This area was where Gothic Vortex played the SMIRF Engine in a long time control game where each engine had up to 24 hours to make a move. Most moves were made much more quickly than this. SMIRF won with a checkmate in 73 moves.
You can replay the game one move at a time by clicking this link:
Vortex vs. SMIRF
[Event "One Move Per Day: Blog Game"]
[Site "GothicChess.BlogSpot.com"]
[Date "2007.10.07"]
[Time "00:25:48"]
[Round "Blog-Game-1"]
[White "Gothic Vortex"]
[Black "Smirf MS-168i"]
[Result "0-1"]
[Annotator "RS"]
[SetUp "1"]
[FEN "rnbqckabnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBQCKABNR w KQkq - 0 1"]
1. Nh3 g6 2. d4 Nh6 3. g3 c6 4. Bf3 Na6 5. Bf4 f6 6. Be3 Qb6 7. Nc3 Qxb2 8. Cd3 Qa3 9. Ne4 Qa5 10. d5 Qxd5 11. Cxd5 Axd5 12. Bxh6+ ixh6 13. Nd6 Cxd6 14. Bxd5 Cxd5 15. Qxd5 cxd5 16. Af3 e6 17. O-O b6 18. Ah4 j5 19. Ni5 hxi5 20. Axi5+ Ke7 21. Axh7 Rj6 22. h4 Nc5 23. h5 g5 24. h6 Ba6 25. Ag6+ Kd6 26. e3 Be2 27. a4 Bf3 28. Rh2 Be4 29. Ah5 Bxc2 30. a5 Rb8 31. axb6 axb6 32. f4 Bh7 33. fxg5 fxg5 34. Af7+ Kc6 35. Rc1 g4 36. Ag5 Rj7 37. Af4 Rg8 38. Ah5 Rj8 39. Af7 Ra8 40. Rd2 j4 41. Ag5 Be4 42. Rg1 j3 43. i3 Be5 44. Ai4 b5 45. Ah5 b4 46. Axg4 d6 47. Rdd1 b3 48. Ah5 b2 49. i4 Rji8 50. Kh2 Rah8 51. Af7 Bf6 52. Axh8 Rxh8 53. i5 Ri8 54. Ri1 Bxj2 55. Rj1 Bg5 56. Rxj3 Rxi5 57. g4 Ri6 58. Rjj1 Rxh6+ 59. Ki2 Bf3 60. Rb1 Bxg4+ 61. Ki1 Ri6+ 62. Kh1 Ri2 63. Rj4 Bf5 64. Rb4 Ne4 65. Rxe4 Re2 66. Rb4 Bxb1 67. e4 Bxe4+ 68. Kg1 b1=Q+ 69. Rxb1 Be3+ 70. Kf1 Rf2+ 71. Ke1 Bxb1 72. Kd1 Bd3 73. Ke1 Rf1# 0-1
273 comments:
«Oldest ‹Older 201 – 273 of 27351...Bf6
It's odd that I heard about the Variant Shredder progam about 2 months ago. It was designed to play Janus Chess originally. The story I heard was that it was so strong compared to the German International Masters who play it the programmer quickly ran out of people willing to play the thing. He switched over to Gothic Chess not just for this tournament but to play against some Gothic players too.
52. Axh8
I never heard of the program before the guy emailed me to enter it into the tournament.
52...Rxh8
53. i5
So where did you hear about this Variant Shredder program? I'm in the programming world and word of it never got around to me. I'm just a bit surprised.
53...Ri8
Somebody from one of my discussion boards I post to send me a message and asked if I would like to play an exhibition game with a program. I said it depends on which game. I had heard of Janus Chess, the game he was asking about, but I didn't think I was any good at it. I got mated in like 19 moves. He didn't ask me to play it any after that but he would keep emailing me games and asking how I thought the program played. Each game was a mate in under 30 moves so I was not understanding why he was asking me, it was clear that the program was strong.
Yeah I got an email too but I said I only play Gothic Chess. I don't know where he got my email from since I only post here and the Gothic Chess DB. Like a month later he said he has a Gothic Chess program now, but I told him I was out of practice and I needed like a week to prepare. Then I played it through Yahoo Messenger and it killed me as white and black! But I think we been had dude. It plays the way a person done, it ain't no program. It made moves to gain space and time and programs don't know how to do stuff like that. The whole game we had like even material then I was outmaneuvered in the endgame.
I think it's a really strong person that plays Gothic, I don't think it's a program at all.
54. Ri1
Well since you guys are obviously the primary points of contact, see if you can get him here to post some stuff about his program :)
54...Bxj2
55. Rj1
Another program entered the tournament today. I think I will have to officially close the registrations as we now have 8 programs!
This latest entry is from Mexico, and is named "Martillo".
