A Visit to My Opponent's Camp: Introducing Man vs. Machine (CC) Challenges and Wchess

What to expect:

Computer chess and correspondence chess have developed at great speed in recent years and it has been a little too fast for me sometimes. When we discussed the details of Steve Ryan's Computer Chess Challenge it became clear that I had lost contact to computer chess for quite some time. The name Wchess, for example, did not ring any bells. After moving from Germany to the United States I was also a little lost in the world of (American) Correspondence Chess. Interest simply focuses on other people here. However, it was a pleasant experience to learn about other people and their work. The article, that you are reading, is (in great part) the result of my attempts to catch up. Many things I came across were new to me. Some things I already knew. I hope that the reader will experience the same. I tried to find a balance between past, present, and future and I think that some things are worth remembering. Young readers may be surprised to hear that computers haven't always been strong at rapid chess. It is also not true that the worlds of Correspondence Chess and Computer Chess are more or less distinct.

This article was written for a general audience who is simply interested in Computer Chess or Correspondence Chess. I tried to explain things as simple as possible, but I suppose that some things are not clear. I also suppose that there are some mistakes in what I have written. That is unavoidable - even more since I am not an expert on computers. I checked sources with care but sometimes they provided contradictory information. The text contains many links and I hope that readers who want to know more will find what they are looking for. Many of these links lead to sites that have pictures. I intend to update and extend this article and I would be grateful for suggestions or information on mistakes or inaccuracies. Feel free to contact me at: Volker@kasparovchess.com

It's a small world after all

Let's start with the first match between Garry Kasparov and Deep Blue at Philadelphia in 1996. If we take a look behind the scenes of this event then we find that the worlds of OTB-chess, Correspondence Chess, and Computer Chess are definitely not distinct. The arbiter of the match, Michael J. Valvo, is a perfect example.

• Valvo has been arbiter and tournament director of computer chess events for many years.
• He is a strong OTB-player holding the title of International Master.
• He has played Correspondence Chess for many years and earned a reputation for his match victory against Deep Thought in 1988-89.

Another man of multiple talents is co-commentator Hans Berliner. In the 1950s Berliner was a top-twenty OTB-player in the United States. Later he dedicated himself to Correspondence Chess and in 1967 he became the 5th World CC Champion. Feeling that he wasn't able to surpass his achievements in Correspondence Chess he stopped playing and put all his efforts into his academic career. He became a professor for computer science at Carnegie Mellon University at Pittsburgh where he headed the development of a chess computer called Hitech. Hitech became one of the strongest chess programs of the time. It is notable that Murray Campbell from the Deep Blue team was a student of Berliner's at Carnegie Mellon and that Berliner also worked briefly with Feng-hsiung Hsu (who was a chip designer then). And so we are back at Deep Blue where we started. Ken Thompson, known in the chess world for his endgame tablebases, also held a job at the event. He served as an advisor to Valvo in computer related matters.

More detailed descriptions of Hans Berliner, Ken Thompson, and Michael Valvo are to follow in the text. For the readers who would like to see some pictures I recommend using the link below. The article of Frederic Friedel on the history of computer chess has pictures of Berliner, Campbell, Hsu, Kasparov, Thompson, and a Deep Blue chip. I don't know about any picture of Valvo on the WWW.

Frederic Friedel: A Short History of Computer Chess at: http://www.clubkasparov.ru/hist/ha03e.htm

For the readers who are interested in the background and details of the first Kasparov vs. Deep Blue match I recommend the book by Monty Newborn, Kasparov versus Deep Blue: Computer Chess Comes of Age, published by Springer-Verlag New York in 1997. It contains a very detailed history of computer chess and many pictures (for example of all people mentioned above including Valvo). Newborn was the organizer of the event and he served as president of the International Computer Chess Association.

