longest 6-man EGTB win

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
petero2
Posts: 610
Joined: Mon Apr 19, 2010 5:07 pm
Location: Sweden
Contact:

Re: longest 6-man EGTB win

Post by petero2 » Sun Oct 27, 2019 12:59 am

syzygy wrote:
Sat Oct 26, 2019 11:01 pm
Dann Corbit wrote:
Thu Oct 24, 2019 10:22 pm
Zenmastur wrote:
Thu Oct 24, 2019 2:38 pm
Anyone have a list of positions of the longest 6-man EGTB wins that meet 50 move rule requirements? I.E. no cursed wins allowed
I think to do this accurately, you must have the complete Lomonosov files.
They have distance to mate.
But they don't take into account the 50-move rule. Whenever they show a mate in more than 50 moves, you can't really be sure about the number of moves it would take under the 50-move rule (it could be more, it could be less).
How could it be less?

If the winning side can force mate in N moves under the 50 move rule, there must exist a proof tree demonstrating that win.

In this proof tree all terminal nodes are at most N moves away from the root and have a score that is "mate" for the winning side.

A (correct) draw claim by the 50 move rule would correspond to a terminal node with a none-mate score, so there can be no draw claims in the proof tree.

In a game without the 50 move rule, the proof tree is a valid game tree since it does not contain any draw claims by the 50 move rule.

In a game without the 50 move rule, this game tree contains all valid moves for the losing side in all positions, since in the game without the 50 move rule, the set of legal moves is always the same or one less than the set of legal moves in the game with the 50 move rule (depending on if the "claim draw" move is valid or not in the game with the 50 move rule).

Therefore the game tree is a proof tree also for the game without the 50 move rule and since all terminal moves are at most N moves away from the root, this proves that it is possible to force mate in at most N moves.


A shorter but maybe less formal proof goes like this:

Assume there is a forced mate in N moves under the 50 move rule. If you remove the 50 move rule, two things happen:

1. The losing side loses the ability to claim a draw in some positions. Taking away options for the losing side cannot give him the ability to defend better. (compare relaxation in mathematical optimization)

2. The winning side loses the ability to claim a draw in some positions, but claiming a draw never wins the game, so losing the ability to claim a draw does not hurt the winning side.

Zenmastur
Posts: 908
Joined: Sat May 31, 2014 6:28 am

Re: longest 6-man EGTB win

Post by Zenmastur » Sun Oct 27, 2019 8:13 am

Seems like a gaping hole in the statistics for 6-man EGTB's.
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

syzygy
Posts: 4719
Joined: Tue Feb 28, 2012 10:56 pm

Re: longest 6-man EGTB win

Post by syzygy » Sun Oct 27, 2019 1:46 pm

petero2 wrote:
Sun Oct 27, 2019 12:59 am
syzygy wrote:
Sat Oct 26, 2019 11:01 pm
Dann Corbit wrote:
Thu Oct 24, 2019 10:22 pm
Zenmastur wrote:
Thu Oct 24, 2019 2:38 pm
Anyone have a list of positions of the longest 6-man EGTB wins that meet 50 move rule requirements? I.E. no cursed wins allowed
I think to do this accurately, you must have the complete Lomonosov files.
They have distance to mate.
But they don't take into account the 50-move rule. Whenever they show a mate in more than 50 moves, you can't really be sure about the number of moves it would take under the 50-move rule (it could be more, it could be less).
How could it be less?
You're right, it cannot be less. I did not think it through completely.

What could be less (or more) is the max DTM when respecting the 50-move rule (max DTM50). It could be more because a position can need more moves to be converted to mate when respecting the 50-move rule. It could be less because a position that can be converted without 50-move rule can be drawn with the 50-move rule (cursed win).

The WDL50 information can be used to detect the cursed wins, so combining Lomonosov/Nalimov's DTM information with Syzygy's WDL50 information can at least give a lower bound on max DTM50.

Paloma
Posts: 970
Joined: Thu Dec 25, 2008 8:07 pm

Re: longest 6-man EGTB win

Post by Paloma » Sun Oct 27, 2019 2:06 pm

It was once reported here that syzygy-DTM is under development.

How far is this project? Or has it been given up?

Does anyone know more? Ronald de Man?

Dann Corbit
Posts: 11506
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: longest 6-man EGTB win

Post by Dann Corbit » Mon Oct 28, 2019 6:03 am

The 50 move rule adds an incredible amount of complexity.
Therefore, it also adds an incredible amount of size to the files.
Every position can potentially have 100 different answers.
I think the problem is too difficult and/or too large.
Maybe noob has the hardware to deal with it, but even that is doubtful.
The worst part, by far, is how do you debug this thing?
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Zenmastur
Posts: 908
Joined: Sat May 31, 2014 6:28 am

Re: longest 6-man EGTB win

Post by Zenmastur » Mon Oct 28, 2019 6:45 am

Dann Corbit wrote:
Mon Oct 28, 2019 6:03 am
The 50 move rule adds an incredible amount of complexity.
Therefore, it also adds an incredible amount of size to the files.
Every position can potentially have 100 different answers.
I think the problem is too difficult and/or too large.
Maybe noob has the hardware to deal with it, but even that is doubtful.
The worst part, by far, is how do you debug this thing?
Hmmm...

