Gomocup 2008, the 9th tournament (April the 19th, 2008)

The ninth tournament took place in Prag in Small Town’s Central library.
By the way there is very interesting construction from books. For the first sight I didn’t believe my own eyes.
We were dealing with technician problems all the morning and we started with delay.

The innovation of 2008 was done on tito, Onix, Swine, Kanec, HGarden, Hewer, Nabamoku.
And there are three entirely new AI: Fiverow by Yaroslav from the Ukraine and PureGM by Paul Dwyer (Australian living in Japan)
and Hector for Gomoku (called HfM) by Jergler Csaba. Hector is a gomoku version of chess program and it is private,
it means that is not available for download.
But we didn’t know about these AI on Saturday.

At first we cast last year’s finalists by lots. And then we cast others by lots.
Then we tried vainly set up network among computers, but it didn’t work, so we launch groups separately.
We used two dual CPU laptops with Intel centrino Duo T2300e @ 1.66Ghz and Intel Core 2 Duo CPU T7500 @ 2.2Ghz.

After closing time of library, we move to the Pivrnec Pub to finish groups. After that we each went home.
Then I fixed some bugs in online broadcast and in the evening we connected through internet using hamachi and we launch final.
It was running all the night and all the following day.
On Saturday’s evening was allmost finish and we have got unpleasant surprise.
We obtained a lot of emails apparently sent on Friday.
There were innovations or new programs of already mentioned AI Tito, Hector, PureGm and Nabamoku.
We decided to partially annul results and to start tournament again.
The results of first and third group was kept because affected AI Hector and Tito promoted anyway to the final.
So the second group and final was annulled.
After second run of second group was the promoted set the same as for the first time.

Goro didn’t promote to final because of sending emails.
A nice surprise was performance of Hector.
Basic groups were win by HGarden, Swine and Tito. Otherwise there was no further surprise.
The results are in the following table.

Results of group number 1

Results of group number 2

Results of group number 3

Composition of AI promoted to the final was similar to the last year, only Trunkat and Noesis was replaced by Hector and Hewer.
Time limits stayed on established boundary 300s/move and 1200s/game, because we haven’t so much computers as last year.
In the last year it was in the final 400s/move and 2200s/move.

Before annulations was showed that the combat of this final will be combat Tito versus Onix.
Onix was fixed and didn’t move into bad coordination, but was still a little bugy and three times allocated to much memory.
Before annulations of results it seems to draw or tight win of Onix.
Innovated Tito (probably due to Onix’s over-allocations) became the winner, even with worse score.
Tito wins the more important games than Onix.
Another bug (EAccessViolation) was showed by Kanec.

Among the ‘win reporters’ the deepest solution was founded 46 moves ahead by swine against Pela.
Then 36 moves ahead by Hewer (but Hewer is sometime mistaken in reporting win) against Kanec and 26 moves ahead by Onix against Hector.

The least thinking was swine in average 2,6s and the most thinking was Pela in average 27s.

Hewer and Hector gained the same count of points, the same score and the mutual draw; they are together on 4th and 5th place.

In the end, the fast-game was launched. Only AI being (at least a little) able to play gomoku fast were chosen.
Tito was the winner again, HGarden was the second and Hewer was the third.
Onix was very often overtime.

You can download complete results and openings here

Home page