logo Welcome, Guest. Please Login or Register.
October 15, 2019, 04:55AM
Home Help Search Login Register
News: Visit a member's site: http://www.boris-is-king.com (plozikou)

  Chess Computer Community
  General Category
  Discussions about Chess Computers
(Moderators: Endspielgott, Overtom, bobosse)
  Phantom 6100 end game bug?
« previous next »
Pages: [1] Reply Notify of replies Send the topic Print 
   Author  Topic: Phantom 6100 end game bug?  (Read 933 times)
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« on: December 1, 2005, 02:17AM »
Reply with quote

Robert,

Quote from: Robert Weck on November 30, 2005, 11:44PM   

{........}
If you still want it, i can look after it...

Robert

Thanks for the offer!

Yes, I would be very much obliged. I find it hard to test the computer if I don't know exactly what the levels stand for.

Best regards,

tom
Report to moderator Logged
Robert Weck
Hero Member
*****

Offline

Posts: 791



I love Boardnation!

View Profile
Re:Phantom 6100 end game bug?
« Reply #1 on: November 30, 2005, 11:44PM »
Reply with quote


Quote from: Overtom on November 30, 2005, 02:05AM   

Would you by any chance have the manual of the Mystery? I have the computer here, but never managed to get a manual.

I'm not sure, if i have it, but i never missed it, because the program is absolutely identical with the Centurion and its "clones"!

If you still want it, i can look after it...


Robert
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #2 on: November 30, 2005, 02:05AM »
Reply with quote


Quote from: Robert Weck on November 23, 2005, 07:00AM   

And even more unbelievable: yesterday evening, my Cosmos played a game against his bigger brother Mystery and i think, it ended with the same bug!!!!

Hello Robert,

Would you by any chance have the manual of the Mystery? I have the computer here, but never managed to get a manual.

Thanks in advance,

tom
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #3 on: November 24, 2005, 07:53AM »
Reply with quote


Quote from: Robert Weck on November 24, 2005, 02:50AM   

... I saw a sign: "Würstchen mit Weck 2,-- DM"

It means: Sausage with roll!

Never heard this before... 


Robert


Ah - my command of the German language is getting better and better ...

So now I know "ein Weck im Heu" means "a roll in the hay"

Best regards,

tom
Report to moderator Logged
Robert Weck
Hero Member
*****

Offline

Posts: 791



I love Boardnation!

View Profile
Re:Phantom 6100 end game bug?
« Reply #4 on: November 24, 2005, 02:50AM »
Reply with quote


Quote from: Overtom on November 24, 2005, 02:32AM   

By the way, I liked your quotation! Never realized you were German. Here in the Netherlands, your surname is mainly associated with appliances for the preservation of fruits and vegetables. After all, the ending "ck" might as well be German as English.

About 10 years ago, when i was in Riegelsberg (a town in the western part of Germany near the border to France) i saw a sign: "Würstchen mit Weck 2,-- DM"

It means: Sausage with roll!

Never heard this before... 


Robert
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #5 on: November 24, 2005, 02:32AM »
Reply with quote


Quote from: Robert Weck on November 24, 2005, 01:33AM   

Hi Tom,

If it was a human, i would have thought, it was an act of desperation, but from a computer? 

But maybe this is an explanation:
Here in Germany we say: "Lieber ein Ende mit Schrecken als ein Schrecken ohne Ende!"

You could translate it with: Better an end with horror than a horror without an end! 

I find it most mysterious, that they play this move absolut immediately even on "infinite"!!

Robert


Well, Im not sure whether this can really be called the "horizon effect", but from the point of view of the computer it's not strange that it opts for a loss with a smaller difference in points. After all, numerical evaluation is what automated chess-playing is all about.

I sometimes try to explain decisions made by chess computers from such an alien point-of-view; I still cherish the illusion that one day I'll be able to complete my Java chess program (overtom.nl/TomChess.html) in such a way that it can beat me - which is not such a terrible achievement taking my very limited playing capabilities into account

By the way, I liked your quotation! Never realized you were German. Here in the Netherlands, your surname is mainly associated with appliances for the preservation of fruits and vegetables. After all, the ending "ck" might as well be German as English.

Best regards,

tom
Report to moderator Logged
Robert Weck
Hero Member
*****

Offline

Posts: 791



I love Boardnation!