Taking on the machines in Correspondence Chess

1) Michael Valvo versus Deep Thought

The match between Michael Valvo and Deep Thought in 1988/89 is probably THE Man versus Machine event in Correspondence Chess (and, in my opinion, may never be topped). As I said in the previous section Valvo knew computer chess very well, but he brought more to the table than just that. At the time he was a very strong OTB-player with an USCF-rating of 2488. [Valvo is still a very strong OTB-player. He is an International Master (FIDE) and his ELO-rating in 1999 was 2375.] Two games were played in the match and the moves were posted on the Internet newsgroup rec.games.chess. Together with his moves, as it is a good tradition in challenge matches now, Valvo also posted commentary. The games lasted for five months (or 48 moves) and are extremely interesting. With the white pieces Valvo played a fine positional game while sacrificing material for a strong attack as Black. He defeated the machine both ways. It is interesting that up to move 12 the game Valvo - Deep Thought is identical with an earlier game of Valvo's (a win in 24 moves over Grandmaster Lev Alburt from the World Open in Philadelphia in 1988).

At the moment my knowledge of Valvo's career in Correspondence Chess is a little limited and I hope to provide more information in a later version of this article. What I can see from Tim Harding's database Mega Corr (of Correspondence Chess games) is that Valvo played CC already in the 1970s. It is probably a good bet that his most famous (human) victim in Correspondence Chess is Juan Sebastian Morgado whom he defeated in the 4th ICCF World Cup preliminaries. However, Morgado may not have been as strong then as he is now. Recently Valvo participated in the final of the 11th United States CC Championship and finished on 9 points out of 14 games.

White: Michael J. Valvo
Black: Deep Thought
Correspondence match, November 1988 - March 1989
Alekhine's Defense, Four Pawns Attack [B 03] 1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. c4 Nb6 5. f4 dxe5 6. fxe5 Nc6 7. Be3 Bf5 8. Nc3 e6 9. Nf3 Bg4 10. Be2 Bxf3 11. gxf3 Qh4+ 12. Bf2 Qf4 13. c5 Nd7 14. Qc1 Qf5 15. Qb1 Qxb1 16. Rxb1 0-0-0 17. f4 Be7 18. Rd1 g5 19. fxg5 Bxg5 20. Bf3 f5 21. 0-0 Nb4 22. Rfe1 Rhg8 23. Kh1 c6 24. a3 Na6 25. b4 Nc7 26. a4 a6 27. Re2 Be7 28. Rb2 Nd5 29. Nxd5 cxd5 [DIAGRAM]

30. b5 axb5 31. axb5 Rg7 32. Ra1 Nb8 33. Rba2 Rdg8 34. Ra8 Bg5 35. b6 Bd8 36. Bh5 Rf8 37. Be2 Rfg8 38. Be3 h5 39. Rb1 Be7 40. Bb5 Bd8 41. Ba4 f4 42. Bxf4 Rf7 43. Bh6 h4 44. Bb5 Be7 45. c6 bxc6 46. Bxc6 Rf3 47. Rba1 Ba3 48. Bd2 1-0

White: Deep Thought
Black: Michael J. Valvo
Correspondence match, December 1988 - March 1989
King's Gambit, Nimzovich's Countergambit [C 36]

1. e4 e5 2. f4 d5 2. exd5 c6 4. Nc3 exf4 5. Nf3 Bd6 6. d4 Ne7 7. dxc6 Nbxc6 8. d5 Nb4 9. Bc4 0-0 10. a3 b5 11. Bb3 Na6 12. Nxb5 Qa5+ 13. Nc3 Nc5 14. Ba2 [DIAGRAM]

14... Ba6 15. b4 Qc7 16. bxc5 Rfe8 17. Ne2 Qxc5 18. c4 Nxd5 19. Qd4 Qxd4 20. Nxd4 Bc5 21. Kd2 Ne3 22. Kc3 Rac8 23. Bb2 Nxg2 24. Raf1 Rcd8 25. Rhg1 Re3+ 26. Kd2 f3 27. Rxf3 Rxf3 28. Rxg2 Rh3 29. Kc1 g6 30. a4 Bb7 31. Rf2 Ba8 32. Bb1 Rb8 33. Ba2 Rd3 34. Rf4 Rd2 35. Kxd2 Rxb2+ 36. Nc2 Rxa2 37. Nc3 Rb2 38. Rf6 Kg7 39. Rf1 f5 40. Nd5 Bxd5 41. cxd5 Rb3 42. h4 Kf6 43. Re1 Rh3 44. Re6+ Kf7 45. a5 Rxh4 46. Rc6 Bb4+ 47. Nxb4 Rxb4 48. Rc7+ Kf6 0-1

More information on this match and the Deep Thought / Deep Blue computer programs: Monty Newborn, Kasparov versus Deep Blue. Computer Chess Comes of Age, Springer-Verlag New York, 1997.

