SYZYGY Base question

[IWB]

Sorry my fault. This is discussed after Ne3 in the CSS fora and I mixed things a bit.

BYe
Ingo

[IWB]

You probably wanted to enter a different PGN, this one is Shredder with Black again.

Yes, sorry, I am getting totaly confused:


[pgn][Event "1 Minutes/Game + 1 Seconds/Move"]
[Site "Engine Match"]
[Date "2014.01.19"]
[Round "1"]
[White "DS12 1T"]
[Black "Stockfish 180114 SYZ 1T"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "B7/8/8/8/5nk1/8/8/4K1B1 w - -"]

1. Kd2 {+M65/1 0s} Nh5 {+0.01/30 2s} 2. Bh2 {+M64/1 0s
(Bc5)} Ng3 {+0.01/33 2s} 3. Kd3 {+M63/1 0s} Kh3 {+0.48/27
0s} 4. Bg1 {+M62/1 0s} Kg4 {+0.48/27 0s} 5. Ke3 {+M61/1 0s
(Bc6)} Kg5 {+0.48/1 0s} 6. Bc6 {+M60/1 0s} Kf6 {+0.48/32
1s} 7. Bf3 {+M59/1 0s (Bd7)} Kg5 {+0.48/31 2s} 8. Bd1
{+M58/1 0s (Bg2)} Nh5 {+0.48/34 1s} 9. Kd4 {+M57/1 0s} Kg6
{+0.48/39 1s} 10. Ke5 {+M56/1 0s} Ng7 {+0.48/38 0s}
11. Bc2+ {+M55/1 0s} Kf7 {+0.48/38 0s} 12. Bd4 {+M54/1 0s
(Bc5)} Ne8 {+0.48/1 0s} 13. Kf4 {+M53/1 0s (Bc5)} Ng7
{+0.48/1 0s} 14. Bb3+ {+M52/1 0s} Kg6 {+0.48/31 0s} 15. Kg4
{+M51/1 0s (Ke5)} Ne8 {+0.48/1 0s} 16. Bc2+ {+M50/1 0s} Kf7
{+0.48/1 0s} 17. Kg5 {+M49/1 0s} Ke6 {+0.48/1 0s} 18. Bb3+
{+M48/1 0s (Bf5+)} Kd6 {+0.48/1 0s} 19. Bf2 {+M47/1 0s
(Kf5)} Nc7 {+0.48/1 0s} 20. Bg3+ {+M46/1 0s} Kc6 {+0.48/19
0s} 21. Ba4+ {+M45/1 0s (Kf5)} Kb6 {+0.48/1 0s} 22. Kf5
{+M44/1 0s (Bf2+)} Nb5 {+0.48/1 0s} 23. Bf2+ {+M43/1 0s}
Ka5 {+0.48/19 0s} 24. Bb3 {+M42/1 0s} Kb4 {+0.48/21 0s}
25. Be6 {+M41/1 0s} Kc3 {+0.48/23 0s} 26. Ke4 {+M40/1 0s}
Kb4 {+0.48/25 0s} 27. Bg8 {+M39/1 0s (Bh4)} Nd6+ {+0.48/1
0s} 28. Kd5 {+M38/1 0s (Ke5)} Nc4 {+0.48/1 0s} 29. Bf7
{+M37/1 0s (Be6)} Kc3 {+0.48/1 0s} 30. Bd4+ {+M36/1 0s} Kb4
{+0.48/23 0s} 31. Bf6 {+M35/1 0s (Bg6)} Nd2 {+0.48/1 0s}
32. Bg6 {+M34/1 0s} Nf3 {+0.48/1 0s} 33. Bd8 {+M33/1 0s}
Nd2 {+0.48/1 0s} 34. Bh4 {+M32/1 0s (Be7+)} Kb3 {+0.48/1
0s} 35. Kc5 {+M31/1 0s (Bf2)} Nf3 {+0.48/1 0s} 36. Bf6
{+M30/1 0s} Ng1 {+0.48/1 0s} 37. Bh5 {+M29/1 0s (Bf7+)} Kc2
{+0.48/1 0s} 38. Kd4 {+M28/1 0s} Kd2 {+0.48/1 0s} 39. Bg5+
{+M27/1 0s} Ke1 {+0.48/22 0s} 40. Ke3 {+M26/1 0s} Kf1
{+0.48/21 0s} 41. Bh4 {+M25/1 0s} Kg2 {+0.47/18 0s} 42. Bg4
{+M24/1 0s} Kh2 {+0.46/19 4s} 43. Kf2 {+M20/1 0s (Be7)}
Nh3+ {+0.44/1 0s} 44. Ke2 {+M19/1 0s} Ng1+ {+0.44/15 0s}
45. Kf1 {+M18/1 0s} Nh3 {+0.43/25 0s} 46. Bf5 {+M17/1 0s
(Be7)} Nf4 {+0.43/1 0s} 47. Kf2 {+M16/1 0s (Bf2)} Nh3+
{+0.43/1 0s} 48. Kf3 {+M15/1 0s (Ke1)} Ng1+ {+0.43/1 0s}
49. Kg4 {+M14/1 0s} Ne2 {+0.43/87 0s} 50. Bd3 {+M13/1 0s
(Be1)} Nd4 {+0.43/1 0s} 1/2-1/2[/pgn]

But this is sorted, I am fine and understood the DTM thing with different detours in the Nalimovs here.

