Xiphos x Ethereal: Who has the best style?

[Ovyron]

See here for context.

The top 2 chess engines I never tested were Xiphos and Ethereal, arguably the top 7 and 6 chess engines.

As usual I ran a match of 10 games after calibrating their strength so they played at the same level and could showcase their styles. (PGN for collectors)

Contrary to other tests of this kind I've run, there's not much to talk about regarding style. No king attacks or losing in amazing fashion, most of the time the engines seemed to just be happy with a draw.

I've chosen to highlight 3 games:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "7"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A04"]
[Result "1/2-1/2"]

1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Bb4 5. a3 Bxc3+ 6. bxc3
O-O 7. e3 c5 8. Bb2 Qa5 9. Nd2 cxd4 10. cxd4 Bd7 11. Qc2
Rc8 12. Bc3 Qa6 13. cxd5 Qa4 14. Qxa4 Bxa4 15. Bb2 Nxd5
16. Bd3 f5 17. O-O a6 18. Rfc1 Nc6 19. Be2 Na5 20. g4 g6
21. g5 Bc2 22. a4 Rc6 23. Bf1 Rac8 24. Bb5 R6c7 25. Be2 Bb3
26. Rxc7 Rxc7 27. Ra3 Bc2 28. h4 Be4 29. Bd1 b6 30. Kf1 Nc4
31. Nxc4 Rxc4 32. Bb3 Rc7 33. Ke2 Bc2 34. Bxd5 exd5 35. a5
b5 36. Rc3 Rxc3 37. Bxc3 Be4 38. Be1 Kf7 39. Bb4 Ke6
40. Bc5 Kd7 41. Ba3 Ke6 42. Bf8 f4 43. f3 Bc2 44. exf4 Kf5
45. Bd6 Bb3 46. Kd2 Ba4 47. Ke2 Bb3 48. Kd2 Ba4 49. Ke2 Bb3
1/2-1/2
[/pgn]

[d]2r3k1/1p5p/p1r1p1p1/nB1n1pP1/P2P4/4P3/1BbN1P1P/R1R3K1 b - -

Xiphos dares Ethereal into taking the Bb5 Bishop, but of course Ethereal doesn't bite.

[pgn]
[Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "2"]
[White "Ethereal"]
[Black "Xiphos"]
[ECO "A40"]
[Result "1/2-1/2"]

1. d4 d5 2. Nf3 Nf6 3. c4 e6 4. Nc3 c6 5. e3 Nbd7 6. Qc2
Bd6 7. Bd3 dxc4 8. Bxc4 O-O 9. O-O b5 10. Bd3 Bb7 11. e4 e5
12. dxe5 Nxe5 13. Nxe5 Bxe5 14. h3 b4 15. Na4 Bd4 16. e5
Bxe5 17. Nc5 Bc8 18. Re1 Qc7 19. Bd2 a5 20. a3 bxa3
21. Rxa3 h6 22. Ra4 Re8 23. b4 Bxh3 24. bxa5 Bc8 25. Ne4
Nd5 26. Rc4 Re6 27. Rc1 Nf4 28. Nc5 Re7 29. Bf1 Ng6 30. Na4
Bb7 31. Nc5 Bc8 32. Na4 Bb7 33. Nc5 Bc8 1/2-1/2
[/pgn]

[d]r3r1k1/2q2pp1/2p2n1p/p1N1b3/RP6/3B3b/2QB1PP1/4R1K1 w - -

Xiphos sacrifices the Bishop for the h3 pawn, but Ethereal doesn't take it and the game ends shortly after.

And that's it.

Yeah, those were the only 2 exciting moves of the entire batch!

I usually end these with two games, but this third one is exemplary:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "10"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A00"]
[Result "1/2-1/2"]

1. g3 d5 2. Nf3 Nf6 3. Bg2 c5 4. O-O e6 5. c4 d4 6. e3 Nc6
7. exd4 cxd4 8. d3 Bd6 9. Na3 O-O 10. Nc2 e5 11. b4 a6
12. Re1 Re8 13. c5 Bf8 14. Bg5 Bf5 15. Nh4 Bd7 16. Nf3 Bf5
17. Nh4 Bd7 18. Nf3 Bf5 1/2-1/2[/pgn]

That's the style of both engines summarized for you, if I didn't know I'd say they prearranged this game behind my back!

