WCCC 2023?

[Eduard]

Fritz is Ginkgo and has just won the software wccc. Now a new Fritz 19 is coming onto the market. Soon.

The good: New software from ChessBase.
The bad: I bought almost all Fritz programs to test them. Powerfritz 18 (Ginkgo) was the worst engine in my ENET-2023 test, last place and even weaker than Rebel 16.2.

The GUI Powerfritz 18 update 17 is buggy. I had to downgrade to update 10 to be able to play with it.

Will everything get better with Fritz 19?

[Gian-Carlo Pascutto]

Fritz won the software championship by beating Deep Sjeng twice with white. All other games in the tournament were drawn. I will send this Deep Sjeng to the rating lists after the tournament so you can get some idea. I doubt that the engine that played here as Fritz (whichever it is) is "weak". That said, we did take huge risks to avoid 100% draws and it succeeded, it's just that the wrong side won

Random observation: the Ryzen 3700X that was used here is probably closer to what actual users have than 52 core / 104 thread AVX512 low-clocked server chips like TCEC uses, which quite frankly as an engine developer is frustrating because you have to optimize for hardware you don't have and really nobody else has. The dual RTX 3080 Ti on 3900X that I used in Gouda is 50% faster than the TCEC hardware. How this is possible I don't know.

At least a 3700X you can get on ebay for 100 bucks

I think our odds are better for the WCCC where Stoofvlees/DeepSjeng are allowed to use the GPU, but we shall see what happens.

[Eduard]

Thanks for the information. I'm interested in the two winning games in PGN. Can someone post them here?

[peter]

Thanks for the information. I'm interested in the two winning games in PGN. Can someone post them here?

[pgn][Event "ICGA WCSC 2023"]
[Site "Valencia"]
[Date "2023.07.11"]
[Round "4.1"]
[White "Fritz"]
[Black "Stoofvlees"]
[Result "1-0"]
[ECO "B12"]
[Annotator "broadcasts"]
[PlyCount "107"]
[EventDate "2023.07.01"]
[EventType "tourn"]
[EventCountry "ESP"]

1. e4 c6 2. d4 d5 3. e5 Bf5 4. Be2 e6 5. Nf3 c5 6. Be3 cxd4 7. Nxd4 Ne7 8. Nd2
a6 9. g4 Be4 10. O-O h5 11. f4 Nbc6 12. Nxc6 Nxc6 13. Nxe4 dxe4 14. gxh5 Qa5
15. Qe1 Bc5 16. Qf2 Bxe3 17. Qxe3 O-O-O 18. c3 Ne7 19. Qxe4 Qb6+ 20. Kh1 Nf5
21. a4 Rd2 22. Rad1 Rhd8 23. Rxd2 Rxd2 24. b4 Nh4 25. a5 Qc7 26. Qc4 Qxc4 27.
Bxc4 Kd8 28. Bb3 Ke7 29. Rg1 Kf8 30. Rd1 Rf2 31. Ba4 g5 32. Kg1 Ra2 33. Rd8+
Kg7 34. Bd1 gxf4 35. Rd4 Nf5 36. Rd7 Nh4 37. Kf1 Nf5 38. Bb3 Ne3+ 39. Ke1 Ng2+
40. Kd1 Ne3+ 41. Kc1 f3 42. h6+ Kf8 43. Rd8+ Ke7 44. Bxa2 f2 45. Bc4 Nxc4 46.
Rd1 Kf8 47. Rf1 Nxe5 48. h3 f5 49. Kc2 Nf7 50. h7 Kg7 51. Rxf2 Kxh7 52. Rd2 Ne5
53. Re2 Nf7 54. Rxe6 1-0
[/pgn]
[pgn][Event "ICGA WCSC 2023"]
[Site "Valencia"]
[Date "2023.07.11"]
[Round "10.1"]
[White "Fritz"]
[Black "Stoofvlees"]
[Result "1-0"]
[ECO "A80"]
[Annotator "broadcasts"]
[PlyCount "91"]
[EventDate "2023.07.01"]
[EventType "tourn"]
[EventCountry "ESP"]

1. d4 f5 2. Nf3 e6 3. Bf4 Nf6 4. e3 Be7 5. Be2 O-O 6. c4 b6 7. Nfd2 Bb7 8. O-O
d6 9. Nc3 Nbd7 10. h3 Rc8 11. Bh2 Kh8 12. Qb3 g5 13. c5 Nd5 14. cxd6 Bxd6 15.
Bxd6 cxd6 16. Nb5 Ba6 17. Rae1 Bxb5 18. Bxb5 N7f6 19. g3 Ne4 20. Nxe4 fxe4 21.
Rc1 h5 22. Rxc8 Qxc8 23. Bc4 Qb7 24. a4 Kg7 25. a5 h4 26. g4 Rc8 27. axb6 axb6
28. Kg2 b5 29. Bxd5 exd5 30. Ra1 Rb8 31. Qb4 Qb6 32. Ra5 Kf6 33. Qa3 Ke6 34.
Ra6 Qc7 35. Ra7 Rb7 36. Ra8 Rb8 37. Qa6 Qd8 38. Rxb8 Qxb8 39. Qc6 b4 40. Kf1 b3
41. Kg1 Qf8 42. Qb5 Qf3 43. Qe8+ Kf6 44. Qd8+ Ke6 45. Qxg5 Qxh3 46. Qf4 1-0
[/pgn]

[Eduard]

Thank you Peter!