There must be a communications issue.

I'm not asking that the files be changed in anyway. Just for a short, representative list, (1 from each of several EGTB files (e.g. 1 from the KQPkrr file for example) of positions which represent a long mate that respects the 50-move rule.

A similar list that doesn't respect the 50-move rule can be found here:http://kirill-kryukov.com/chess/longest ... ngth.shtml

I don't even need a full list. A half dozen example positions (hopefully from different endgame files) where the mate is, say 100 moves or longer, and it's not a cursed win would be more than sufficient I think.

I'm surprised no one has started such a list.

Regards,

Zenmastur
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

User avatar
hgm
Posts: 24730
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: longest 6-man EGTB win

Post by hgm » Mon Oct 28, 2019 10:16 am

The problem is that this information is not in any of the existing EGT. The DTM tables don't respect the 50-move rule, so that positions with large DTM might actually be draws. The DTZ tables do tell you which positions can be won respecting the 50-move rule, but they do not tell you their DTM.

duncan
Posts: 11765
Joined: Mon Jul 07, 2008 8:50 pm

Re: longest 6-man EGTB win

Post by duncan » Mon Oct 28, 2019 10:44 am

hgm wrote:
Mon Oct 28, 2019 10:16 am
The problem is that this information is not in any of the existing EGT. The DTM tables don't respect the 50-move rule, so that positions with large DTM might actually be draws. The DTZ tables do tell you which positions can be won respecting the 50-move rule, but they do not tell you their DTM.


I tried to look for the 10 longest DTM in descending order. The longest I could find was mate in 125. Does this look at all accurate to you. ?



duncan
Posts: 11765
Joined: Mon Jul 07, 2008 8:50 pm

Re: longest 6-man EGTB win

Post by duncan » Mon Oct 28, 2019 2:26 pm

duncan wrote:
Mon Oct 28, 2019 10:44 am
hgm wrote:
Mon Oct 28, 2019 10:16 am
The problem is that this information is not in any of the existing EGT. The DTM tables don't respect the 50-move rule, so that positions with large DTM might actually be draws. The DTZ tables do tell you which positions can be won respecting the 50-move rule, but they do not tell you their DTM.


I tried to look for the 10 longest DTM in descending order. The longest I could find was mate in 125. Does this look at all accurate to you. ?

Also there are no draws with the longest legal 6 piece endings.

Paloma
Posts: 970
Joined: Thu Dec 25, 2008 8:07 pm

Re: longest 6-man EGTB win

Post by Paloma » Mon Oct 28, 2019 3:09 pm

duncan wrote:
Mon Oct 28, 2019 10:44 am
I tried to look for the 10 longest DTM in descending order. The longest I could find was mate in 125. Does this look at all accurate to you. ?
Mate in 125 ? It is a mate in 68 !!

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/1q6/8/4Q3/8/8/B4n2/NK1k4 w - - 0 1"]
[Move_number "0"]
[Is_demo_mode "false"]

1. Bb3+ Kd2 2. Qb2+ Ke3 3. Nc2+ Kf4 4. Qf6+ Kg3 5. Nd4 Ng4 6. Qf5 Qh1+ 7. Kb2 Qh2+ 8. Ka3 Qd2 9. Qf3+ Kh4 10. Qh1+ Kg3 11. Qg1+ Kh3 12. Be6 Qc3+ 13. Ka4 Qf3 14. Qe1 Qf4 15. Qb4 Kh4 16. Ka5 Ne5 17. Ka6 Kg5 18. Qc5 Qe4 19. Bf5 Qa8+ 20. Kb5 Qb8+ 21. Ka4 Kf6 22. Be4 Kf7 23. Nf5 Qe8+ 24. Kb3 Kf6 25. Nd4 Qg8+ 26. Ka3 Qd8 27. Ka2 Qb8 28. Ka1 Qd8 29. Kb1 Qe7 30. Qb6+ Kg5 31. Ne6+ Kf6 32. Bd5 Qh7+ 33. Kb2 Qh2+ 34. Kb3 Qg3+ 35. Ka4 Nf7 36. Nd8+ Nd6 37. Nf7 Qg4+ 38. Ka3 Qd7 39. Kb4 Ke7 40. Qc5 Qb5+ 41. Qxb5 Nxb5 42. Kxb5 Kf6 43. Kc5 Kg6 44. Nd6 Kf6 45. Kd4 Kg6 46. Ke5 Kg7 47. Be4 Kg8 48. Kf6 Kf8 49. Nf7 Kg8 50. Bd3 Kf8 51. Bh7 Ke8 52. Ne5 Kd8 53. Be4 Ke8 54. Bd5 Kf8 55. Nd7+ Ke8 56. Ke6 Kd8 57. Kd6 Ke8 58. Bc4 Kd8 59. Bf7 Kc8 60. Nc5 Kd8 61. Nb7+ Kc8 62. Kc6 Kb8 63. Nd6 Ka8 64. Bc4 Kb8 65. Kb6 Ka8 66. Kc7 Ka7 67. Nc8+ Ka8 68. Bd5# 1-0

Post Reply