How would you compare the best Shogi program to normal chess

[pichy]

Has Gekisashi one of the best Japanese Shogi chess program reached the level of a master or grandmaster level when you compare it to our normal or standard chess

http://www.aifactory.co.uk/newsletter/2 ... p_2008.htm

[Bill Rogers]

Although I have not followed it very closely I would think they are on par with Tom's TSCP program. Shogi is immensly more difficult.
Bill

[pichy]

Although I have not followed it very closely I would think they are on par with Tom's TSCP program. Shogi is immensly more difficult.
Bill

Probably Larry Kaufman would have to correct me, but top professional players seemed to agree that the best computer programs are at Shoreikai 3 Dan level. They may or may not have made a significant progress since then, I really don't know. Maybe another great human vs computer match would be just what is needed to bring the computers' strength into focus again. (I mean a game like the Watanabe vs Bonanza match.

Bonkras versus Yonenaga (2011–2012)

On 21 December 2011, computer program Bonkras crushed retired 68 year old Kunio Yonenaga, the 1993 Meijin. They played 85 moves in 1 hour, 3 minutes 39 seconds on Shogi Club 24. Main time was 15 minutes then additional 60 seconds per move. Yonenaga was gote (white) and played 2. K-62. This move was to confuse the computer by playing a move not in Bonkras's joseki (opening book).[15] On 14 January 2012, Bonkras again defeated Yonenaga. The game had 113 moves. Time allowed was 3 hours and then 1 minute per move. Bonkras moved first and used a ranging rook opening. Yonenaga made the same second move, K-6b, as in the previous game he lost. Bonkras ran on a Fujitsu Primergy BX400 with 6 blade servers to search 18,000,000 moves per second. Yonenaga used 2 hours 33 minutes. Bonkras used 1 hour 55 minutes.[16] Bonkras evaluated its game with Yonenaga in January 2012.

In order to compare a Shogi computer playing at Shoreikai 3 Dan level to normal standard chessI would say it is equal to 2600 in Computer chess rating

[lkaufman]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

[pichy]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

[lkaufman]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

I don't agree with this statement at all. I believe the top five chess programs could play on one core with just one hour and easily win a match against any human player with nine hours. Of course the human will score some draws, but probably no wins.

[pichy]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

I don't agree with this statement at all. I believe the top five chess programs could play on one core with just one hour and easily win a match against any human player with nine hours. Of course the human will score some draws, but probably no wins.

How did Nakamura managed to win so easy versus Rybka and how can they program it to avoid blocked positions

[Uri Blass]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

I am not sure that more time help humans in chess when we compare 9 hours per side with 2 hours per side even if we use some hardware that is slow enough that top humans are equal at 2 hours per side.

computers have a significant advantage at long time control and computers are never tired and have perfect memory when humans after a lot of calculation may forget their calculations(remember that making notes is not allowed in chess).

[lkaufman]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

I don't agree with this statement at all. I believe the top five chess programs could play on one core with just one hour and easily win a match against any human player with nine hours. Of course the human will score some draws, but probably no wins.

How did Nakamura managed to win so easy versus Rybka and how can they program it to avoid blocked positions

That was pretty funny, but Nakamura only scores something like 10% against computers on ICC, usually by blocking up the game and waiting to win on time or by the engine desperately trying to avoid a 50 move draw as in this game. Actually we solved the problem of blocked positions over 20 years ago, when our engines (Rexchess, Alpha, Socrates) generally scored the best against human GMs despite being not the best based on engine games. It is quite easy to solve this problem. The only reason every program doesn't do this is that they are judged by engine-engine ratings, which don't benefit from avoiding blocked positions. A suitable opening book will also be very helpful in avoiding blocked positions.

[pichy]

We don't have much to go by, but probably Yonenaga still plays at around Shoreikai 3 Dan level, so these games indicate that the computer is stronger than that. I would not be surprised if they are already mid-pro level or stronger, so perhaps in FIDE chess terms 2650 or even higher. It partly depends on the time limit, the slower the game, the lower the rating for the computer against humans. Perhaps it could already defeat Habu or Watanabe in a match at 30 minutes per side, but would lose badly at the Meijin time limit of 9 hours per side.

I did not know that Shogi could go up to 9 hours per side, if you give 9 hours to GM Kramnik or Aronian it would be very hard for any current top programs to beat them in normal classical chess.

I don't agree with this statement at all. I believe the top five chess programs could play on one core with just one hour and easily win a match against any human player with nine hours. Of course the human will score some draws, but probably no wins.

I finally came to the conclusion that a computer playing time handicup against the top 5 FIDE GM, but if you limit the computer to 15 seconds per move and either GM Aronian, Anand, or Kramnik with 5 minutes per move the result could be interesting and only a real match could determine the outcome.

PS: Nowadays programs like Hiarcs 13.31 on Iphone 4S can keep up with the best top players at blitz time control but I still believe that they are no match for the top 5 GM in standard time control


Here is another game this game was played at game in 25 minutes Fritz 5.32 played on an intel Duo Core 2.50 GHz which could give Fritz a rating of 2830

For comparison reason this game Fritz 5.32 played versus Judith Porgal in a laptop 350 Mhz, and My Iphone4 with Hiarcs 13.31 is playing a la par to fritz 5.32


[Event "Hiarcs vs Fritz 5.32"]
[Site "Pichy"]
[Date "2012.03.20"]
[Round "?"]
[White "HIARCS"]
[WhiteElo "2900"]
[Black "Fritz 5.32"]
[BlackElo "2830"]
[Result "1/2"]

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 e5 6. Ndb5 d6 7. Bg5 a6 8.
Na3 b5 9. Nd5 Be7 10. Bxf6 Bxf6 11. c3 O-O 12. Nc2 Bg5 13. a4 bxa4 14. Rxa4 a5
15. Bc4 Rb8 16. b3 Kh8 17. Nce3 g6 18. O-O f5 19. Qf3 fxe4 20. Qxe4 Bb7 21. Qd3
Bh6 22. Rfa1 Qh4 23. f3 Qg5 24. Ng4 e4 25. Qxe4 Rbe8 26. Qd3 Bg7 27. Kh1 h5 28.
Nf2 Ne5 29. Qd1 Nxc4 30. bxc4 Bxd5 31. cxd5 Bxc3 32. Ne4 Rxe4 33. Rxe4 Bxa1 34.
Qxa1+ Kh7 35. Qxa5 Qc1+ 36. Qe1 Rc8 37. h4 Rc5 38. Qxc1 Rxc1+ 39. Kh2 Rd1 40.
Re7+ Kh6 41. g4 hxg4 42. fxg4 Rxd5 43. Kg3 Rd3+ 44. Kf4 Rd4+ 45. Kf3 Rd3+ 46.
Ke4 Rd1 47. Rd7 Rh1 48. Kf4 Rf1+ 49. Kg3 Rg1+ 50. Kf3 Rf1+ 51. Kg2 Rd1 52. Ra7
g5 53. h5 Rd4 54. Kf3 Rd3+ 55. Ke2 Rd4 56. Kf3 Rd3+ 57. Kg2 Rd4 58. Kg3. Black offered a draw 1/2



Sent from my iPhone