Author: Matthew Herman
Date: 08:05:46 01/05/99
Go up one level in this thread
On January 04, 1999 at 20:44:38, Dann Corbit wrote: >On January 04, 1999 at 20:09:15, Eugene Nalimov wrote: > >>On January 04, 1999 at 19:52:21, Dann Corbit wrote: >> >>>After having stepped through these end-games, I think that the problem is the >>>horizon effect. The draws are so far away, that they cannot be seen. Which >>>brings up an interesting question: >>>Do endgame tablebase files contain draw data *as well as* checkmate data, so >>>that when you are behind a draw could be discovered? This seems (to my feeble >>>way of thinking) to be just as valuable as checkmate data. If not, it seems >>>worthwhile to construct draw endgame tablebase files just as construction of >>>checkmate database files is valuable. >> >>Based on my understanding, any program that has access to >>the corresponding TB must declare those positions as draws >>(I myself have no chess program here, so I cannot check my >>statement). >After turning logging on, I see that Crafty did see that the position was drawn, >but is "trying for swindle": > >White(1): setboard 8/6kP/5N2/5K2/8/8/8/8 w - - >White(1): go >end-game phase > clearing hash tables >Drawn at root, trying for swindle. > time surplus 0.00 time limit 12:50 (12:50) > depth time score variation (1) > 6-> 0.07 6.91 1. Ke6 Kh8 2. Kd6 Kg7 3. Ke5 Kh8 > 7 0.07 6.91 1. Ke6 Kh8 2. Kd6 Kg7 3. Ke5 Kh8 4. > Ke6 > 7-> 0.14 6.91 1. Ke6 Kh8 2. Kd6 Kg7 3. Ke5 Kh8 4. > Ke6 > 8 0.16 6.99 1. Ke6 Kh8 2. Kd6 Kg7 3. Ke5 Kh8 4. >etc. >also: >White(1): setboard 5k2/6r1/4Q3/8/8/8/8/7K b - - > puzzling over a move to ponder. >end-game phase > clearing hash tables > depth time score variation (1) >Black(1): Rh7+ [pondering] > clearing hash tables >Drawn at root, trying for swindle. > time surplus 0.00 time limit 12:50 (12:50) > depth time score variation (1) > 6-> 0.06 4.60 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 > 7 0.13 4.60 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 > 7-> 0.14 4.60 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 > 8 0.31 4.60 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 6. Qh8 > 8-> 0.41 4.60 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 6. Qh8 > 9 1.13 4.65 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 6. Qf7+ Kd6 7. Qg8 > 9-> 1.22 4.65 2. Kg2 Re7 3. Qf6+ Ke8 4. Kg3 Kd7 5. > Kf4 Re6 6. Qf7+ Kd6 7. Qg8 Here is another of one of those positions 5k1K/7P/8/8/8/8/8/N7 w - - It is drawn as white (with the knight) cannot "lose a tempo" and drive the king away from f7/f8. (with black to move or the knight on a2 it is a win). Programs don't see this either (Junior at depth 60 says +6.96) Also as Hyatt said.. if it is in the tablebase it knows if it is drawn or a win .. and will say drawn at root etc.. (this is what fouled me up when using crafty under the winboard adaptor to get access to the tablebases via chessbase (it wouldn't show = 0.00 at the drawn positions but rather say +2.00 (in a KRKB drawn ending))
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.