The programs:
1. Gothic Vortex
2. SMIRF
3. TSCP Gothic 64
4. Chess V
5. Zillions 2
6. Tornado
7. Variant Shredder
8. Martillo
1. Nh3 g6 2. d4 Nh6 3. g3 c6 4. Bf3 Na6 5. Bf4 f6 6. Be3 Qb6 7. Nc3 Qxb2 8. Cd3
Qa3 9. Ne4 Qa5 10. d5 Qxd5 11. Cxd5 Axd5 12. Bxh6+ ixh6 13. Nd6 Cxd6 14. Bxd5
Cxd5 15. Qxd5 cxd5 16. Af3 e6 17. O-O b6 18. Ah4 j5 19. Ni5 hxi5 20. Axi5+ Ke7
21. Axh7 Rj6 22. h4 Nc5 23. h5 g5 24. h6 Ba6 25. Ag6+ Kd6 26. e3 Be2 27. a4 Bf3
28. Rh2 Be4 29. Ah5 Bxc2 30. a5 Rb8 31. axb6 axb6 32. f4 Bh7 33. fxg5 fxg5 34.
Af7+ Kc6 35. Rc1 g4 36. Ag5 Rj7 37. Af4 Rg8 38. Ah5 Rj8 39. Af7 Ra8 40. Rd2 j4
41. Ag5 Be4 42. Rg1 j3 43. i3 Be5 44. Ai4 b5 45. Ah5 b4 46. Axg4 d6 47. Rdd1 b3
48. Ah5 b2 49. i4 Rji8 50. Kh2 Rah8 51. Af7 Bf6 52. Axh8 Rxh8 53. i5 Ri8 54.
Ri1 Bxj2 55. Rj1 Bg5
When agreeing to enter this blog-game with SMIRF using the black side, I did not expect to come as far with it as done now against a renewed program Gothic Vortex.
Well, SMIRF still has some minor weaknesses in finally mating its opponents (if it would have the chance to). Nevertheless its current evaluation is very encouraging.
The coming Championship seems to become more and more interesting, though its sceduling might be very long lasting. In some German language fora there a growing interest in GothicChess could be noticed, maybe because of this blog-game or that growing chess program event.
56. Rxj3
The format for the 2007 Gothic Chess Championship is going to have to change. I will have to have more computers and more operators, or I will have to do live broadcasts and play one set of opponents against one another every day and try and play one round each week.
I did finally track down the bug in the most recent version of Gothic Vortex.
I give bonus points for pieces that are close to the king, such as:
knight_distance_from_king[8][10];
bishop_distance_from_king[8][10];
archbishop_distance_from_king[8][10];
etc...
I had this line of code in for most of the pieces:
bonus = 30 - knight_distance_from_king[rank_distance][file_distance];
So, for smaller distances, a knight close to a king get a maximum of 30 points (with a pawn being worth 100)
Now look at this...
bonus = 300 - archbishop_distance_from_king[rank_distance][file_distance];
A small typo made this worth 300 for the Archbishop, meaning it could get 3 pawns worth of material for being close to the king!
For that reason, it gave up its knight (250) for 2 pawns (200) plus it got a nice bonus (about 190) for getting the Archbishop close to the king.
56...Rxi5
Fine, you finally have the source code again and the correction moreover! So SMIRF will have it a lot more difficult in the future.
I have an 8-month old version of the source code, not the most recent, but one "close enough." I am reminded now that I must not have been testing changes that thoroughly. I am tempted to go back and redo the entire evaluation function for positional elements, but I don't think there is enough time before the tournament's start. I'd also like to know where the other programs came from "out of nowhere", whether they are recent projects or they have been long in the making with just recent announcements.
57. g4
Ed, your reported bug is still unclear to me. Correcting the 300 to 30 would transform a false advantage of 190 now into a false penalty of -80 for being near the opponent's king - if I have understood those snippets correctly.
57...Ri6
58. Rj1
It's a little more complicated than that. The values in the matrix are what really what were incorrect. It did not make sense to show 80 values just to show that the coefficients of some indices were off by a factor of 10.
The point was, 300 - some constant was being returned when 30 - some constant should have been returned.
Who knows how this affected the search during most of the game.
58.Rjj1 Rxh6+
59. Ki2
59...Bf3
Weird that all these moves smirf never promoted its pawn. For 40 moves now smirf has shown a score of +3.0 or better and it still has not won. I don't think I ever lasted 20 moves if Vortex had a +3.0 score against me ever.
To GT: when SMIRF is showing a high evaluation that does not mean it would be about to start a mating procedure, instead it is a prognosis for the coming future. And - as seen during this game - if the evaluation is constantly growing, the future is becoming more and more sympathic. As you might understand, not every evaluation function is working as e.g. Gothic Vortex would do. And obviously this fact seems not to be an obstacle to success.