View Profile
Re:Phantom 6100 end game bug?
« Reply #6 on: November 24, 2005, 01:33AM »
Reply with quote

Hi Tom,


Quote from: Overtom on November 23, 2005, 01:05PM   

Just imagine you're a chess computer  and try to figure out why:

In a few moves, one of the white pawns will be promoted, so white would then have a pawn, a rook and a queen (about 15 or 16 points), against black only a rook (5 points). So the difference would be about 10 points.

Now that the black rook captures one pawn, the promotion is shifted behind the horizon and white is left with one rook and one pawn (6 points). So for the near future, the difference between white and black has become less.

Does make sense, doesn't it?


If it was a human, i would have thought, it was an act of desperation, but from a computer? 

But maybe this is an explanation:
Here in Germany we say: "Lieber ein Ende mit Schrecken als ein Schrecken ohne Ende!"

You could translate it with: Better an end with horror than a horror without an end! 


I find it most mysterious, that they play this move absolut immediately even on "infinite"!!


Robert
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #7 on: November 23, 2005, 01:05PM »
Reply with quote

Hi Robert,

Just imagine you're a chess computer  and try to figure out why:

In a few moves, one of the white pawns will be promoted, so white would then have a pawn, a rook and a queen (about 15 or 16 points), against black only a rook (5 points). So the difference would be about 10 points.

Now that the black rook captures one pawn, the promotion is shifted behind the horizon and white is left with one rook and one pawn (6 points). So for the near future, the difference between white and black has become less.

Does make sense, doesn't it?

Best regards,

tom
Report to moderator Logged
Robert Weck
Hero Member
*****

Offline

Posts: 791



I love Boardnation!

View Profile
Re:Phantom 6100 end game bug?
« Reply #8 on: November 23, 2005, 07:00AM »
Reply with quote

Hi Tom,

Quote from: Overtom on November 17, 2005, 02:43AM   


If you care to see REAL bugs, then here are two of them:

Some time ago, I tested Saitek/Mephisto's Chess Challenger. In a game against Mephisto MMV, this position was reached:


Meph Ch Ch - Meph MMV (white to move)

Needless to say white could have continued 38 Rd1-d7+ Kb7-a6, 39 a2-a4 ... and mate by Rd7-a7+ is unavoidable.

But white played 38 Rh8xh6 ... and the game ended in a draw.

really a mysterious bug: even on "infinite", my Cosmos plays also Txh6 immediately; but when you put the rook on h7, then it plays correct!!

And even more unbelievable: yesterday evening, my Cosmos played a game against his bigger brother Mystery and i think, it ended with the same bug!!!!


[Event "?"]
[Site "?"]
[Date "2005.11.22"]
[Round "?"]
[White "Mystery, 30m"]
[Black "Cosmos, 30m"]
[Result "1-0"]
[ECO "A34"]
[Annotator "Weck,Robert"]
[PlyCount "107"]
[EventType "game"]

1. c4 c5 2. Nc3 Nc6 3. g3 Nf6 4. Nf3 d5 5. cxd5 Nxd5 {*} 6. Bg2 {*} e6 7. O-O
Be7 8. Nxd5 {0,0} exd5 {0,0} 9. d4 cxd4 10. Nxd4 O-O 11. Be3 Bf6 12. Rc1 Qa5
13. a3 Re8 14. Rc2 Qa6 15. Bxd5 Nxd4 16. Bxd4 Bxd4 17. Qxd4 Be6 18. Bxe6 Rxe6
19. e3 Rd6 20. Qf4 Rd7 21. Rfc1 Rad8 22. Rc8 Rd1+ 23. Rxd1 Rxc8 {1,7} 24. Qc7 {
1,6} Rf8 25. Rd7 h6 26. Qxb7 Qxb7 27. Rxb7 Ra8 28. b3 {1,7} a6 {1,7} 29. Kg2 g6
30. a4 Kg7 31. b4 Re8 32. Kf3 Kf6 33. b5 {2,0} axb5 {2,0} 34. Rxb5 Rd8 35. a5
Rd1 36. Kf4 Ke6 37. f3 Ra1 38. e4 Kd6 39. Rb7 Rxa5 40. Rxf7 Ke6 41. Rh7 h5 42.
h4 Kf6 43. g4 hxg4 44. fxg4 Ke6 45. e5 {0,6} Rxe5 {2,0} 46. Rh6 {2,1} Kd5 47.
Rxg6 {2,2} Re4+ {2,0} 48. Kf3 Re1 49. h5 Rf1+ 50. Kg2 Rf8 {2,4} 51. h6 {2,4}
Kc5 52. Rg7 Rh8 53. g5 Rxh6 54. gxh6 1-0

the 53rd black move has been played immediately and even on "infinite". And needless to say, that the Mystery played the same (it has the same program, but running at 12 MHz instead of 10)

The situation was not so clear like in your match, because the game was already lost for black, but normally it was to early to sacrifice the rook! 


