Patricia 4

[Whiskers]

https://github.com/Adam-Kulju/Patricia/releases/tag/4

When performing final testing with Pohl, I came to the uncomfortable realization that she is a good deal less aggressive in unbalanced books than more balanced books. At her strength, the reason is obvious - the only chance of winning a balanced opening against another strong engine is to go ham with the sacrifices, so nearly all her wins were sacrifices leading to an artificially high EAS score.

Pohl's testing still supports it being the world's most aggressive engine, however; she's also much stronger and it's probably going to take me at least a couple months to get her EAS back to 400,000+ in conditions where it matters. So, release time it is. I hope you all enjoy this new Patricia.

[pohl4711]

My testrun of Patricia 4.0 is finished: +172 Celo to Patricia 3.1. EAS-Score: 340432 (!)

https://www.sp-cc.de/files/uho_full_list.txt

Learn more about Patricia 4.0 and replay some insane sac-games of Patricia 4.0 from my testrun:
https://www.sp-cc.de/patricia_eas_engine.htm

[Joerg Oster]

https://github.com/Adam-Kulju/Patricia/releases/tag/4

When performing final testing with Pohl, I came to the uncomfortable realization that she is a good deal less aggressive in unbalanced books than more balanced books. At her strength, the reason is obvious - the only chance of winning a balanced opening against another strong engine is to go ham with the sacrifices, so nearly all her wins were sacrifices leading to an artificially high EAS score.

Pohl's testing still supports it being the world's most aggressive engine, however; she's also much stronger and it's probably going to take me at least a couple months to get her EAS back to 400,000+ in conditions where it matters. So, release time it is. I hope you all enjoy this new Patricia.

Thank you!

[pohl4711]

Jim Ablett made Windows/Linux and Android binaries, which do not have and Virus-alerts or Windows-permission errors, like the Patricia 4.0 binary on Patricia GitHub. Download:
https://drive.proton.me/urls/0GWGH1T6WG#wXMl5y7fwG35

(On my new AMD Ryzen machines, patricia_4_64_ja_bmi2 binary is the fastest (slightly faster than the Patricia_4.0_avx2 binary from GitHub!), but mention, older AMD CPUs support bmi2 only per emulation, so other binaries should run faster there: Just try the binaries by Jim out on your system)

[Werewolf]

https://github.com/Adam-Kulju/Patricia/releases/tag/4

When performing final testing with Pohl, I came to the uncomfortable realization that she is a good deal less aggressive in unbalanced books than more balanced books. At her strength, the reason is obvious - the only chance of winning a balanced opening against another strong engine is to go ham with the sacrifices, so nearly all her wins were sacrifices leading to an artificially high EAS score.

Pohl's testing still supports it being the world's most aggressive engine, however; she's also much stronger and it's probably going to take me at least a couple months to get her EAS back to 400,000+ in conditions where it matters. So, release time it is. I hope you all enjoy this new Patricia.

2 things:

1) Thank you so much for your hard work and for giving us this great gift!
2) Please make aggression - attacks on the king - the main thing and not Elo. There are many strong engines to choose from, but few aggressive ones.

[pohl4711]

Dont worry and look at my EAS-ratinglist (build from my UHO full Ratinglist gamebase).
https://www.sp-cc.de/files/uho_full_list.txt
Top10:

Code: Select all

                                 bad  avg.win 
Rank  EAS-Score  sacs   shorts  draws  moves  Engine/player 
-------------------------------------------------------------------
   1    432481  51.61%  38.03%  05.49%   66   Patricia 3.1 avx2  
   2    340432  46.96%  29.39%  03.46%   70   Patricia 4.0 avx2 
   3    253682  37.88%  23.88%  08.79%   75   Rebel Extreme avx2  
   4    221083  32.46%  23.04%  09.25%   71   Velvet 8.1.0 risky  
   5    204351  34.75%  22.22%  10.27%   73   Velvet 8.0.0 risky  
   6    200919  31.18%  29.46%  17.09%   71   Revenge 1.0 avx2  
   7    195881  21.25%  26.18%  09.40%   70   Stockfish 16 230630  
   8    193732  22.08%  24.73%  08.71%   70   Stockfish 16.1 240224  
   9    189267  25.24%  24.69%  13.42%   72   Velvet 7.0.0 avx2  
  10    187447  24.86%  20.78%  10.07%   75   Rebel EAS 2.0 avx2  

That there is an EAS regression to Patricia 3.1 is no surprise for me: Gaining +172 Celo must lead to such a regression. But IMHO the EAS-Score is still very, very impressive: Better than all other aggressive playing engines and way better than Stockfish 16. And Stockfish 17 has only 163866 EAS-points, so Patricia 4 has more than twice as many EAS-points as the latest Stockfish. Awesome!
And, Patricia 4.0 has the smallest number of bad draws (3.46%) of all tested engines (better than Patricia 3.1 here !!!)