Additional information on Michael Valvo:

Michael J. Valvo (* 1942) certainly knows computer chess very well having been the arbiter and tournament director of computer chess events for many years. His most prominent assignment certainly was as arbiter in the first match between Kasparov and Deep Blue at Philadelphia in 1996. In the rematch in 1997 he worked as a commentator (together with Yasser Seirawan and Maurice Ashley). Valvo is also famous for his abilities in blindfold play. In the early days of computer chess he used to play blindfolded against the entire field of machines.

More information on Michael Valvo can be found at the IBM website: http://www.research.ibm.com/deepblue/watch/html/c.13.html

2) CC World Champions versus ChessGenius 5

In March 1997, the German CC Magazine Fernschach International (Correspondence Chess International) organized a very interesting event which also features four Humans versus Computer games. Here, two former CC World Champions, Dr. Fritz Baumbach (GER) and Mikhail M. Umansky (RUS), play ChessGenius 5 in two-game matches. The pace of play is slow. At the moment only about 15 moves are played in each game. Unfortunately, it seems (I receive the magazine by surface mail from Europe and hence there is a time delay of about two months) that the games have come to a complete stop. The magazine Fernschach International has not published new moves between April and November 1999. I hope that the break is only temporary.

Additional information on Dr. Fritz Baumbach and Mikhail M. Umansky

Since the future of the above event seems uncertain I just repeat some basic facts. Dr. Fritz Baumbach (* 1935) and Mikhail Markovich Umansky (* 1952) are the 11th and 13th Correspondence Chess World Champion, respectively. Dr. Baumbach's style of play is clearly positional whereas (according to Hans Berliner, a World Champion colleague of the two,) "What Umansky does, he is sort of like Tal, he makes a tornado come out of a peaceful country scene." ( Chess Mail 10/1997) Dr. Baumbach was General Secretary of the International Correspondence Chess Federation (ICCF) and is still the President of the German CC Federation (Deutscher Fernschachbund). He (like some of his World Champion colleagues) has written a book on Correspondence Chess. The German title is (I have to write that from memory) 52-54 Stop: Tips und Tricks vom Fernschach-Weltmeister. I don't know if the book is available in English.

More information on Dr. Fritz Baumbach and Mikhail Umansky can be found at the Chess Mail website. The links are:

http://www.chessmail.com/baumbach.html http://www.chessmail.com/umansky.html

Note that these websites allow to access (part of) the interviews below in PDF format. They can be viewed with (Adobe) Acrobat Reader.

Sources in print:

CHESS MAIL 2/1999, Tim Harding: Governments fall, chess is eternal. (Biographical notes and interview with Dr. Fritz Baumbach).
CHESS MAIL 2/1997, Tim Harding: Through the eyes of the World Champion. (Biographical notes and interview with Mikhail M. Umansky).

My opponent - the long story

The chess computers that hit the department stores in the late 1970s weren't exactly a revelation. They could produce legal moves and recognized stalemate and checkmate. Some of them did not bother to check entered moves for legality and so it could happen that at some point of the game the computer suddenly crashed. The playing strength was very limited but this was not necessarily the programmers' fault. Hardware in general (CPUs and memory chips) was expensive and suitable hardware for chess programming cost even more. The manufacturers used hardware that kept the price within reasonable limits but, despite all these flaws, chess computers were a commercial success. There were some programmers who believed that computer chess had potential and were not afraid to work under the constraints given by available hardware. The American, David Kittinger, the programmer of my opponent Wchess, is one of these pioneers. He made a strong impression with his program for the NOVAG Super Constellation (SC). For readers who are not familiar with the early days it is probably hard to understand what made the SC so special. But keeping in mind that the first generation of computers failed to produce anything chess-like it was quite amazing to see the SC, with better hardware than the first generation models, play sequences of natural moves - even sacrifices! That's why one advertiser recently described the SC as legendary. The program was successful in competition (for example, in the 1983 World Microcomputer Chess Championship at Budapest it finished fourth on equal score with the second and third) and showed that computers could handle tactics well. Michael Scheidl, the author of the excellent web-publication Schach & Computer: Eine perfekte Kombination (see link below), observed: "SC's style of play was enterprising and appeared creative." Scheidl proves his point by showing the following game from "a small private tournament" at Vienna in 1991. Indeed, SC's piece sacrifice on move 25 is remarkable given the slow speed of the processor and the depth of the combination.