Wait, all games were drawn except for one, where is it? It was a game where Ethereal outplays Xiphos into a Queen and Pawn vs. Rook endgame and then Etheral becomes Leela and trolls Xiphos for many moves (no kidding, that game took 2 hours to run! ), it's not included because it shows strength mostly, and wasn't fair for Xiphos anyway.

So Xiphos has the better style due to those examples shown, just for a little bit. No exchange sacrifices or huge material imbalances which make these kind of tests fun. I wonder if Ethereal would play exciting chess when messing with her Contempt settings, which seem advanced, and which could help considering strength will be calibrated anyway.

But in the meantime, not much to see here. Move along.

[AdminX]

if I didn't know I'd say they prearranged this game behind my back!

Now that's funny! Thank you for sharing these games.

[abgursu]

See here for context.

The top 2 chess engines I never tested were Xiphos and Ethereal, arguably the top 7 and 6 chess engines.

As usual I ran a match of 10 games after calibrating their strength so they played at the same level and could showcase their styles. (PGN for collectors)

Contrary to other tests of this kind I've run, there's not much to talk about regarding style. No king attacks or losing in amazing fashion, most of the time the engines seemed to just be happy with a draw.

I've chosen to highlight 3 games:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "7"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A04"]
[Result "1/2-1/2"]

1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Bb4 5. a3 Bxc3+ 6. bxc3
O-O 7. e3 c5 8. Bb2 Qa5 9. Nd2 cxd4 10. cxd4 Bd7 11. Qc2
Rc8 12. Bc3 Qa6 13. cxd5 Qa4 14. Qxa4 Bxa4 15. Bb2 Nxd5
16. Bd3 f5 17. O-O a6 18. Rfc1 Nc6 19. Be2 Na5 20. g4 g6
21. g5 Bc2 22. a4 Rc6 23. Bf1 Rac8 24. Bb5 R6c7 25. Be2 Bb3
26. Rxc7 Rxc7 27. Ra3 Bc2 28. h4 Be4 29. Bd1 b6 30. Kf1 Nc4
31. Nxc4 Rxc4 32. Bb3 Rc7 33. Ke2 Bc2 34. Bxd5 exd5 35. a5
b5 36. Rc3 Rxc3 37. Bxc3 Be4 38. Be1 Kf7 39. Bb4 Ke6
40. Bc5 Kd7 41. Ba3 Ke6 42. Bf8 f4 43. f3 Bc2 44. exf4 Kf5
45. Bd6 Bb3 46. Kd2 Ba4 47. Ke2 Bb3 48. Kd2 Ba4 49. Ke2 Bb3
1/2-1/2
[/pgn]

[d]2r3k1/1p5p/p1r1p1p1/nB1n1pP1/P2P4/4P3/1BbN1P1P/R1R3K1 b - -

Xiphos dares Ethereal into taking the Bb5 Bishop, but of course Ethereal doesn't bite.

[pgn]
[Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "2"]
[White "Ethereal"]
[Black "Xiphos"]
[ECO "A40"]
[Result "1/2-1/2"]

1. d4 d5 2. Nf3 Nf6 3. c4 e6 4. Nc3 c6 5. e3 Nbd7 6. Qc2
Bd6 7. Bd3 dxc4 8. Bxc4 O-O 9. O-O b5 10. Bd3 Bb7 11. e4 e5
12. dxe5 Nxe5 13. Nxe5 Bxe5 14. h3 b4 15. Na4 Bd4 16. e5
Bxe5 17. Nc5 Bc8 18. Re1 Qc7 19. Bd2 a5 20. a3 bxa3
21. Rxa3 h6 22. Ra4 Re8 23. b4 Bxh3 24. bxa5 Bc8 25. Ne4
Nd5 26. Rc4 Re6 27. Rc1 Nf4 28. Nc5 Re7 29. Bf1 Ng6 30. Na4
Bb7 31. Nc5 Bc8 32. Na4 Bb7 33. Nc5 Bc8 1/2-1/2
[/pgn]

[d]r3r1k1/2q2pp1/2p2n1p/p1N1b3/RP6/3B3b/2QB1PP1/4R1K1 w - -

Xiphos sacrifices the Bishop for the h3 pawn, but Ethereal doesn't take it and the game ends shortly after.

And that's it.

Yeah, those were the only 2 exciting moves of the entire batch!

