|
Commentary & photos from 10th CSA, Tokyo 8-10th March 2000
Jeff Rollason
The article published with permission of author
Part I: Introduction
The 10th CSA, 8th-10th March 2000, was held at the Tokyo Bay Sheraton, and these are some of my impressions of this event, with a bias to coverage of the western entries. Of course there is also the Reijer Grimbergen's report from the event, and readers should turn to this for a more complete, broader and more Shogi-literate report.
Lunch in the peaceful Sheraton gardens provided an escape from mid-round stress! Jeff Rollason on the left and Reijer Grimbergen on the right.
This is the first time that the CSA tournament has had to span three days. This had become a matter of necessity as the number of entrants has increased, making the difficulties of running two rounds in one day increasingly impossible.
Now the first day only saw non-qualifying programs from the previous year and new programs battle it out in round 1 for 8 places in the second round. These 8 qualifying programs then combine with the 4th to 22nd seeds (16) who were either already in round 2 the previous year or qualified that year. From this 2nd round 5 programs qualify to join the top three seeds to enter an all-play-all to determine the placings of the top programs.
Expectations
Of course, Shotest was going to enter as it was already a top 3rd place seed, and therefore did not need to compete on days 1 or 2. To repeat my comments from the previous report: Although Shotest is essentially my project, I could not be in this position without help from others. Calvin Hutt of Oxford Softworks (who comes to the CSA) provides much patient strategic and practical support. Mick Reiss has provided valuable auto-testers, Martin Smith help on tree-searching knowledge and coding of the book and Chris Whittington for tree searching and evaluation. Essential shogi expertise has been also provided by Reijer Grimbergen and Marc Theeuwen. Thanks for all the support!
I expected this year to be tough. Shotest was in better shape than last year, when it was perhaps lucky to secure a second 3rd placing in two successive years. However work throughout the year had been diluted by other pressures, and so I was still not as happy as I might be about the state of the program. I considered my goal to try and keep in the top three, if I could.
The previous year the newly entered (and 7 year old) program Shocky written by Pauli Misikangas of Finland had made its debut. It failed to qualify from round one, but was running on a very slow machine. This year I knew that this program was worthy of regard as it had beaten the 9th CSA Shotest a few times in testing, and therefore looked capable of doing well. As Pauli and I had been in on contact on and off through much of 1999, we agreed to reduce our costs by sharing one of the otherwise spacious and luxurious rooms of the Sheraton. This allows us to do some pre-round testing with our hardware cluttering our living space. This highlighted bugs in my serial interface and also gave us a measure of how strong our respective programs were.
The only other western entry was of course Reijer Grimbergen's SPEAR. This interesting program is oriented towards intelligent pattern matching rather than mechanical brute force, and consequently has a very very low node rate, some 200 times slower than the more conventional programs. Despite this SPEAR wins games against programs that must be able to easily out-search it. SPEAR has been in the competition during the last few years, but has probably not been allowed to fulfill its potential because of lack of time to test it and remove bugs. Unlike many of the top authors, Reijer is far from full time, with many other pressures on his time, particularly this year. Both SPEAR and Shocky would be competing in round 1.
Part II: Pre-day diversions!
|