White: NOVAG Super Constellation
Black: Fidelity Avangarde
Vienna 1991
French Defense [C 19]

1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Ne7 7. a4 Nbc6 8. Nf3 Qa5 9. Bd2 Bd7 10. g3 c4 11. Ra2 Nd8 12. Qa1 Nc8 13. Bg2 Nb6 14. 0-0 Bxa4 15. Bc1 Nc6 16. Ba3 h6 17. Bc5 Qa6 18. Rb1 f6 19. exf6 gxf6 20. Re1 Kf7 21. Bh3 f5 22. Ne5+ Nxe5 23. Rxe5 Rhg8 24. Kh1 Qb5 25. Bxf5! [DIAGRAM]

25... exf5 26. Re7+ Kg6 27. Bxb6 axb6 28. Qe1 Rgf8 29. g4! f4 30. h4 Rae8 31. Qe6+ Rf6 32. h5+ Kg5 33. Rg7+ Kh4 34. Qxf6+ Kh3 35. Ra1 Qb2 36. Rg1 Qxc2 37. Qxf4 Qe4+ 38. Qxe4 1-0 (It is checkmate in four moves.)

Michael Scheidl: Schach & Computer: Eine perfekte Kombination (in German but with some pictures) at: http://www.computerschach.de/einleit/start.htm

Introducing Wchess

After receiving fame (about fortune I don't know) with the NOVAG Super Constellation Kittinger wrote the program Wchess. It is based on the operating system DOS (see below for some remarks on this fact). Like most chess computers and chess programs Wchess has been tested extensively by (the members of) the Swedish Computer Chess Association (SSDF). For information on test procedures and the newest list here is the link: http://home.swipnet.se/~w-36794/ssdf

There has been much discussion on how to compare SSDF-ratings to ELO-ratings. Not being an expert on this I simply avoid the topic. However, from the way the SSDF-list is generated it allows one to compare the strength of the different programs (with certain error margins). The September 1999-list of the SSDF contains about 200 entries, among them Wchess 1.03 and Wchess 1.06. The hardware specification associated with Wchess 1.03 is a 486 / 50-66 MHz processor while Wchess 1.06 is associated with a 90 MHz Pentium processor. I guess that the programs were designed for PCs with these specifications.

Fritz 3.0, like Wchess 1.06, is also a program for a 90 MHz Pentium PC. Both programs competed successfully in the 8th World Computer Chess Championship in 1995 and hence it is interesting to take a look at their respective ratings. In the quoted September-1999 list, Fritz 3 has an SSDF-rating of 2342 (+- 19 points error margin) versus Wchess's 2335 (+- 20 points error margin). Given the fact that both programs have more than 1200 rated games one can say that they are of about identical playing strength. But, of course, the truth is not entirely in the numbers. In the cited World Championship in 1995 both programs played the Deep Blue Prototype and ... well, see the tournament report below. By the way, we find that the NOVAG Super Constellation (6502 / 4MHz processor) has an SSDF-rating of 1731 (+- 18 points error margin) based on 1626 games.

It is understood that in chess programming five years are a lot of time. Wchess is not one of the top programs anymore. All of today's top software is now based on the operating system Windows which has some obvious advantages over DOS: 1) much more RAM (computer memory) is available for the program and 2) the RAM can be accessed much faster. These advantages allow the programmers to implement much more ambitious algorithms (and still get good results in games played under blitz and/or tournament conditions). Millennium GmbH recently released David Kittinger's newly developed program Wchess 2000, which they say shows significant differences in its playing style compared to its DOS-predecessor. However, the veteran WChess has achieved already what all those new programs are competing for. Two highlights of Wchess's career were its performances in the Harvard Chess Club Challenge in 1994 and in the 8th World Computer Chess Championship in 1995, which I wish to present in detail to the reader now.

