Suicide chess tablebases (stalemated player wins)

[niklasf]

Since I had only 5 piece tables (with pawns) anyway I somewhat wastefully regenerated all DTZ tables. The new md5sums (https://syzygy-tables.info/suicide/2-3- ... s.stbz.md5 ) differ for 66 tables. My probing code and the stats now agree on everything except the 4 positions you pointed out

Now trying to find that mistake ... WDLs match the expected result in all of the root positions.

[niklasf]

Alright, found it. My probe_ab was reporting a "threat" when a capture could be forced, even if the position was still losing then. And then probe_dtz just returned 2/102.

All stats in agreement now.

[syzygy]

Since I had only 5 piece tables (with pawns) anyway I somewhat wastefully regenerated all DTZ tables. The new md5sums (https://syzygy-tables.info/suicide/2-3- ... s.stbz.md5 ) differ for 66 tables. My probing code and the stats now agree on everything except the 4 positions you pointed out

Did you not yet generate any 6-piece tables with pawns for giveaway? Then the damage remains small.

If you did generate a substantial number of 6-piece tables with pawns, then let me know so that I will create a small program to check whether they need to be regenerated. (I will probably make such a thing for personal use anyway.)

[niklasf]

I started generating pawnless 6 piece tables first, and that's not even complete. So I am in luck, here.

[syzygy]

I started generating pawnless 6 piece tables first, and that's not even complete. So I am in luck, here.

Now let's hope there are no further hidden problems

I have been able to reduce the number of potentially problematic DTZ tables to 595 now. Some are false positives (e.g. 90 tables with 2-5 pieces compared to the 66 you found), but regenerating 595 tables is a lot better than regenerating all 4830 tables with pawns.

[niklasf]

Status update from me:

- The 2310 pawnless 6 piece tables are now complete. As always I uploaded them to https://syzygy-tables.info/suicide/ including checksums.
- I've ported giveaway table probing to Daniel Dugovic's variant-playing Stockfish: https://github.com/ddugovic/Stockfish

Next:

- Generate 6 piece giveaway tables with pawns
- Online lookup

[Henk]

What a bad name for a game. No need to advertise psychic diseases. It may even be contagious. Don't mention the war.

[syzygy]

Status update from me:

- The 2310 pawnless 6 piece tables are now complete. As always I uploaded them to https://syzygy-tables.info/suicide/ including checksums.

All checksums are identical to mine.

In the meantime I have regenerated the 4/5/6-piece suicide DTZ tables with pawns that were potentially affected by the cursed win bug. Surprisingly, their total size is now 13952 bytes less. This seems to be caused mainly by KKPPvQP.stbz, which went from 5655440 bytes to 5627344 bytes. I suspect the slight changes in the uncompressed data caused the generator to pick a different permutation for that table which just happened to compress a bit better.

Good luck generating 6-piece pawnful giveaway tables

[niklasf]

After round about 4 years of CPU time the 6 piece giveaway tables are finally complete.
The 2310 pawnless and 3444 giveaway tables together occupy 317G WDL + 68G DTZ = 384G total (and then additional 10G for the 2646 tables with less pieces).

Stats, checksums and downloads can be found at https://syzygy-tables.info/suicide/.

Online lookup based on a modified version of Daniel Dugovic's variant playing Stockfish is hosted by (the free software chess website) lichess.org as a public HTTP API.

An online user interface for analysing antichess games is also available: https://lichess.org/analysis/antichess#opening. FENs can be pasted below. Tablebases will be shown in the lower right corner. (Btw. Watkins antichess proof trees can also be selected as an opening book in the lower right corner. Syzygy tables will be used once the position reaches tablebase land.)

For offline usage Daniel Dugovic's Stockfish and my library python-chess support probing variant Syzygy bases.

I have not yet decided if I will also compute the pawnful tables with FICS rules.

Thanks again, Ronald de Man, for the amazing generator and your help with the probing code. This was a fun project

[George Tsavdaris]

I have also added a generator for atomic chess just in case someone is interested.

Hi, does the TB generator here https://github.com/syzygy1/tb for Atomic Chess support 7 men generation for Atomic Chess?
I would like to generate certain 7 men TBS if possible so i wonder if it supports them.
Also how the command line to generate Atomic Chess TBs works anyway? Any info?

BTW is this (https://github.com/syzygy1/tb) able to generate the 7 men Chess TBs? Because the info on the page talks only about 6 men Chess EGTBs. So is just the info on the page just outdated?