I usually end these with two games, but this third one is exemplary:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "10"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A00"]
[Result "1/2-1/2"]

1. g3 d5 2. Nf3 Nf6 3. Bg2 c5 4. O-O e6 5. c4 d4 6. e3 Nc6
7. exd4 cxd4 8. d3 Bd6 9. Na3 O-O 10. Nc2 e5 11. b4 a6
12. Re1 Re8 13. c5 Bf8 14. Bg5 Bf5 15. Nh4 Bd7 16. Nf3 Bf5
17. Nh4 Bd7 18. Nf3 Bf5 1/2-1/2[/pgn]

That's the style of both engines summarized for you, if I didn't know I'd say they prearranged this game behind my back!

Wait, all games were drawn except for one, where is it? It was a game where Ethereal outplays Xiphos into a Queen and Pawn vs. Rook endgame and then Etheral becomes Leela and trolls Xiphos for many moves (no kidding, that game took 2 hours to run! ), it's not included because it shows strength mostly, and wasn't fair for Xiphos anyway.

So Xiphos has the better style due to those examples shown, just for a little bit. No exchange sacrifices or huge material imbalances which make these kind of tests fun. I wonder if Ethereal would play exciting chess when messing with her Contempt settings, which seem advanced, and which could help considering strength will be calibrated anyway.

But in the meantime, not much to see here. Move along.

Thanks for sharing these games! To be honestly, I like Xiphos' style more than Ethereal's, while I like Ethereal more than Xiphos.

[chysiddh14]

etheral simple forward effective

[Alayan]

Neither Xiphos or Ethereal have a speculative play style, so if you adjust them to be about equally strong and play games from short balanced openings, the game won't often feature exciting moves and positions.

You need some source of imbalance. The start position, the difference in strength, and the difference in playstyle are potential sources. A short TC that make mistakes more frequent can also help in this regard.

Take a Leela-Allie game from the start position, it will look boring because neither will really be able to surprise the other, and the opportunity for sound but exciting moves won't appear. While Stockfish-Leela has more potential for interesting positions because they have much more disagreement even when adjusted for equal strength.

Here is a bullet game of Ethereal 12.24 @ 90s+0.9 vs Stockfish20191117 @ 30s+0.3 that Ethereal won, you get an exchange sac :

[pgn][Event "Ethereal_test"]
[Site "https://lichess.org/s7sayD4Q"]
[Date "2020.06.14"]
[Round "13"]
[White "Stockfish 17/11/2019"]
[Black "Ethereal"]
[Result "0-1"]
[WhiteElo "?"]
[BlackElo "?"]
[Variant "From Position"]
[TimeControl "-"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkb1r/p1pppppp/5n2/1p6/8/P6P/1PPPPPP1/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 a6 2. Bf4 Bb7 3. e3 e6 4. Nf3 d5 5. Be2 c5 6. O-O Nc6 7. c3 c4 8. Nbd2 Na5 9. b4 cxb3 10. a4 b2 11. Ra2 Nc4 12. Qb3 Bc6 13. Nxc4 bxc4 14. Qc2 Nd7 15. Rb1 Be7 16. a5 O-O 17. Rbxb2 Re8 18. Rb1 Bb5 19. Nh2 Nf8 20. Bh5 Bd6 21. g3 Rb8 22. Bf3 Bxf4 23. exf4 Re7 24. Ng4 Reb7 25. Qc1 Ng6 26. Ne3 Ne7 27. Nc2 Bc6 28. Nb4 Nf5 29. Qe1 Rxb4 30. Rxb4 Rxb4 31. cxb4 Nxd4 32. Bd1 Nf5 33. g4 Ne7 34. Qc3 Bb5 35. Qd4 Qd7 36. Kf1 h5 37. Ra1 Nc6 38. Qc5 d4 39. Ke1 c3 40. Bf3 d3 41. Qxc3 Nd4 42. Bd1 Qd6 43. Rc1 hxg4 44. hxg4 Qxf4 45. Qd2 Nf3+ 46. Bxf3 Qxf3 47. Qe3 Qxg4 48. f4 Qg2 49. Rc8+ Kh7 50. Kd1 e5 51. Rc5 Qh1+ 52. Kd2 Qh2+ 53. Kc1 exf4 54. Qe4+ Kh6 55. Qe1 f6 56. Qd1 Qh3 57. Kb2 g5 58. Rxb5 axb5 59. a6 Qg2+ 60. Kc3 Qc6+ 61. Kxd3 Qxa6 62. Qh1+ Kg7 63. Qd5 Qa3+ 64. Kc2 Qxb4 65. Qd7+ Kg6 66. Qe8+ Kf5 67. Qc8+ Ke5 68. Qe8+ Kd4 69. Qd8+ Ke3 70. Qd3+ Kf2 71. Qf5 f3 72. Qxf6 g4 73. Kd3 Qc4+ 74. Kd2 Kg2 75. Qb6 f2 76. Qb7+ Kg3 77. Qb8+ Kh3 78. Qh8+ Kg2 79. Qa8+ Kh2 80. Qa1 g3 81. Ke3 f1=Q 82. Qxf1 Qxf1 83. Kd4 g2 84. Kd5 g1=Q 85. Ke6 Qd4 86. Ke7 Qff6+ 87. Ke8 Qdd8# { Black wins by checkmate. } 0-1[/pgn]