Harvard Cup Chess Challenge

The first Harvard Cup Chess Challenge was held at Harvard University (Cambridge, Massachusetts) on October 29, 1989. It was a Man vs. Machine event where a team of humans (Lev Alburt, Maxim Dlugy, Boris Gulko, and Michael Rhode) faced a team of computers (Chiptest-M, Deep Thought, Hitech, and Mephisto Portorose) in rapid chess. Each human played each computer once at the reflection time of 30 minutes for the whole game. The result was a fantastic 14.5 - 1.5 victory for humankind. The machines managed only two scores: Deep Thought was able to win one game and Mephisto contributed a draw. So it seemed that even in rapid chess computers were no threat to professional players. However, in the following years the computer teams improved their results at Harvard. Already in the third Harvard Cup in 1992 the machines came rather close to the 50% barrier by scoring 7 points out of 18 games. In 1994's Harvard Cup six grandmasters faced eight computers. Again, each human played each computer and the humans won by the large margin of 29.5 - 18.5. Wchess, however, caused a sensational upset. The program won four games and drew the remaining two making it an impressive score of 5 points out of 6 games. In the following year the humans seemed eager to put Wchess back into its place. They succeeeded. The program was defeated four times and achieved only one win and one draw.

[The games of these events are included in databases at Frank Quisinsky's website, see the link (with explanations) below.]

8th World Computer Chess Championship

The first World Computer Chess Championship (WCCC) was held in Stockholm (Sweden) in 1974. The Russian program Kaissa scored a perfect 4-0 and clinched the title. It ran on an English ICL 4/70 computer, which enabled the program to evaluate about 200 (!) positions per second. Well, times were different then (In 1996, Deep Blue was able to evaluate more than 500 million positions per second). Runner-up was the American program Chess 4.0 written by Slate and Atkin. Three years later - the WCCC is held every three years - Slate and Atkin won the title with Chess 4.6. The 8th World Computer Chess Championship in 1995 was held at the Chinese University in Hong Kong. The event was a 5-round Swiss with the usual rate of play as in OTB-tournaments: 2 hours for the first 40 moves, then 1 hour for each additional 20 moves. It was a special event since six years earlier, in the 6th WCCC at Edmonton (Canada), Deep Thought had swept the opposition 5-0 and everybody was curious to see the new program. It seemed that history was about to repeat itself since after three rounds Deep Blue had won each of its games and was leading the field by one full point. In round four it was paired against Wchess, which it had defeated in a tournament the year before. However, the game between the giant and the underdog developed quite unexpectedly. After an unspectacular opening (okay, a boring opening) the programs opted for an endgame with unequal material (Diagram 1). Deep Blue's bishop together with the two connected passed pawns seem to be good compensation for the missing rook. On the other hand, Wchess's rooks are placed in textbook style to stop the advance of the pawns. One is placed in front of them, the other behind them. So, I think, the position is in dynamical balance. The following moves brought quite a surprise since Wchess did not only stand its ground but also managed to win the a-pawn (Diagram 2). The game may still be a draw objectively but, in my opinion, Deep Blue must play this endgame very accurately. I would prefer 50. Kxe4 over the played move 50. Kf2 and think that this would have given some practical winning chances.

White: Wchess
Black: Deep Blue (Prototype)
Hong Kong, 05/28/1995
8th World Computer Chess Championship, Round 4

Sicilian Defense, Alapin Variation [B 22]

1. e4 c5 2. c3 d5 3. exd5 Qxd5 4. d4 Nf6 5. Nf3 e6 6. Be2 Nc6 7. 0-0 cxd4 8. cxd4 Be7 9. Nc3 Qd6 10. Nb5 Qd8 11. Bf4 Nd5 12. Bg3 a6 13. Nc3 0-0 14. Qb3 Nf6 15. Rfd1 b5 16. a3 Bb7 17. Qa2 Na5 18. b4 Rc8 19. Rac1 Nc6 20. Bf4 Re8 21. d5 exd5 22. Nxd5 Nxd5 23. Qxd5 Qxd5 24. Rxd5 Bxb4 25. axb4 Rxe2 26. Be3 Re8 27. Rd7 Ba8 28. Nd2 Nxb4 29. Kf1 R2xe3 30. fxe3 Nd5 31. Kf2 [Diagram]