BYe
Ingo

[syzygy]

Mate in 45- Either Nalimov is wrong with a Mate in 51 or SYS-Fish is not defining optimal. OR I make something wrong - which I will and can't exclude completly!

Again, Syzygybases are not DTM. SF+TB on the losing side is not maximising the number of moves to mate, but the number of moves to a capture or pawn push. (And it tries to do this in an intelligent way, so it will only really limit itself to the "DTZ50" optimal move if there is some chance of getting to 50 moves. So it will not refuse to capture a queen just because a capture will zero the 50-move counter immediately.)

The position is indeed a win for white. White can force a capture in 38 moves.

[IWB]

Mate in 45- Either Nalimov is wrong with a Mate in 51 or SYS-Fish is not defining optimal. OR I make something wrong - which I will and can't exclude completly!

Again, Syzygybases are not DTM. SF+TB on the losing side is not maximising the number of moves to mate, but the number of moves to a capture or pawn push. (And it tries to do this in an intelligent way, so it will only really limit itself to the "DTZ50" optimal move if there is some chance of getting to 50 moves. So it will not refuse to capture a queen just because a capture will zero the 50-move counter immediately.)

The position is indeed a win for white. White can force a capture in 38 moves.

Ok, so SYZ-Fish might lose faster if it is lost anyway while Nalimov is offering more resistance (as in this case!).

Actualy this bahaviour is not that optimal for analysing a positon. Playing a game is a different story (It will not lose are draw unessesarily) but writing a book and stating a # in 45 while it is 51 is "suboptimal"

I am just asking because until now I used the 6pc Nalimov only for analysing a positon and thought about changing to SYSbases. Now I have the advantage of taking care of the 50 move rule but the disadvantage of less resistance ...
Isn't there something for both?

Thx for your time to explain things
Ingo

[Sven]

Isn't there something for both?

Just g**gle for "tablebase dtm50" and you might find *very* recent information on it

AFAIK DTM50 is the combination of DTM and DTZ50 that you are looking for. But it has a price, of course.

[syzygy]

Playing a game is a different story (It will not lose are draw unessesarily) but writing a book and stating a # in 45 while it is 51 is "suboptimal"

So use any of the various Nalimov web interfaces when writing the book.

I am just asking because until now I used the 6pc Nalimov only for analysing a positon and thought about changing to SYSbases. Now I have the advantage of taking care of the 50 move rule but the disadvantage of less resistance ...

What kind of positions are you analysing? "Trivial" 5- and 6-piece positions with a long-known outcome (and for which the 50-move rule does not play a role)? Then use a webinterface. Positions with say 10 pieces? Then use my tables.

If you don't care about the 50-move rule (e.g. for correspondence games under the new rules), then set Syzygy50MoveRule to false (in SF).

[IWB]

Playing a game is a different story (It will not lose are draw unessesarily) but writing a book and stating a # in 45 while it is 51 is "suboptimal"

So use any of the various Nalimov web interfaces when writing the book.

It is just that people usually want to have the shortest possible mate with "best" defense and not a "It is lost anyway so here is a shorter helping mate"!

If you don't care about the 50-move rule (e.g. for correspondence games under the new rules), then set Syzygy50MoveRule to false (in SF).

Whats the difference to Nalimov if it doesnt obey the 50 move rule?

Another question is how often something like draw by 50 move rule really happen and if it is statisticaly relevant. But that is more a statistical thing ...

Thx
Ingo

[syzygy]

If you don't care about the 50-move rule (e.g. for correspondence games under the new rules), then set Syzygy50MoveRule to false (in SF).

Whats the difference to Nalimov if it doesnt obey the 50 move rule?

- size of the tables (maybe only matters if you still have to download them).
- is it still true that Nalimov 6-piece tables take 20 minutes to initialise upon engine start?
- probing speed (you can probe closer to the leaves with less overhead).

[IWB]

If you don't care about the 50-move rule (e.g. for correspondence games under the new rules), then set Syzygy50MoveRule to false (in SF).

Whats the difference to Nalimov if it doesnt obey the 50 move rule?

- size of the tables (maybe only matters if you still have to download them).
- is it still true that Nalimov 6-piece tables take 20 minutes to initialise upon engine start?
- probing speed (you can probe closer to the leaves with less overhead).

- I have them on a 1.5TB WD Caviar Green. But size is a feature for sure!
- From that really slow HD it needs 210s to initialize (quite similar with all engines - 20min? Where does this come from?)
-Speed is an interesting point, no doubt.

Thx
Ingo

[syzygy]

- From that really slow HD it needs 210s to initialize (quite similar with all engines - 20min? Where does this come from?)

I've seen that number often. For example in this thread:

6-man tables, excluding 5-1 positions, 1.12TB. Recommended that you load the table onto a 2TB internal HD and that you have 16GB RAM loaded on your machine. (You need the RAM for initial table loading. It takes me about 20 minutes with that much.)

For some reason I never realized that additional ram helped so much with loading of the 6-man tablebases. 20 minutes to load the whole set is amazing! I have 4GB ram and it takes such a long time to load that I've never even tried to load more than about 500GB worth.

20 minutes is amazing!

Mine don't take 210 seconds to initialise either. On a freshly booted system maybe 1 second to read and cache the directories where the files are to be found.