It's not hand-picked, just the first win in 20 games I generated to get evaluation tweak ideas.

[BrendanJNorman]

See here for context.

The top 2 chess engines I never tested were Xiphos and Ethereal, arguably the top 7 and 6 chess engines.

As usual I ran a match of 10 games after calibrating their strength so they played at the same level and could showcase their styles. (PGN for collectors)

Contrary to other tests of this kind I've run, there's not much to talk about regarding style. No king attacks or losing in amazing fashion, most of the time the engines seemed to just be happy with a draw.

I've chosen to highlight 3 games:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "7"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A04"]
[Result "1/2-1/2"]

1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Bb4 5. a3 Bxc3+ 6. bxc3
O-O 7. e3 c5 8. Bb2 Qa5 9. Nd2 cxd4 10. cxd4 Bd7 11. Qc2
Rc8 12. Bc3 Qa6 13. cxd5 Qa4 14. Qxa4 Bxa4 15. Bb2 Nxd5
16. Bd3 f5 17. O-O a6 18. Rfc1 Nc6 19. Be2 Na5 20. g4 g6
21. g5 Bc2 22. a4 Rc6 23. Bf1 Rac8 24. Bb5 R6c7 25. Be2 Bb3
26. Rxc7 Rxc7 27. Ra3 Bc2 28. h4 Be4 29. Bd1 b6 30. Kf1 Nc4
31. Nxc4 Rxc4 32. Bb3 Rc7 33. Ke2 Bc2 34. Bxd5 exd5 35. a5
b5 36. Rc3 Rxc3 37. Bxc3 Be4 38. Be1 Kf7 39. Bb4 Ke6
40. Bc5 Kd7 41. Ba3 Ke6 42. Bf8 f4 43. f3 Bc2 44. exf4 Kf5
45. Bd6 Bb3 46. Kd2 Ba4 47. Ke2 Bb3 48. Kd2 Ba4 49. Ke2 Bb3
1/2-1/2
[/pgn]

[d]2r3k1/1p5p/p1r1p1p1/nB1n1pP1/P2P4/4P3/1BbN1P1P/R1R3K1 b - -

Xiphos dares Ethereal into taking the Bb5 Bishop, but of course Ethereal doesn't bite.

[pgn]
[Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "2"]
[White "Ethereal"]
[Black "Xiphos"]
[ECO "A40"]
[Result "1/2-1/2"]

1. d4 d5 2. Nf3 Nf6 3. c4 e6 4. Nc3 c6 5. e3 Nbd7 6. Qc2
Bd6 7. Bd3 dxc4 8. Bxc4 O-O 9. O-O b5 10. Bd3 Bb7 11. e4 e5
12. dxe5 Nxe5 13. Nxe5 Bxe5 14. h3 b4 15. Na4 Bd4 16. e5
Bxe5 17. Nc5 Bc8 18. Re1 Qc7 19. Bd2 a5 20. a3 bxa3
21. Rxa3 h6 22. Ra4 Re8 23. b4 Bxh3 24. bxa5 Bc8 25. Ne4
Nd5 26. Rc4 Re6 27. Rc1 Nf4 28. Nc5 Re7 29. Bf1 Ng6 30. Na4
Bb7 31. Nc5 Bc8 32. Na4 Bb7 33. Nc5 Bc8 1/2-1/2
[/pgn]

[d]r3r1k1/2q2pp1/2p2n1p/p1N1b3/RP6/3B3b/2QB1PP1/4R1K1 w - -

Xiphos sacrifices the Bishop for the h3 pawn, but Ethereal doesn't take it and the game ends shortly after.

And that's it.

Yeah, those were the only 2 exciting moves of the entire batch!

