|
Commentary & photos from 10th CSA, Tokyo 8-10th March 2000
Jeff Rollason
The article published with permission of author
Part V: Day 3 - Introduction to The Final
Martin Muller, Pauli Misikangas, Reijer Grimbergen and Jeff Rollason watch the changing fortunes of the Shocky vs IS-Shogi game.
Now it was my turn to sweat! I already felt exhausted in following Spear and Shocky in the first two rounds. I knew now that I faced very strong programs, including Shocky. The latter should not have really made me very afraid as pre-round tests in my room had given Shotest a perfect 9:0 score. Shocky was running for most of these games on a slower laptop, but we gave Shocky proportionally more time, so this should not have made much difference.
I expected to beat Shocky, but I had seen these pre-round games, and in some it had at given Shotest a very hard time. I therefore knew that Shocky was capable of winning. Bizarrely Pauli ran a 6 month old version of Shocky as his newer versions all played inferior Shogi! He worked very hard in the last few days trying to make his theoretically superior new version play better, but in the end gave up and resigned himself to using an old version which he did not have the exact sources for! This kind of crazy phenomenon is not so rare. It is common that blatant bug-fixes and obvious improvements often cause a program to play much worse. I have several ridiculous bug-fixes which are suppressed for this very reason. It's stuff like this which sends games programmers round the bend.
Shocky would also now be running on a much faster 750MHz machine rather than the 450MHz laptop of rounds 1 and 2 (the laptop only performed at about the speed of a 233MHz machine though). This upgrade was made possible by Reijer generously donating his machine to the cause, as he had not qualified.
Part VI: The Final Rounds
|