Robert
Report to moderator Logged
Mike Watters
Hero Member
*****

Offline

Posts: 1605



Another day, another blunder.

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #9 on: November 18, 2005, 01:23PM »
Reply with quote

Tom

You have found some good bugs there. The Mephisto Chess Challenger one is almost as if it suddenly thought the black pawns were going in the opposite direction.

We have a few bugs with the Resurrection. One amusing one I described on the Resurrection thread. A black king on h8 took a white pawn on h7 promoting itself to a white piece on h7 in the process.  This was a handicap game requiring use of the set-up board function at the beginning so it is not representative but funny nevertheless.

Ismenio, the Par Excellence manual says the same things about draws as the Phantom manual. I guess this lack of knowledge of chess rules is not really noticed unless you have the Phantom on Autoplay. It is normal in chess play to recognise a dead draw before it happens and you just stop playing. I shall look out to see what other chess computers of that era do with draws. As Asiryangguk says many will probably fail even on the rules most chess players know.

Mike
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #10 on: November 17, 2005, 02:43AM »
Reply with quote

Mike,

I suppose this is the programmer's choice. After all, if you have played against a Novag Amber in such a way that only two kings are left on the board, you must be a fairly strong chess player and are bound to know a game with two kings is undecided.

Even 32K is a limited amount of memory. Taking up all possible positions that cannot be won will probably mean a (slightly) lower ELO rating. And after a certain number of useless moves, the computer will claim a draw anyway because no pieces have been captured or pawns moved.

So I tend to disagree with Ismenio's suggestion that this is a bug.

If you care to see REAL bugs, then here are two of them:

Some time ago, I tested Saitek/Mephisto's Chess Challenger. In a game against Mephisto MMV, this position was reached:


Meph Ch Ch - Meph MMV (white to move)

Needless to say white could have continued 38 Rd1-d7+ Kb7-a6, 39 a2-a4 ... and mate by Rd7-a7+ is unavoidable.

But white played 38 Rh8xh6 ... and the game ended in a draw.

Another unexpected ending (which also involved a Saitek/Mephisto machine) was in a game between two portables: Mephisto Maestro and Excalibur Escort. After 35 moves, this was the position:


Meph Maestro - Exc Escort (white to move)

Here, almost any continuation would have won the game for white, for instance: 36. g8Q Ke1 37. Rg1+ mate. But what did Maestro play?

36. Kd2 stalemate!!


I described these two games in some more detail at:

  overtom.nl/weblog/13FEB05.html
  overtom.nl/weblog/13NOV04.html

Buggy regards,

tom
Report to moderator Logged
Ismenio
Horsie to pointy guy 6
Administrator
*****

Offline

Posts: 1753





View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #11 on: November 16, 2005, 03:48PM »
Reply with quote

Thanks folks for your comments. I find it interesting that the programs would not recognize this type of draw based on the fact that it is very easy to check that you only have the two kings on the board 

I should have checked the manual! That will teach me!



Ismenio
Report to moderator Logged

Map of chess computer friends: http://www.frappr.com/chesscomputers
Mike Watters
Hero Member
*****

Offline

Posts: 1605



Another day, another blunder.

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #12 on: November 16, 2005, 03:32PM »
Reply with quote

Tom

That is a funny game. It is programming quirks like that which make stopping any computer game early and assuming the result a bit problematic. Did you see Martin's Mephisto III game. In that case the Mephisto was so keen on pushing passed pawns it did not spot a load of Mates in 2 and 3. Your game is almost the opposite with the Novag drastically undervalueing its passed pawns.

Ismenio the manual says that the Phantom can recognize and claim three different types of draws: draw by stalemate, draw by the 50-move rule, and draw by three fold repetition. So presumably it has not been programmed to recognise draw by insufficient material to deliver checkmate. You would have thought that a 1988 machine would have covered that.  I think it is the same "2100" program as the Par Excellence, Designer 2100 etc. I will check the Par Excellence to see if it is the same.