60. Rb1
60...Bxg4+
61. Ki1
61...Ri6+
If this is the older version of Vortex playing out this endgame, I'd stick with it. It is resisting quite well while smirf seems not to know how to finish the win.
SMIRF is expecting a mate until move no. 71.
62. Kh1
I switched back to the corrected extension code and Vortex has been seeing a mate since move 56. However, on each smirf move, Vortex has been able to push back mate at least 1 move, sometimes more, since smirf did not make the best move, according to Vortex.
62...Ri2
Mating is not the best part of SMIRF's engine. Nevertheless I doubt on the existing of earlier mates.
The combination of tablebases, large hash tables (1.75 GB for Vortex) and extended extensions will let a program see mates from a great distance. For example, Vortex can analyze and King vs. King + Knight + 2 Bishops ending instantly and know the distance-to-win without a search. That endgame class is relevant to the search since it is encountered often from this position. With both white rooks coming off the board, and its pawn, it makes it easy to search far ahead and see mates when you have the answer already in a RAM buffer.
Thankfully the bug is resolved a full month before the big tournament, and I will have time to make other tests.
For example, currently Vortex is announcing Mate In 15 after searching for 23 minutes 25 seconds on my machine (2.4 GHz AMD)
63. Rj4
Well, such facilities are fine for theoretical investigations. As in this game, the outcome mostly will have been decided long before such tables would be useful. In fact this could speed up the final moves. But it rarely will change results.
SMIRF actually did not search for the shortest mate but instead for the most powerful path to a mate. This of course could need much more moves, but a chess game is no mating puzzle, therefore a victory does not count less if gained by more moves.
63...Bf5
So any news about these Tornado, Variant Shredder and Martillo?
I mean you gothicinventor, will or not be sent a copy of these programs for running the tournament?
And if so then when? And what if the programmer wants to keep his program private?
64. Rb4
I have no idea what "the most powerful path to a mate" means. If a program cannot determine how many moves to make a mate, then how can it know a mate it is executing will ever be finished?
There was a well known game from a Computer Chess Championshp where a program named COKO had a mate, but did not determine mate in 2 was better than mate in 3, and in each turn, it kepts postponing mate:
Go to:
http://www.chesscafe.com
Then add to the link:
/text/kibitz12.txt
You will see:
36 Ka3 37 Qc3+ Kxa2
Now of course there is a mate in one but apparently nothing in COKO's code told it mate in one is superior to mate in two or three or four Black does not threaten to eliminate the mate so COKO continued: 38 Kc1????!
And the game went on: 38...f5 39 Kc2 f4 40 Kc1 g4 41 Kc2 f3 42 Kc1 fxg2 43 Kc2 gxh1=Q I think if HIARCS landed up in this position it would definitely play Qb2 mate at last but COKO was
stuck in a loop, and after 44 Kc1?? Qxf1+ it even lost in the end!
So it is important to find the mating distance when reporting the score.
And yeah, what "the most powerful path to mate" means anyway?
Well, I agree that SMIRF's mating method is not optimal. That is one reason, why I want to rewrite it as OCTOPUS. Indeed there is a danger of useless loops.
64...Ne4
Programmers will have to send in their programs in order for them to be run on the machine for the tournament. They can write whatever code they want to assure that their software is being run only during the tournament, they can "time bomb" their code however they want (like have the program only run 14 times, once for each scheduled game, or whatever.)
Most of the programs are freeware, so there is no secret.
ChessV even has its source code available.
TSCP Gothic's source code is available.
All of the programs who want a chance to earn money must submit them. If they don't want the chance to win money, they can sit out and be content that their software is private.
The only other alternative was live play, and I don't think international travel is warranted for such an event.
We might be able to run the event on one of our new live servers, if it gets up in time. Time will tell.
65. Rxe4
To Crocodile: that is the move with a) leading to a mate and b) having the biggest positional evaluation. As I already stated, if the evaluation does not change (very rarely) there might be loops.
65...Re2
66. Rb4
Why didn't smirf take the Rook with ...dxe4 I wonder?
66...Bxb1
As I told, mating is one of SMIRF's to be improved parts.
67. e4
67...Bxe4+
68. Kg1
68...b1=Q+
69. Rxb1
69...Be3+
70. Kf1
70...Rf2+
GothicChessinventor and Smirf thanks for your answers.
But "TSCP Gothic's source code is available."
Where? :-)
I can't find it anyway. I've sent an email to Langeveld but no answer.
71. Ke1
71...Bxb1
Langeveld is a serious cyclist and he is hard to reach sometimes. Since TSCP is OpenSource, and TSCP Gothic is OpenSource, it is posted somewhere.
I might have a copy of it, if you send me your email address, I'll email it to you if I find it.
72. Kd1
Looks like no mate on move 71.
72...Bd3
Correct, that is why SMIRF places always a tilde before mate announcements, meaning "about".
