In his 1984 thesis Computer-Chess Tactics and Strategy, Alex Szabo elaborates on tactics and strategy exemplified by his thesis chess program Tech 3. In the spirit of the original Tech program by James Gillogly, he claims that knowledge is best applied at the top of the search tree as positional presort rather than at the leaf nodes using complex evaluation.
Tech3's performance on Reinfeld'sWin at Chess (WAC) problem set is 274/300, which compares favourable with Belle's, considering machine power. He developed a technology curve as a tool for measuring the effectiveness of knowledge encoding, and found in this respect Nuchess as the best chess program of that time [2].
TechMate
During the mid 80s, Alex Szabo developed TechMate for the Atari ST. It was derived from Tech 3 with the addition of a simple strategic component, a more profound evaluation function. TechMate was commercially market by Szabo Software[3].
As published in their 1988 ICCA Journal paper [6], Alex and Barbara Szabo revisited the technology curve by playing 6882 games between copies of TechMate set at different time rates, with the conclusion that the advantage of improved technology rapidly decreases when machines and algorithms become more powerful. Ernst A. Heinz in his self-play memo on the experiment of the Szabos [7]:
The Szabos determined the technology curve of their chess program TechMate that self-played 6,882 games on two Atari ST computers. The number of games per match between longer and shorter searching versions of the program varied strongly from a minimum of 32 to a maximum of 1367. The gain in playing strength averaged at 156 rating points per doubling of available search time (computing power). The experimental data indicated slight diminishing returns at longer search times. However, the Szabos simply did not play enough games at long times to draw reliable conclusions.
#230 [Rb4 does not win. The main line is, 1... Rb4!? 2.cxb4 a4
3.b5+ Kxb5 4.Ba3 c3 5.Re2! Kc4 6.f4 Kxd4 7.f5 exf5 8.e6 Kd3
9.e7 Bd7 10.Kf3 d4 11.Rh2 Kc4 12.Rh8 b2 13.Rb8 d3
14.Bxb2 cxb2 15.Rxb2 a3 16.Rb7 Be8 17.Ra7 Kb3 18.Ke3 a2
19.Kxd3 Kb2 20.Rb7+ Kc1 21.Ra7 =]
Ra7, Rb6, Rb5, Rd7, Rf7, Rg7, Rh7, Bd7, Kd7, Kb6, Kb5, a4, and Rc7
are just as good as the book solution Rb4 -- they all hold the game.
Table of Contents
Alexander (Alex) Szabo,
a computer scientist, physicist, chess player, at times professional computer chess programmer, and along with Barbara Szabo conducting a computer programming services company, located in San Antonio, New Mexico [1]. In the 80s they were active in computer chess business and research, Szabo Software located in Borrego Springs, California, distributed their commercial chess program TechMate for the Atari ST. Alex Szabo holds a M.Sc. in physics, 1980, and a M.Sc. in CS, 1984, both from University of British Columbia, Vancouver, British Columbia, Canada.
Tech 3
In his 1984 thesis Computer-Chess Tactics and Strategy, Alex Szabo elaborates on tactics and strategy exemplified by his thesis chess program Tech 3. In the spirit of the original Tech program by James Gillogly, he claims that knowledge is best applied at the top of the search tree as positional presort rather than at the leaf nodes using complex evaluation.Tech3's performance on Reinfeld's Win at Chess (WAC) problem set is 274/300, which compares favourable with Belle's, considering machine power. He developed a technology curve as a tool for measuring the effectiveness of knowledge encoding, and found in this respect Nuchess as the best chess program of that time [2].
TechMate
The Technology Curve
As published in their 1988 ICCA Journal paper [6], Alex and Barbara Szabo revisited the technology curve by playing 6882 games between copies of TechMate set at different time rates, with the conclusion that the advantage of improved technology rapidly decreases when machines and algorithms become more powerful. Ernst A. Heinz in his self-play memo on the experiment of the Szabos [7]:WAC 230
#230 [Rb4 does not win. The main line is, 1... Rb4!? 2.cxb4 a4 3.b5+ Kxb5 4.Ba3 c3 5.Re2! Kc4 6.f4 Kxd4 7.f5 exf5 8.e6 Kd3 9.e7 Bd7 10.Kf3 d4 11.Rh2 Kc4 12.Rh8 b2 13.Rb8 d3 14.Bxb2 cxb2 15.Rxb2 a3 16.Rb7 Be8 17.Ra7 Kb3 18.Ke3 a2 19.Kxd3 Kb2 20.Rb7+ Kc1 21.Ra7 =] Ra7, Rb6, Rb5, Rd7, Rf7, Rg7, Rh7, Bd7, Kd7, Kb6, Kb5, a4, and Rc7 are just as good as the book solution Rb4 -- they all hold the game.In August 2010, Dann Corbit posted an allegedly refutation of Szabo's refutation found by Stockfish 1.8 [11].
Selected Publications
[12]Forum Posts
External Links
References
What links here?
Up one level