31... h6 32. Nf1 Nb4 33. Nd2 Bd5 34. Rb1 Be6 35. Ra7 Nd3+ 36. Ke2 Nc5 37. Rb4 Bd5 38. g3 Ra8 39. Rxa8+ Bxa8 40. Rd4 Kh7 41. Rd8 Bb7 42. Rb8 Bh1 43. Rc8 Ne6 44. e4 Bg2 45. Ke3 Bh3 46. Rc6 f5 47. Rxa6 Nc5 48. Rd6 Nxe4 49. Nxe4 fxe4 [Diagram 2]

50. Kf2 Bg4 51. Rb6 Bf3 52. Rxb5 g5 53. Ke3 Kg7 54. Rb7+ Kg6 55. Rb6+ Kg7 56. Re6 h5 57. Rd6 h4 58. g4 Bxg4 59. Kxe4 Bh3 60. Rd3 Bg4 1/2-1/2

This draw and the round 5-loss to Fritz 3 proved costly for the Deep Blue team. The resulting 3,5 points proved not to be enough to get an entry in the winners list. (In World Computer Chess Championships usually only the winner and the runner-up is listed, making it effectively only two medals.) Fritz 3 went on to win the World Champion title after a tiebreak game with Star Sokrates. The Wchess program remained undefeated (one win, four draws) and came in 6th. It is quite remarkable that two programs running on an ordinary Pentium 90 Megahertz computer scored against the much more powerful IBM machine.

[The games of this event are included in a database at Frank Quisinsky's website, see link (with explanations) below.]

The link to the download area at Frank Quisinsky's website is: http://www.in-trier.de/~quisinsky/schach/download/download-q.htm

The names of the most relevant files are Turniere 70-89 and Turniere 90-2000. As the names indicate they contain computer games from tournaments held 1970-1989 and 1990-2000, respectively. There is a PGN (portable game notation) and CBH (new ChessBase format) version of each file. Note that these files are big and zipped (compressed). The first one has more than 700 games, the second more than 4,000. Note also that Wchess appears in the second file as WChess.

For newbies: Before you can look at the games you need to "unzip" (decompress) these files. There are several programs that can do that. Some are available for free download on the Internet. My knowledge on this topic is very limited but if you need help you can contact me. In principle, PGN files can be opened with a text editor. The format is "plain text" with the usual English abbreviations for the pieces and short algebraic notation. However, this is inadvisable for files containing many games since the process takes a lot of time (even with a fast computer) and memory. It is much more efficient to use a PGN reader. I heard there are many good programs but most of them I don't know. It is probably best to ask the experts on "The Correspondence Chess Message Board" (TCCMB). Ralph Marconi's homepage has a link.

Wchess 2000

For those readers who are more interested in what is going on at the moment I would like to point out that Wchess 2000 is competing (with many of today's top programs) under tournament conditions in a 7-round Swiss tournament organized by Jrn Gronemann. Internet coverage is provided in English and German at the address below. The site contains hardware specifications and other tournament details. The games of each round are available for online viewing and for download. Gronemann also provides round reports. One of the most exciting games so far is the game between Wchess 2000 and the reigning World Computer Chess Champion Shredder 4.0.

White: Wchess 2000
Black: Shredder 4.0
Private computer chess tournament
Ahrensburg (round 2), January 2000

Sicilian Defense, Najdorf Variation [B 92]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Be2 e5 7. Nb3 Be7 8. 0-0 Be6 9. f4 Qc7 10. g4 0-0 11. f5 Bxb3 12. axb3 Qc5+ 13. Kh1 h6 14. Qd3 Nc6 15. Be3 Nb4 16. Qd2 Qc6 17. Ra4 d5 18. exd5 Nfxd5 19. Nxd5 Nxd5 20. Rc4 Nxe3+ 21. Rxc6 Nxf1 22. Qd7 bxc6 23. Qxe7 Nd2 [DIAGRAM]

24. f6 Rfe8 25. Qd7 Red8 26. Qxc6 Rac8 27. Qb6 Re8 28. fxg7 Kxg7 29. c3 Re6 30. Qb7 Rcc6 31. g5 hxg5 32. Bh5 Rf6 33. Bxf7 Rcd6 34. Bd5+ Kg6 35. Qa8 g4 36. c4 Kg5 37. Qg8+ Kh4 38. Qh7+ Kg5 39. Qd3 Nf3 40. c5 Rd8 41. c6 Rh8 42. Qe3+ Kg6 43. Be4+ Kg7 44. Kg2 Rxh2+ 45. Kg3 Rh8 46. Bd5 Rhf8 47. c7 a5 48. Qc5 Nd4 49. c8Q Ne2+ 50. Kxg4 Rg6+ 51. Kh3 Nf4+ 52. Kh4 Rh6+ 53. Kg4 Rg6+ [DIAGRAM]

