor.
30th April 2006, Tomas Kubes