Zeppelin Chess Engine

[Tibono]

A nice short game by Zeppelin 1.1.0 vs Darky 0.5d
Time control 40/10. Analysis by Fritz 19 (7s/move, 2 cores).

[pgn][Event "Arena Tournament.at"]
[Site "Arena"]
[Date "2026.02.07"]
[Round "2"]
[White "Zeppelin 1.1.0 NNUE"]
[Black "Darky 0.5d"]
[Result "1-0"]
[Annotator "Tactical Analysis 7.0"]
[ECO "D47"]
[WhiteElo "2071"]
[BlackElo "2028"]
[WhiteFideId "-1"]
[BlackFideId "-1"]
[PlyCount "53"]
[GameId "2274980901636559"]
[EventDate "2026.??.??"]
[TimeControl "40/600:40/600:40/600"]

1. d4 d5 2. Nf3 Nf6 3. c4 c6 4. Nc3 e6 5. e3 Nbd7 6. Bd3 dxc4 7. Bxc4 b5 8. Bd3 b4 9. Ne4 Bb7 {30} 10. Nxf6+ {19} Qxf6 {28} 11. O-O {19} c5 {27 [#] White is slightly better.} 12. Ne5 $1 {19} O-O-O {25} 13. Nxd7 {19} Rxd7 {24} 14. Qa4 {19} cxd4 {23} 15. Qxa7 {19 Hoping for Bc1-d2.} e5 $2 {22} (15... Qe5 $16 {is tougher.}) 16. Bd2 $18 {19} dxe3 {20} (16... Bd6 $142 17. Rac1+ Kd8) 17. Rac1+ {19} ({Not} 17. Qxe3 Kb8 $16) 17... Rc7 {19} 18. Rxc7+ {19} Kxc7 {18} 19. fxe3 {19} (19. Rc1+ $142 Qc6 20. Rxc6+ Kxc6 21. Be4+ {Skewer, Deflection} (21. Qxe3 Bc5 $18) 21... Kd7 22. Qxb7+ Ke6 23. Bd5+ (23. Bxe3 h6 24. Qc8+ Kf6 25. Qc6+ Ke7 26. Bc5+ Kd8 27. Bf5 Bxc5 28. Qd7#) 23... Kf6 24. Qxf7+ (24. Bxe3 Be7 25. Qc6+ Kf5 26. Be4+ Kg4 27. Bf3+ Kh4 28. Qe4#) 24... Kg5 25. Bxe3+ Kh4 26. g3+ Kh3 27. Qh5#) 19... Qe7 {17 Black is weak on the light squares} 20. Rc1+ {19 [#] Deflection} Kd7 {17} 21. Qxb7+ {19} (21. Bf5+ $142 Qe6 22. Qxb7+ (22. Bxe6+ Kxe6 23. Qxb7 Bd6 $18) 22... Ke8 23. Bxe6 Be7 24. Rc8+ {Skewer} Bd8 25. Qxf7#) 21... Ke6 {16} 22. Qc8+ {19} Qd7 {15} 23. Rc6+ {3 White mates.} Bd6 {14} 24. Rxd6+ {0} Kxd6 {13} 25. Bxb4+ {0} Ke6 {13} 26. Qa6+ {0} Qc6 {12} 27. Qxc6# {0 Weighted Error Value: White=0.47/Black=1.69 Loses game: --- Black=1 Mistake: --- Black=1 Inaccurate: White=2 Black=2 OK: White=3 --- Best: White=2 Black=1} 1-0

[/pgn]

Tournament Elo perf. for Zeppelin 1.1.0: 2170
All the best,
Tibono

[ChickenMan4236]

Another version of Zeppelin - Zeppelin v1.2.0
I've managed to improve move ordering, it's got some ELO increase.

Thanks for Your testing Eric, I think this situation, when Zeppelin missed the mate, must have happened when it was extremely low on time. I was trying to reproduce the this bug and I've noted, that something similar happens if there is less than one second left for Zeppelin in that position. I'm not completely sure if it's related to time control, but if it is, then I'm also not sure if it's a "fixable" problem. For now I'm more focused on increasing overall strength of the engine, I'm currently waiting for my NNUE training data to be generated, but at some point I'm gonna investigate problems like this more deeply

[ChickenMan4236]

Hello,
I would like to announce a new version - Zeppelin v1.3.0

I've played around with search extensions, the new version can reach a little higher depths. I've also implemented a basic pondering support and retrained the NNUE. As always - strength increase confirmed by SPRT test

http://zeppelin.bot.nu/
https://github.com/jszczerbinsky/zeppel ... tag/v1.3.0

[ChickenMan4236]

Hello,
Another version is coming - this time I've trained my NNUE with over 100million positions, the previous ones were trained using only around 10million.

Check out Zeppelin v1.4.0
https://github.com/jszczerbinsky/zeppel ... tag/v1.4.0
https://zeppelin.bot.nu/versions

Recently I created a mirror of Zeppelin's code on codeberg, I'm thinking about moving my repo permamently there, due to Microsoft's AI policy and using github repos for training LLMs.
https://codeberg.org/jszczerbinsky/zepp ... tag/v1.4.0

[ChickenMan4236]

Hello,
Minor update released: Zeppelin v1.4.1 - some improvements in move ordering and skipping moves with low SEE in quiescence search

https://zeppelin.bot.nu
https://github.com/jszczerbinsky/zeppel ... tag/v1.4.1
https://codeberg.org/jszczerbinsky/zepp ... tag/v1.4.1

[ChickenMan4236]

Hello,
New release just came out - Zeppelin v1.4.2!
This time I focused mainly on optimizations related to CPU architecture.

• I've implemented PEXT move generation

• I've added support for CPUs with popcount

Download the new version:
https://github.com/jszczerbinsky/zeppel ... tag/v1.4.2
https://codeberg.org/jszczerbinsky/zepp ... tag/v1.4.2
https://zeppelin.bot.nu/versions