Is there now coming changes to syzygy databases?

[jpqy]

TCEC Live+Chat via Twitch:

http://tcec.chessdom.com/live.php

Rybka chat:

http://rybkaforum.net/cgi-bin/rybkaforum/chat_show.pl

JP.

[Dann Corbit]

I have adapted my generator for DTM, but DTM tables cannot replace WDL/DTZ. But they can be used in addition to WDL/DTZ to get mate distances (huge waste of disk space, though).

I am not yet finished.

Would it be possible to replace some DTZ tables with DTM if those tables have no cursed wins or blessed losses?

It would not be a perfect replacement.

Suppose the engine is on the losing side in KQvKR. The opponent can always win it within the 50-move rule, but he might play several suboptimal moves that bring the 50-move counter to, say, 25 moves with 26 moves remaining until the capture of the rook. So now the engine has a "blessed loss" within its reach, but without DTZ table it might not be able to tell and it might not play the right moves. (DTM would try to postpone the mate as much as possible, possibly at the cost of losing the rook a move or so earlier.)

Anyway, current Cfish is able to deal at the root with any of WDL+DTZ, WDL+DTM and WDL+DTZ+DTM.

In the search, Cfish will probe WDL at TB positions and then only probe DTM if the normal TB-win score is does not lead to a cutoff (i.e. if a win is already certain and the engine starts to look for the shortest win). This fits nicely with the early_mate approach. It is also possible to use 6-piece WDL(+DTZ) with 5-piece DTM.

Dumb question (I am sure others already know this):
So are there now Syzygy DTM files or does Cfish work with other DTM files like Gaviota TB?

[syzygy]

So are there now Syzygy DTM files or does Cfish work with other DTM files like Gaviota TB?

Cfish works only with Syzygy DTM files, which are not yet available

[phhnguyen]

TCEC Live+Chat via Twitch:

http://tcec.chessdom.com/live.php

Rybka chat:

http://rybkaforum.net/cgi-bin/rybkaforum/chat_show.pl

JP.

Thanks a lot!

Look like I cannot find what I need from a live chat 🙁

[Uri Blass]

Is WDL+DTZ+DTM enough to ensure perfect play under the 50 move rule or do we need DTM50?

Another idea you may already be doing is splitting tables where one side has two bishops into two parts, one with one bishop of each color and one with both bishops on the same color. This would likely improve compression and would improve caching.

Adding DTM50 is also not enough to ensure perfect play against non perfect opponent because it is possible to draw by the repetition rule.

Suppose the engine is on the losing side in KQvKR. The opponent can always win it within the 50-move rule, but he might play several suboptimal moves that bring the 50-move counter to, say, 25 moves with 24 moves remaining until the capture of the rook so move A that is the best based on the DTM50 tables is losing the game

After move B the opponent can in theory win with 23 moves until the capture of the rook but in order to do it the opponent has to allow draw by the repetition rule and without allowing draw by the repetition rule the opponent has no way to win the game because of the fifty move rule.

Move B is the best move that force a draw but all the tablebases in this thread are not going to help the engine to find move B.

The only practical way to solve the problem is to build specific tablebases that use the history of the specific game and the rules of the specific game that may define some positions that already happened twice in the game as a draw if they happen again so practically tablebases should consider 3 factors:
1)material on the board
2)fifty move counter
3)positions that are defined to be a draw(in most cases 3 is empty and in that case the DTM50 tablebases are enough but not always).

The number of possibilities for 3 is too big and the only practical solution may be to build the relevant tablebases during the game if at least one position already happened twice.

[jshriver]

So are there now Syzygy DTM files or does Cfish work with other DTM files like Gaviota TB?

Cfish works only with Syzygy DTM files, which are not yet available

I'd be willing to help with that have the generator code or linux binaries? I have about 4 machines I could let burn then seed the set once it's done.

[Jouni]

After changes ALL syzygy engines and GUIs must be re-programmed ! Fritz GUI: Nalimov DTM -> syzygy DTZ -> syzygy DTM - is the final solution ?

[syzygy]

After changes ALL syzygy engines and GUIs must be re-programmed ! Fritz GUI: Nalimov DTM -> syzygy DTZ -> syzygy DTM - is the final solution ?

No, WDL+DTZ already is the perfect solution.

DTM only adds a bit of cosmetic.

[Nordlandia]

It was said there is several options in regards to overall size.

5-men DTM format candidates.

Smallest set at 2 GB

Middle 2.6 GB and

Largest 5 GB - this one is to some extent faster than 2.6 GB version. Maybe 5 GB version benefit more from fast SSDs.

One set for download avoid unnecessary confusion. However if it is true 5 GB version is better for analysis than 2.6 GB, maybe it is a good idea to include for voluntary download
option regardless which set is to be preferred.