Additionally, Patricia 4 has a clearly improved time-management and her evals are much less extreme and much less unstable/fluctuating than Patricia 3.1.
And Patricia 4 is much stronger in the endgame. This alone lowers the EAS-scoring, because more endgame-wins (after a game without sacs) lower the sac-ratio, because the EAS-sac-stats are calculated from percent-values of won games (compared to Patricia 3.1, not winning such endgames)

[pohl4711]

Here a very impressive win of Patricia 4 in my 14000 games testrun. 6 moves UHO openings. 3min+1sec singlethread.
Patricia sacs a full rook and 2 pawns (!!!) and mates Obsidian 8.0 in 31 moves. I doubt, that an engine can play much more aggressively...

[pgn]
[Event "UHO Ratinglist"]
[White "Patricia 4.0 avx2"]
[Black "Obsidian 8.0 avx2"]
[Site ""]
[Round "281"]
[Annotator "IWS-Tool: 5+ PawnUnits Sacrifice found in this game"]
[Result "1-0"]
[Date "2024.12.01"]
[PlyCount "61"]
[TimeControl "180+1"]

1. e4 c5 2. Nf3 Nc6 3. Bb5 e6 4. Bxc6 bxc6 5. b3 Qf6 6. e5 Qf5 7. d3 f6 8. Nc3 fxe5 9. Ne4 d6 10. O-O Be7 11. Nh4 Qf8 12. f4 exf4 13. Bxf4 Nf6 14. Qf3 Bd7 15. Bg3 O-O-O 16. b4 cxb4 17. a3 b3 18. Rab1 bxc2 19. Qe3 Nxe4 20. dxe4 Bxh4 21. Bxh4 cxb1=N 22. Rxb1 c5 23. Qb3 Bc6 24. Qb8+ Kd7 25. Qxa7+ Ke8 26. Qc7 Bb5 27. Rxb5 d5 28. Rb7 Qf2+ 29. Bxf2 Rd7 30. Qxd7+ Kf8 31. Qc8# 1-0
[/pgn]

[Archimedes]

Patricia 4.0 for Android:
https://sourceforge.net/projects/chess- ... p/download

Patricia 4.0 for Android (APK):
https://sourceforge.net/projects/chess- ... k/download

[jefk]

in the game P4-Obsidian, the move 15...0-0-0? is a mistake. But with
fast time controls and only one thread, such mistakes can happen (some
human players with a sense of danger, the also would think twice
before long castling into such an unsafe kings position...).

For normal humans the move 18.Rab1! would deserve a double exclamation mark
(other options is to cxb3 but that gives less advantage). Then cxb2 and
Qe3! gets another exclamation mark.

PS started with some test games with P4 as well with my gambit book, eg.
against Crafty (normal book) 5/2 time control and some more threads
and maybe an upload of the (very) interesting wins later on.

[Whiskers]

https://github.com/Adam-Kulju/Patricia/releases/tag/4

When performing final testing with Pohl, I came to the uncomfortable realization that she is a good deal less aggressive in unbalanced books than more balanced books. At her strength, the reason is obvious - the only chance of winning a balanced opening against another strong engine is to go ham with the sacrifices, so nearly all her wins were sacrifices leading to an artificially high EAS score.

Pohl's testing still supports it being the world's most aggressive engine, however; she's also much stronger and it's probably going to take me at least a couple months to get her EAS back to 400,000+ in conditions where it matters. So, release time it is. I hope you all enjoy this new Patricia.

2 things:

1) Thank you so much for your hard work and for giving us this great gift!
2) Please make aggression - attacks on the king - the main thing and not Elo. There are many strong engines to choose from, but few aggressive ones.

Thank you so much! I agree that aggression is more important than Elo, right up until a couple days ago all my testing was showing Patricia 4 is far more aggressive than Patricia 3, but sadly this doesn’t apply on some books and I wanted to release what I have instead of delaying Patricia 4 by weeks/months. Next version will have higher EAS don’t worry