54. Kf3 Rxc8 55. Qxc8 Nxd5 56. Qd7+ Kf6 57. Qd6+ Kf5 58. Qxd5 Re6 59. Qd7 1-0 Coverage of Jrn Gronemann's private 7-round Swiss computer tournament (in English and German): http://www.computerschach.de/tourn/joern00.htm

Good advice for playing computers from people who know them

Ken Thompson:

Kenneth L. Thompson (* 1943) is a pioneer of computer chess whose influence on the subject can hardly be overestimated. In the 1970s, Thompson (together with Joe Condon) worked on building a chess computer. One important ingredient was to develop special-purpose, high-speed hardware to run a strong chess program on it. The result was a machine called "Belle" which could only play chess. Around 1980, Belle was able to search 120,000 chess positions per second. To achieve this task it needed 1,700 especially designed chips (it seems they were all the same). In tournament chess Belle was able to look four to four and a half moves deep which was enough to compete successfully at master level. It dominated computer chess events in the years from 1980, when it won the World Computer Chess Champion title in Linz (Austria), to 1983.

Games of Belle are included in the file Turniere 70-89 in the download area of Frank Quisinsky's website: http://www.in-trier.de/~quisinsky/schach/download/download-q.htm

(Note: The subchapter 8th World Computer Chess Championship of this article contains some explanations on downloading and viewing chess games.)

Thompson also examined the connection between search depth and playing strength of a chess machine. He concluded that a computer must look at least seven moves deep to be a match for a player like Kasparov. In the chess world today Thompson is probably best-known for the so called endgame tablebases (EGTB). Using EGTB, which come on CDs because of the enormous volume of data, a computer can play endgames with five chessmen or less with perfection. These tablebases contain all crucial information for all legal position of these types. Examples for the positions covered by EGTB are those where White has a King and a Queen and Black has only a King and a Rook. In most of these positions White has a forced win. The EGTB indicate if a given position is a forced win with best play for the stronger side (White in the example above) or if the weaker side (Black in the example above) can hold. The tablebases do not only know the outcome for each position (best play assumed) but also know the outcome after each possible move in this position. So if the stronger side was to move and a win was possible then the EGTB would say what the best move is (often there is more than one) and how many moves it takes against the best defense. If the game is a draw one can see which moves are good enough for a draw and which would lose. Thompson has already solved some endings with six pieces and posted the results on his webpages at Bell Laboratories (use the link below and click Chess endgames). Thompson does not provide much documentation but the reader will probably find out by herself / himself what is meant and how it works.

(Note: For more details on this and photos of Ken Thompson and Belle, respectively, see the article by Frederic Friedel below.)

But let's turn to computer chess of recent days, new developments, and Thompson's opinion about it. Like Thompson and Condon, the Deep Blue team also used a special-purpose chip. A couple of weeks ago, Feng-hsiung Hsu released an open letter that contained some information on the chip and its importance for a chess playing machine. He wrote that he bought the right from IBM to commercialize the chip and stated that [with the chip] it would be possible to build a PC based machine that should be every bit as strong as Deep Blue, if not stronger. It seems to me that Hsu is indirectly saying (true or not) that the strength of a chess computer depends mostly on the available hardware. This would mean that improving the software, for example in the area of the evaluation function, is not seen as very difficult or critical. In this context it is interesting to hear what Ken Thompson said when asked about Deep Blue's success over Kasparov. In an interview (details see below) he said: "... Mostly they [the Deep Blue team] had a lot more positional evaluations. It was hardware evaluations." He added that, in his opinion, Kasparov was still stronger than the computer and had beaten himself. Questioned about his assessment of Deep Blue's chances in a tournament like Linares Thompson gave the following general answer. "... I think they [the Deep Blue team] would fail against a lot of the top players, against most of those players, but they'd end up winning some games anyway. That's the way computers have always played. They defend tenaciously until the human does something stupid to try to win and lose. ..." So we know now what to expect from a strong chess computer.