I usually end these with two games, but this third one is exemplary:

[pgn][Event "XE"]
[Site "Merah"]
[Date "2020.06.19"]
[Round "10"]
[White "Xiphos"]
[Black "Ethereal"]
[ECO "A00"]
[Result "1/2-1/2"]

1. g3 d5 2. Nf3 Nf6 3. Bg2 c5 4. O-O e6 5. c4 d4 6. e3 Nc6
7. exd4 cxd4 8. d3 Bd6 9. Na3 O-O 10. Nc2 e5 11. b4 a6
12. Re1 Re8 13. c5 Bf8 14. Bg5 Bf5 15. Nh4 Bd7 16. Nf3 Bf5
17. Nh4 Bd7 18. Nf3 Bf5 1/2-1/2[/pgn]

That's the style of both engines summarized for you, if I didn't know I'd say they prearranged this game behind my back!

Wait, all games were drawn except for one, where is it? It was a game where Ethereal outplays Xiphos into a Queen and Pawn vs. Rook endgame and then Etheral becomes Leela and trolls Xiphos for many moves (no kidding, that game took 2 hours to run! ), it's not included because it shows strength mostly, and wasn't fair for Xiphos anyway.

So Xiphos has the better style due to those examples shown, just for a little bit. No exchange sacrifices or huge material imbalances which make these kind of tests fun. I wonder if Ethereal would play exciting chess when messing with her Contempt settings, which seem advanced, and which could help considering strength will be calibrated anyway.

But in the meantime, not much to see here. Move along.

The games are simply too "perfect"...you rarely see brilliancies between Carlsen and Caruana either although even that is much lower level than these guys.

What I usually do to test "style" is place Xiphos or Ethereal (or whomever) in gauntlets against a host of interesting 2800-2900 engines (Thinker, Rodent, Frenzee, Junior, Hiarcs, Fizbo, ChessBrainVB, Gaviota, etc) and make the opening uniform, but sharp and flexible (maybe a Najdorf starting from 6.Bg5 for example, with no opening books).

Throw both engines in the same gauntlet and see who produces the most beautiful smashes against weaker opponents.

[Frank Quisinsky]

Hi there,

for the middlegame is Xiphos more speculative, more aggressive.
Not to compare with Fizbo but very strong too.

On the other hand, the more Elo from Xiphos for the newer versions comes "unfortunately" from a higher endgame strength. So older Xiphos versions have also a nice middlegame style.

I make only middlegame stats and try to find out here a little bit.
Endgame isn't interesting because all the TOP-Engines plays the playing phase clearly better as the other playing phases.
So the endgame is +-100 Elo allways the same from engines with a CEGT playing strength 3.050 - 3.200 with one core.

Very boring for myself, so I am looking only in the early and late middlegame.
Much more interesting!

Best
Frank

Really a great style in middlegames have Fritz 17 (Ginkgo).
But I await that after Gromit / Anaconda in the past have also a great attacker style.
I think Ginkgo is stronger as Fizbo in king attacks. Fizbo is more speculative if I compare to Ginkgo.

Topic is Xiphos / Ethereal!
Xiphos have definitiv a nice style. For Ethereal I can't see that the middlegame is for the high Elo the engine have "extra strong". I think a average style to the Elo the engine have.

[carldaman]

The games are simply too "perfect"...you rarely see brilliancies between Carlsen and Caruana either although even that is much lower level than these guys.

What I usually do to test "style" is place Xiphos or Ethereal (or whomever) in gauntlets against a host of interesting 2800-2900 engines (Thinker, Rodent, Frenzee, Junior, Hiarcs, Fizbo, ChessBrainVB, Gaviota, etc) and make the opening uniform, but sharp and flexible (maybe a Najdorf starting from 6.Bg5 for example, with no opening books).

Throw both engines in the same gauntlet and see who produces the most beautiful smashes against weaker opponents.

Long live the 'cannon fodder' engines! That's exactly what I do when I need a quick assessment of engine style. If I have the time, I may even try some sparring games against the more promising engine opponents, to see how they beat me.

[AndrewGrant]

Engine "style" is a human concept. People attribute fantastic "style" to the engines they happen to like. They invent a "style" out of a subset of the games they see. Time and time again it has been shown that people cant distinguish games played by different engines like SF and Leela.
Engines don't have a style. They have a raw calculation. Nothing more.

[carldaman]

There's raw calculation, but there are also ways to prioritize certain things - preferences or biases if you will. There are many engines that play as if not to lose. Not very exciting to many of us. Others take chances, playing speculatively, in stark contrast.

There's no denying that, and we shouldn't lose sight of such distinctions.