Mike
Report to moderator Logged
Overtom
Moderator
Co-admin
Hero Member
*****

Offline

Posts: 628



"Life is too short for Chess" (Lord Byron)

View Profile WWW E-Mail
Re:Phantom 6100 end game bug?
« Reply #13 on: November 15, 2005, 06:44PM »
Reply with quote

Ismenio,

I admit programs like Fritz claim a draw when they have insufficient material to mate their opponent.

But I wonder if you have often seen any dedicated computers claim a draw under these conditions.

But things can even be worse:

Yesterday, I was testing two quite strong handheld computers of the Overtom collection: Kasparov Travel Champion 2080 and Novag Amber, both computers that will beat more than 95% of human players.

Since the Novag computer has much faster hardware than the Kasparov machine, I gave the latter twice as much time as the Novag computer.

In one of the games, this was the position after black's 78th move:


Novag Amber - Kasparov Travel Champion

After 79. Qxf4+ Kg2, white could have won with 80. Qxh2+ Kxh2, and white has no problem promoting his queen side pawns.

But what happened? 80. Qg4+ Kf2 81. Qf5+ Kg1 82. Qg6+ Kf2 83. Qe4 Kg1 84. Qb1+ Kg2 85. Qc2+ Kg1 86. Qd1+ Kg2 87. Qd5+ Kg1 88. Qg8+ Kf2 89. Qh7 Kg1 90. Qg7+ Kf1 91. Qf8+ Kg2 92. Qg8+ Kf2 93. Qd5 Kg1 94. Qc5+ Kg2 95. Qg5+ Kh1 96. Qd5+ Kg1 1/2-1/2

Can you imagine my infinite surprise?

With flabbergasted regards,

tom
Report to moderator Logged
Asiryangguk
Newbie
*

Offline

Posts: 34

To beat a chess computer, unplug it.

View Profile
Re:Phantom 6100 end game bug?
« Reply #14 on: November 15, 2005, 02:28PM »
Reply with quote

Dear Mr. Sousa,

Nowadays one only expects Novag Computers to come even close (and not completely even then) to the exhaustive and manifold conditions that lead to a legal draw in the game of chess, conditions that FIDE has chosen to extrapolate on for many years.  Very few people in the world know all the conditions that lead to a draw when considering these FIDE extrapolations.  It should not shock most people that computers may not know all these draws, albeit it seems surprising that the Fidelity 6100 overlooked this simplest of stalemates.

If you'd like to further experiment, set up a King, Rook, and Bishop (no pawns) against a King and Rook (no pawns) and watch the computer play itself.  Barring a capture or a checkmate, the computer should declare a draw NOT by the fifty-move-rule, but by the seventy-five-move-rule, a condition set by the otherwise obscure FIDE Congress at Salonika in November 1988.  I would totally understand if you began laughing, and no one can truly expect a programmer, except possibly President Kirsan himself, to put in all those obscure situations into some computer.

The simplest solutions about why some computers may not have exhaustive knowledge of legal drawing situations lies in how it appears cheaper to produce and sell them that way.

I hope this helps (smiles)!
Report to moderator Logged
Ismenio
Horsie to pointy guy 6
Administrator
*****

Offline

Posts: 1753





View Profile WWW E-Mail
Phantom 6100 end game bug?
« Reply #15 on: November 14, 2005, 04:46AM »
Reply with quote

Hi all,

I was showing the Phantom 6100 to some chess players this weekend and in one of the games where the Phantom was playing against itself I ran into something I had never seen before.

In the end, we had only the 2 kings and 1 rook. The white king took the black rook but the Phantom kept on playing! 
I don't recall ever seeing an end game with only the two kings on the board. I stopped the game I think after a couple of moves as I was getting ready to pack and go.

Has anyone else seen it? Is this a normal behaviour?

Thanks,

Ismenio
Report to moderator Logged

Map of chess computer friends: http://www.frappr.com/chesscomputers
Pages: [1] Reply Notify of replies Send the topic Print 
« previous next »
Jump to:  

Login with username, password and session length

Chess Computer Community
Get your Ad Free Message Board at Boardnation!   © 2013 Boardnation  Web hosting by Nubis