[The interview I am quoting from was part of an article by Frederic Friedel called The endgame database man and was posted at the ChessBase website during the winter holidays 1998/99. Unfortunately, it is not there anymore.]

Additional information on Ken Thompson

In April 1999, Ken Thompson was awarded the U.S. National Medal of Technology from President Bill Clinton in a ceremony at the White House. He developed the operating system UNIX in the 1960s and later also contributed to the development of the computer language "C."

Page of Ken Thompson at Bell Laboratories http://cm.bell-labs.com/who/ken/ Frederic Friedel: A Short History of Computer Chess at: http://www.clubkasparov.ru/hist/ha03e.htm Hans Berliner

Hans Berliner (* 1929) began his career in chess as an OTB-player. He learned chess at the relatively late age of 13 but soon (sometime in the 1950s) he had managed to work himself into the top-twenty list of the United States - his best position in the ranking list was 12th. He considers his fifth place in the 1957/58 US-Championship at New York as his best success in OTB-chess. Berliner came in behind Fischer, Reshevsky, Sherwin, and Lombardy but was ahead of Denker and Mednis. He finished on a positive score (+4 -3 =6). Berliner later dedicated himself to Correspondence Chess and achieved incredible results. In 1968, he won the 5th World Correspondence Chess Championship (in CC, other than in OTB-chess, the world champion is determined in a tournament and not in a match) by an incredible margin of three full points. This achievement is without parallel and may be a record never to be broken. Feeling that he could not surpass his achievements in CC he stopped playing Correspondence Chess and put all his efforts into his academic career instead. He became a professor for computer science at Carnegie Mellon University at Pittsburgh. There, Berliner headed the development of a chess computer called Hitech, which became one of the strongest machines of the time. Berliner's creation, like Belle and Deep Blue, included special-purpose hardware. In the 5th World Computer Chess Championship, at Cologne (Germany) in 1986, Hitech finished second (Cray Blitz won the event with equal score but better tiebreaker).

Games of Hitech are included in the files Turniere 70-89 and Turniere 90-2000 in the download area of Frank Quisinsky's website. http://www.in-trier.de/~quisinsky/schach/download/download-q.htm

(Note: the subchapter: 8th World Computer Chess Championship of this article contains some explanations on downloading and viewing chess games.)

It is notable that Murray Campbell from the Deep Blue team was a student of Berliner's at Carnegie Mellon and that Berliner also worked briefly with Feng-hsiung Hsu (who was a chip designer then). The article by Frederic Friedel (see link below) contains a picture of Hans Berliner and some related information.

In an interview with Allan Savage (it is the second of the two below) Berliner describes the right way to play against a computer as follows. "...The right way to do it is to get a minuscule edge against the computer, and then don't do much. And then let it say, oh my God if I do this, I will be a tenth of a pawn down and if I do that I am only going to be an eleventh of a pawn down, so it is better to do that. While it is doing this it is retreating all the time and some of this retreating decisions are going to be wrong."

Frederic Friedel: A Short History of Computer Chess at: http://www.clubkasparov.ru/hist/ha03e.htm More information on Hans Berliner can be found at the Chess Mail website http://www.chessmail.com/berliner.html

Sources in print:

CHESS MAIL 10/1997, Allan G. Savage: Berliner tells all! (Part one of a long interview with Hans Berliner).
CHESS MAIL 11/1997, Allan G. Savage: Hans Berliner on games, artificial intelligence and the road to Deep Blue. (Part two of a long interview with Hans Berliner).
FERNSCHACH 1/1999, Fritz Baumbach: Rckblick / Interview Ex-FS-WM Dr. Fritz Baumbach mit Ex-FS-WM Dr. Hans Berliner. (Biographical notes and an interview with Berliner; in German).

Thank You

I am grateful to the following people for their help with this project. I would like to thank Steve Ryan, Stephen Ham, and the people of the Millennium GmbH for providing me with information. Of course, mistakes in the text are my responsibility and nobody else's. I also would like to thank Ralph Marconi for his enthusiasm when I suggested the project. Finally, I thank my wife for proofreading the manuscript and her many suggestions on how to improve upon it.