I looked at Stefan Pohl's page and I see that the WCh from 2022 (GinkGo), as the later ChessBase Engine Powerfritz 18, only occupies place 29 at Stefan. Rebel 16.1 is in 15th place with 140 Elo more than Fritz.

Stefan Pohl Rating List:
7 CSTal 2.0 avx2 : 3685 7 7 10000 49.4% 3690 78.2%
15 Rebel 16.1 : 3621 5 5 15000 49.3% 3625 76.4%
29 PowerFritz 18 avx2 : 3478 5 5 11000 48.9% 3485 62.7%

That agrees with my test. Nobody really needs a world champion who plays so weak, right? Let's hope that the new Fritz 19 will play better, but I have doubts because all other programs are also getting better, see ChessSytem Tal 2 which now plays significantly better than Rebel 16.1

[Gian-Carlo Pascutto]

They did a Fritz Neuronal update so there must be a newer version with NNUE. I don't know about PowerFritz.

The Rebel/CSTal are based on another engine IIRC so it's not relevant to WCCC, and ChessBase has probably learned their lesson with selling GPL clones

[Eduard]

ICGA is now already promoting ChessBase:

https://icga.org/

World Chess Software Champion
Posted on July 11, 2023 by Jonathan Schaeffer
The winner is Fritz. Fritz is one of the best-selling chess programs in the history of computer chess products. The first Fritz became available around 30 years ago. An early version of the program won the 1995 World Computer Chess Championship, with a famous win over Deep Blue.

Congratulations to Frank Schneider, the book developer Wolfgang Zugrav, and the rest of the Fritz team.

Thanks!
If you are not familiar with the context, you might want to believe that Frank Schneider won against Deep Blue with Fritz.

It should be noted that Fritz is actually GinkGo who also played last year. But now suddenly Fritz is written as a special program that already won against Deep Blue in 1995. But that was a completely different Fritz. Fritz 2023 is GinkGo, and GinkGo 2022 was last sold as Powerfritz 18. Unfortunately, this engine cannot convince in any ranking.

[Eduard]

I compared the first game with PlayChess engine games since 2021. Move 8 for black is not in my book. I've been playing all sorts of openings on PlayChess for years, almost every day. My engine always played 8...Nc6. As things progressed, Stoofvlees blundered on move 24. Overall, Caro-Kann is not suitable for ENG games if you want to win. It has nothing to do with risk, you won't win with Black against an equally strong ENGINE opponent with Caro-Kann. My engine draws 99%. And for a draw you have to be able to fight well with Caro-Kann, that's not easy.

[Event "ICGA WCSC 2023"]
[Site "Valencia"]
[Date "2023.07.11"]
[Round "4.1"]
[White "Fritz"]
[Black "Stoofvlees"]
[Result "1-0"]
[ECO "B12"]
[Annotator "broadcasts"]
[PlyCount "107"]
[EventDate "2023.07.01"]
[EventType "tourn"]
[EventCountry "ESP"]

1. e4 c6 2. d4 d5 3. e5 Bf5 4. Be2 e6 5. Nf3 c5 6. Be3 cxd4 7. Nxd4 Ne7 8. Nd2 a6 {Not in my universal tournament book} (8... Nbc6 {Most played move} 9. N2f3 Be4 10. O-O Bxf3 11. Nxf3 Qc7 12. Rc1 Rd8 13. c3 Ng6 14. Bd4 Be7 {Most played variation}) 9. g4 Be4 10. O-O h5 11. f4 Nbc6 $146 {New Move in my Engine Big Book wth 2,5 Mio games.} (11... hxg4 12. Nxe4 dxe4 13. f5 Nxf5 14. Nxf5 Qxd1 15. Raxd1 exf5 16. Rxf5 Nc6 17. Bxg4 g6 18. Rf4 Nxe5 $11) 12. Nxc6 Nxc6 13. Nxe4 dxe4 14. gxh5 Qa5 15. Qe1 Bc5 16. Qf2 Bxe3 17. Qxe3 O-O-O 18. c3 $14 Ne7 19. Qxe4 Qb6+ 20. Kh1 Nf5 21. a4 Rd2 22. Rad1 Rhd8 23. Rxd2 Rxd2 24. b4 Nh4 $2 {Blunder} (24... Qc7) 25. a5 Qc7 26. Qc4 Qxc4 27. Bxc4 Kd8 28. Bb3 Ke7 29. Rg1 Kf8 30. Rd1 Rf2 31. Ba4 g5 32. Kg1 Ra2 33. Rd8+ Kg7 34. Bd1 gxf4 35. Rd4 Nf5 36. Rd7 Nh4 37. Kf1 Nf5 38. Bb3 Ne3+ 39. Ke1 Ng2+ 40. Kd1 Ne3+ 41. Kc1 f3 42. h6+ Kf8 43. Rd8+ Ke7 44. Bxa2 f2 45. Bc4 Nxc4 46. Rd1 Kf8 47. Rf1 Nxe5 48. h3 f5 49. Kc2 Nf7 50. h7 Kg7 51. Rxf2 Kxh7 52. Rd2 Ne5 53. Re2 Nf7 54. Rxe6 1-0