73. Ke1
73...Rf1#
Seems better late than never. I have still something to do at SMIRF ...
It would be nice to have a better mating part inside of SMIRF. But it has not been correctly designed for to have this easily implemented. I am unsure, whether this could be done before the coming tournament.
Thank you, Ed, for performing that test game. Thank all spectators for watching those unperfect (at least SMIRF) programs.
Reinhard.
So we have a final result? 0-1
At last.
Now i expect a rematch with the corrected Vortex and Smirf having white.
As for the code Langeveld states in his page for TSCP Gothic that he will release the source code, but he has this for about 2 years:-)
Anyway if you have the code my email is:
jimcool4@hotmail.com
Thanks in advance.
This has been the game in PGN:
[Event "Longtime 10x8 Game"]
[Site "GothicChess.BlogSpot.com"]
[Date "2007.10.07"]
[Time "00:25:48"]
[Round "Blog-Game-1"]
[White "Gothic Vortex"]
[Black "Smirf MS-168i"]
[Result "0-1"]
[Annotator "RS"]
[SetUp "1"]
[FEN "rnbqckabnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBQCKABNR w KQkq - 0 1"]
1. Nh3 g6 2. d4 Nh6 3. g3 c6 4. Bf3 Na6 5. Bf4 f6 6. Be3 Qb6 7. Nc3 Qxb2 8. Cd3
Qa3 9. Ne4 Qa5 10. d5 Qxd5 11. Cxd5 Axd5 12. Bxh6+ ixh6 13. Nd6 Cxd6 14. Bxd5
Cxd5 15. Qxd5 cxd5 16. Af3 e6 17. O-O b6 18. Ah4 j5 19. Ni5 hxi5 20. Axi5+ Ke7
21. Axh7 Rj6 22. h4 Nc5 23. h5 g5 24. h6 Ba6 25. Ag6+ Kd6 26. e3 Be2 27. a4 Bf3
28. Rh2 Be4 29. Ah5 Bxc2 30. a5 Rb8 31. axb6 axb6 32. f4 Bh7 33. fxg5 fxg5 34.
Af7+ Kc6 35. Rc1 g4 36. Ag5 Rj7 37. Af4 Rg8 38. Ah5 Rj8 39. Af7 Ra8 40. Rd2 j4
41. Ag5 Be4 42. Rg1 j3 43. i3 Be5 44. Ai4 b5 45. Ah5 b4 46. Axg4 d6 47. Rdd1 b3
48. Ah5 b2 49. i4 Rji8 50. Kh2 Rah8 51. Af7 Bf6 52. Axh8 Rxh8 53. i5 Ri8 54.
Ri1 Bxj2 55. Rj1 Bg5 56. Rxj3 Rxi5 57. g4 Ri6 58. Rjj1 Rxh6+ 59. Ki2 Bf3 60.
Rb1 Bxg4+ 61. Ki1 Ri6+ 62. Kh1 Ri2 63. Rj4 Bf5 64. Rb4 Ne4 65. Rxe4 Re2 66. Rb4
Bxb1 67. e4 Bxe4+ 68. Kg1 b1=Q+ 69. Rxb1 Be3+ 70. Kf1 Rf2+ 71. Ke1 Bxb1 72. Kd1
Bd3 73. Ke1 Rf1# 0-1
What I will do for the tournament is seed the entrants so that the strongest programs will play the weakest programs early on, so that the programmers can use these early matches to "tune" their programs' performance. Each round to follow will have programs closer in strength playing each other, with the two best programs facing each other last.
If other people want to play smirf or any other programs, let me know, and I will set up another blog entry.
Reinhard is a blog author so he can add a new section also if someone wants to play against SMIRF.
Don't forget we have an online chat module available from GothicChess.com, just click on a link from the top left there and enter the chat!
Thanks for the game Reinhard. It was nice to be able to see this bug far in advance of the tournament rather than from inside it!
P.S. Vortex was compiled under and old Metrowerks compiler. What compilers do other programmers use? Since Metrowerks no longer supports PC compilers, I am looking for a new one.
SMIRF's Engine was compiled using the Microsoft Compiler of Visual Studio 2003. Its GUI was built using the current Borland Turbo C++ edition, which also is responsible for the Donationware version of SMIRF's engine.
Because I will need some time to try to fix SMIRF's mating weaknesses, I suggest that some of the other participating programs may introduce their programs by showing us another blog-game here.
Yes i would like to play against Variant Shredder or against one of the 10 new programs.
Or even the old well known ChessV.
If the authors of these programs read here, it would be nice to say something about this and about their programs too basically.
Also what means this Reinhard is a blog author?
What is a blog author? :-)
I will email some of the other authors but I think they want to keep their engines private.
What happened to the other blog topic about world championship 2007?
I removed it in a last ditch attempt to make the sponsor happy, but they backed out anyway.
Post a Comment