Swordfish Engine

    Swordfish is a Stockfish derivative that is reported to have achieved good results in test suites.  Supposedly it can be useful for analyzing sharp positions where regular Stockfish can overlook some tactics.

    Swordfish is based on Stockfish, but is designed to be more aggressive and tactical. It balances calculation strength with a highly active, fighting style that often finds sharp, complex tactical lines in positions where Stockfish might settle for a less aggressive move. 

    Unlike Stockfish, which is optimized for absolute perfection, Swordfish leans heavily towards proactive attacks and sharp, forcing lines. This trait can make it a useful tool in some situations and it might be worth downloading which you can do HERE.

    I have not found any online rating for Swordfish, but on my laptop five game (4 minute plus 2 seconds per move) against Stockfish were drawn. Analysis was done with the Reckless engine and both engines scored a 100% accuracy rating. Therefore, I played another match against the weakest engine on my laptop, Lc0, and Swordfish scored 5-0. 

    In all five games Lc0 held its own until committing a catastrophic blunder that lost instantly. The following game was the most interesting because at the critical moment it took both Stockfish and Swordfish some time to realize that capturing the P on move 29 was a mistake. 

    As can be seen on the CCLR site, Stockfish is still the top rated engine, but only by a whisker and for most analysis purposes any of the top engines will do an adequate job. 

 

[Event "Blitz 4.0min+2.0sec"] [Site "PC"] [Date "2026.??.??"] [Round "?"] [White "Lc0"] [Black "Swordfish"] [Result "0-1"] [ECO "A09"] [Annotator "Stockfish 18/et al"] [PlyCount "66"] {[%evp 11,66,19,20,23,18,14,18,20,14,18,15,19,18,24,20,18,15,3,3,4,0,3,8,6,13, 1,4,0,0,0,0,3,0,0,2,0,1,-4,0,0,0,0,0,0,0,-8,0,-7,-8,-399,-395,-394,-392,-387, -382,-382,-379] A09: Reti Opening} 1. Nf3 d5 2. g3 Nf6 3. Bg2 c5 4. O-O e6 5. c4 d4 6. d3 Nc6 7. e3 Bd6 8. exd4 cxd4 9. Bg5 h6 10. Bxf6 Qxf6 11. a3 a5 12. Nbd2 {[%mdl 32]} Qe7 13. Ne4 Bc7 14. b4 axb4 15. axb4 Rxa1 16. Qxa1 f5 17. Nc5 O-O {In human games both 17...O-O and 17...Nxb4 have been played. Both moves are equally as goog as Swordfish's.} 18. Re1 Nxb4 19. Nxe6 {All the engines immediately saw this moves. Retreating the N would lose.} Nc2 20. Nxf8 Qxe1+ $1 21. Qxe1 Nxe1 22. Nxe1 Kxf8 {[%mdl 4096] The tactics are over and the evaluation is 0.00.} 23. Nc2 Bb6 24. Na1 Bc5 25. Nb3 b6 26. f4 g5 27. Kf2 Bd6 28. Kf3 Bd7 {The critical position.} 29. Nxd4 {This is the losing move. All the engines were evaluating the position at 0.00 and assigned the loss to 30. Nc2. It took Stockfish and Swordfish severak minutes to realize that capturing the P was an error and that 29.Bf1 keeps things equal.} (29. Bf1 Bc6+ 30. Ke2 gxf4 31. gxf4 Bxf4 32. Nxd4 Bd7 {would lead to a draw.}) 29... Bc5 {This is the position where all the engines show white losing after this move and suggesting 30.Ke3 would maintain equality.} 30. Nc2 {[%mdl 8192]} (30. Ke3 { It takes some lengthy analysis, but the win is there.} g4 31. Bd5 Kg7 32. Bg2 Kf6 33. Bd5 h5 34. Bg2 h4 35. Bd5 hxg3 36. hxg3 Be8 37. Bg2 Ke7 38. Bd5 Kd6 39. Bg2 Bf7 40. Bb7 Be6 41. Bg2 b5 42. cxb5 Bd5 43. Bf1 Bf3 44. b6 Bxb6 {White is in Zugzwang.}) 30... Bc6+ {Black is winning.} 31. Ke2 Bxg2 32. d4 Bd6 33. fxg5 Be4 {White resigned, Accuracy: White = 80%, Black = 100%.} 0-1