[Modern Times]

Nobody really needs a world champion who plays so weak, right?
29 PowerFritz 18 avx2 : 3478 5 5 11000 48.9% 3485 62.7%

A 3,500 engine is not weak.

[Ajedrecista]

Hello Eduard:

I compared the first game with PlayChess engine games since 2021. Move 8 for black is not in my book. I've been playing all sorts of openings on PlayChess for years, almost every day. My engine always played 8...Nc6. As things progressed, Stoofvlees blundered on move 24. Overall, Caro-Kann is not suitable for ENG games if you want to win. It has nothing to do with risk, you won't win with Black against an equally strong ENGINE opponent with Caro-Kann. My engine draws 99%. And for a draw you have to be able to fight well with Caro-Kann, that's not easy.

[pgn][Event "ICGA WCSC 2023"]
[Site "Valencia"]
[Date "2023.07.11"]
[Round "4.1"]
[White "Fritz"]
[Black "Stoofvlees"]
[Result "1-0"]
[ECO "B12"]
[Annotator "broadcasts"]
[PlyCount "107"]
[EventDate "2023.07.01"]
[EventType "tourn"]
[EventCountry "ESP"]

1. e4 c6 2. d4 d5 3. e5 Bf5 4. Be2 e6 5. Nf3 c5 6. Be3 cxd4 7. Nxd4 Ne7 8. Nd2 a6 {Not in my universal tournament book} (8... Nbc6 {Most played move} 9. N2f3 Be4 10. O-O Bxf3 11. Nxf3 Qc7 12. Rc1 Rd8 13. c3 Ng6 14. Bd4 Be7 {Most played variation}) 9. g4 Be4 10. O-O h5 11. f4 Nbc6 $146 {New Move in my Engine Big Book wth 2,5 Mio games.} (11... hxg4 12. Nxe4 dxe4 13. f5 Nxf5 14. Nxf5 Qxd1 15. Raxd1 exf5 16. Rxf5 Nc6 17. Bxg4 g6 18. Rf4 Nxe5 $11) 12. Nxc6 Nxc6 13. Nxe4 dxe4 14. gxh5 Qa5 15. Qe1 Bc5 16. Qf2 Bxe3 17. Qxe3 O-O-O 18. c3 $14 Ne7 19. Qxe4 Qb6+ 20. Kh1 Nf5 21. a4 Rd2 22. Rad1 Rhd8 23. Rxd2 Rxd2 24. b4 Nh4 $2 {Blunder} (24... Qc7) 25. a5 Qc7 26. Qc4 Qxc4 27. Bxc4 Kd8 28. Bb3 Ke7 29. Rg1 Kf8 30. Rd1 Rf2 31. Ba4 g5 32. Kg1 Ra2 33. Rd8+ Kg7 34. Bd1 gxf4 35. Rd4 Nf5 36. Rd7 Nh4 37. Kf1 Nf5 38. Bb3 Ne3+ 39. Ke1 Ng2+ 40. Kd1 Ne3+ 41. Kc1 f3 42. h6+ Kf8 43. Rd8+ Ke7 44. Bxa2 f2 45. Bc4 Nxc4 46. Rd1 Kf8 47. Rf1 Nxe5 48. h3 f5 49. Kc2 Nf7 50. h7 Kg7 51. Rxf2 Kxh7 52. Rd2 Ne5 53. Re2 Nf7 54. Rxe6 1-0[/pgn]

Artemiev played 8.- ..., a6 against So at Meltwater 2021, finishing in a white win (ChessGames link). It is true that Nbc6 was the next move by black.

The chessdb.cn opening explorer currently gives +0.10 to 8.- ..., Nbc6 (top choice) and +0.37 to 8.- ..., a6 (second best choice).

Regards from Spain.

Ajedrecista.