endgame db for facing anaother computer or a human body ?

[TAILLE]

Hi,

A weird question for any endgame db conceptor : do you think it could be interesting to have a db for facing another computer and a different db when facing a human ?

Normaly, anybody having the 6 pieces endgame db knows that the following position is a draw because the DTC for this position is 52 (greater or equal to 50 => draw according to teh 25 move rule).

1) Did you really take into account the 25 move rule ?
2) If yes did you store the DTC 52 in the DTC db though the position is marked as a draw ?
3) has somebody an idea of the propagation of such draw in the db. I guess the impact should be almost negligeable but I am not able to measure it.

For the time being Damy does not take into account such draw because with the 6 pieces db these positions are very exceptionnal.

The point now is is that with the 7 pieces db we obviously encounter far more examples like the one above so that the question of these draws becomes relevant, at least when facing another computer with DTC db !

When facing a human I wondering if it is not better to mark the position as a win !

What is your feeling ?
Gérard

[ildjarn]

Gerard,

Could you show the solution for this? I used Michel Grimminck's databases and in the main line, white plays 28-22 on the 23rd move, so just in time.

[TAILLE]

Gerard,

Could you show the solution for this? I used Michel Grimminck's databases and in the main line, white plays 28-22 on the 23rd move, so just in time.

Hi,
1.19-10 41-47 2.10-15 47-36 3.15-04 36-47 4.01-06 47-38 5.23-05 38-49 6.04-13 49-40 7.13-27 40-29 8.27-36 29-47 9.05-23 47-38 10.23-34 38-21 11.36-04 21-26 12.34-25 26-12 13.04-36 12-21 14.25-03 21-16 15.36-13 16-07 16.13-02 07-18 17.02-16 18-01 18.03-17 01-29 19.16-02 29-18 20.02-24 18-36 21.24-47 36-18 22.17-21 18-09 23.21-26 09-18 24.47-33 18-04 25.33-50 04-18 (draw due to the 25 moves rule!) 26.26-21 18-04 27.28-23
This kind of position is very useful to debug DTC db. Where do you improve white play ?
Gérard

[steenslag]

Hi,

A weird question for any endgame db conceptor : do you think it could be interesting to have a db for facing another computer and a different db when facing a human ?

(...)

Gérard

From the regulations of the last Kramnik-Fritz match:

ENDGAME TABLEBASES
The use of a database of endgame positions ("Tablebase") is permitted only if the tablebase contains positions with a total five total pieces or less, including kings.

When Deep Fritz identifies the board position in a tablebase, it must inform the Arbiter, who will then stop the clocks.

In the presence of the Arbiter, the Operator will inform Mr. Kramnik that the position has been located in the tablebase.

If the position is evaluated by the tablebase as winning for the side played by Deep Fritz, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. The game will continue, unless Mr. Kramnik chooses to resign.

If the position is evaluated by the tablebase as winning for the side played by Mr. Kramnik, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. The game will continue unless the Deep Fritz Operator chooses to resign.

If the position is evaluated by the tablebase as a draw, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. This will constitute an offer of a draw. The game will continue, unless the offer is accepted prior to the completion of Mr. Kramnik's next move.

It is recognized that the program will access tablebases in its calculations. The above rules apply only when the position on the board is present in the tablebase.

(from http://www.chesspro.ru/_events/2006/fritz3_en.html )

These regulations look reasonable to me. There is no real point in playing against a perfect playing machine, from a sports point of view.

Under these regulations, a relatively small database (for instance 6 pieces) could be better for the computer player than a larger database, leaving the human player more opportunities to missplay the larger positions.

[mschribr]

Hi,

A weird question for any endgame db conceptor : do you think it could be interesting to have a db for facing another computer and a different db when facing a human ?

(...)

Gérard

From the regulations of the last Kramnik-Fritz match:

ENDGAME TABLEBASES
The use of a database of endgame positions ("Tablebase") is permitted only if the tablebase contains positions with a total five total pieces or less, including kings.

When Deep Fritz identifies the board position in a tablebase, it must inform the Arbiter, who will then stop the clocks.

In the presence of the Arbiter, the Operator will inform Mr. Kramnik that the position has been located in the tablebase.

If the position is evaluated by the tablebase as winning for the side played by Deep Fritz, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. The game will continue, unless Mr. Kramnik chooses to resign.

If the position is evaluated by the tablebase as winning for the side played by Mr. Kramnik, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. The game will continue unless the Deep Fritz Operator chooses to resign.

If the position is evaluated by the tablebase as a draw, the Operator will inform Mr. Kramnik of that fact in the presence of the Arbiter. This will constitute an offer of a draw. The game will continue, unless the offer is accepted prior to the completion of Mr. Kramnik's next move.

It is recognized that the program will access tablebases in its calculations. The above rules apply only when the position on the board is present in the tablebase.

(from http://www.chesspro.ru/_events/2006/fritz3_en.html )

These regulations look reasonable to me. There is no real point in playing against a perfect playing machine, from a sports point of view.

Under these regulations, a relatively small database (for instance 6 pieces) could be better for the computer player than a larger database, leaving the human player more opportunities to missplay the larger positions.

Taille’s question was if the information in the database should be different depending on if the opponent is human or computer. Steenslag’s answer talks about the size of the database 6 or 7 pieces, not the content.

With the Kramnik-Fritz match rules it would still benefit the computer with a larger database. If the computer finds a win for the human in the database then the human must play perfectly to win. While the computer has the database and does not need to compute the moves that will delay the win with the longest sequence of moves. This means the human may not find the winning moves and instead draw the game. While a smaller database will force the computer to compute the best moves letting the human have an easier time finding the winning moves.

What about the other rules of the Kramnik Fritz match? Should the other rules also apply to a computer vs human draughts match? Here are some of those rules. 1) When the computer is in its opening book of moves the human is allowed to see the computer’s screen. The screen showing all moves, including all statistics (number of games, performance, score) from grandmaster games and the move weighting of the computer. 2) 2 months before the match the human gets the same program running on the same computer as during the match. For 2 months before the match the human can use a different program to find the computer’s flaws. 3) The human can adjourn the game for the night. The human’s assistants can try moves in his room with the identical copy of the program given to him running on an identical computer given to him. They can do this the whole night until they find a flaw in the computer. Then the next day the human wakes up and replays the moves found for him. With all this help the world champion still lost the match. If the human needs this much help to win then forget the match. We already know who is better. Mark Schreiber