If you are interested....

Alexandria-4.0-x86-64-bmi2
==================

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases (only for Revenge_1.0_x64_bmi2)
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result:





+21 -17 =62

+14 Elo points in 100 games for Alexandria-4.0-x86-64-bmi2

Revenge 1.0 x64 has 3445 Elo points (CCRL Blitz rating list).

Alexandria 3.5 x64 has 3386 Elo points (CCRL Blitz rating list).

A nice improvement.

ClasicAra 1.0.5 (CPU version)
==================

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases (only for ClassicAra 1.0.5)
  -NNs: MadChess 3.1 x64 is a HCE chess engine; ClassicAra 1.0.5 uses a DCNN & a Monte-Carlo Graph Search based on a
        directed acyclic graph (instead of MCTS)
  -Ponder: off
  -OS: Windows 10 Home.

The final result:





+47 -22 =31

+92 Elo points in 100 games for ClassicAra 1.0.5

MadChess 3.1 x64 has 2698 Elo points (CCRL Blitz rating list).

ClassicAra 0.9.5.post0 x64 (the last version tested by CCRL) has 2654 Elo points (CCRL Blitz rating list).

An improvement of about 150 Elo points for this innovative chess engine. I wonder how well the GPU version performs (I have only Intel integrated graphics on my PC destined to computer chess). Impressive! Congratulations to the authors for not abandoning the project!

[pgn]
[Event "NN Test383"]
[Site "ISR 3"]
[Date "2023.08.11"]
[Round "35"]
[White "ClassicAra 1.0.5"]
[Black "MadChess 3.1 x64"]
[Result "1-0"]
[BlackElo "2200"]
[ECO "B01"]
[Opening "Scandinavian"]
[Time "15"]
[Variation "Marshall, 4.Nf3 g6"]
[WhiteElo "2200"]
[TimeControl "240+2"]
[Termination "normal"]
[PlyCount "58"]
[WhiteType "program"]
[BlackType "program"]

1. e4 d5 2. exd5 Nf6 3. d4 Nxd5 4. Nf3 g6 5. c4 {(c2-c4 Nd5-b6 Nb1-c3
Bf8-g7 c4-c5 Nb6-d5 Bf1-c4 c7-c6 O-O O-O Rf1-e1 Bc8-g4 h2-h3 Bg4xf3 Qd1xf3
e7-e6) +0.49/16 9} Nf6 {(Nd5-f6 Nb1-c3 Bc8-f5 Bc1-f4 Rh8-g8 h2-h3 h7-h5
d4-d5 c7-c5 Nf3-e5 Nb8-d7 Ne5xd7 Bf5xd7 Bf1-d3 h5-h4) -0.81/15 7} 6. Nc3
{(Nb1-c3 Bf8-g7 Bf1-e2 O-O O-O c7-c6 h2-h3 Bc8-f5 Bc1-e3 Nf6-e4 Qd1-b3
Ne4xc3 b2xc3 Qd8-c7 Rf1-e1 Nb8-d7) +0.65/16 9} Bf5 {(Bc8-f5 Bc1-f4 Rh8-g8
h2-h3 c7-c6 Bf1-e2 Bf8-g7 O-O Nb8-d7 g2-g4 Bf5-e6 d4-d5 c6xd5 c4xd5)
-0.95/14 6} 7. Qb3 {(Qd1-b3 Qd8-c8 h2-h3 Bf8-g7 Bc1-e3 O-O Bf1-e2 c7-c6 O-O
Nb8-d7 Rf1-d1 Rf8-e8) +1.06/12 6} b6 {(b7-b6 Bf1-e2 Bf8-g7 O-O O-O Qb3-a3
a7-a5 d4-d5 Nb8-a6 Nf3-d4 Bf5-d7 Nd4-c6 Bd7xc6 d5xc6) -0.94/14 6} 8. h3
{(h2-h3 Bf8-g7 g2-g4 Bf5-c8 Bf1-g2 Bc8-b7 O-O O-O Bc1-f4 Nb8-c6 Ra1-d1
Nc6-a5) +1.36/12 6} Bg7 {(Bf8-g7 g2-g4 Bf5-c8 Bc1-f4 Bc8-b7 d4-d5 O-O
Ra1-d1 Nb8-a6 Qb3-a3 Rf8-e8 Nf3-e5 Na6-c5 g4-g5) -0.76/14 13} 9. g4 {(g2-g4
Bf5-c8 Bf1-g2 Bc8-b7 O-O O-O Bc1-f4 Nb8-c6 Ra1-d1 Nc6-a5 Qb3-a4 Bb7-c6
Qa4-c2) +1.37/13 4} Bc8 {(Bf5-c8 Bc1-f4 Bc8-b7 d4-d5 a7-a6 Ra1-d1 Nb8-d7
Qb3-c2 O-O Bf1-e2 Ra8-c8 Nf3-d4 Nd7-c5 O-O Nf6-d7) -0.86/15 6} 10. Bg2
{(Bf1-g2 Bc8-b7 O-O O-O Bc1-f4 Nb8-c6 Ra1-d1 Nc6-a5 Qb3-a4 Bb7-c6 Qa4-c2
Bc6xf3 Bg2xf3 c7-c6) +1.40/14 9} Bb7 {(Bc8-b7 Bc1-f4 h7-h6 Ra1-d1 g6-g5
Bf4-e3 Nb8-d7 d4-d5 c7-c6 O-O Qd8-c7 d5xc6 Qc7xc6 Nc3-b5) -0.88/14 6} 11.
O-O {(O-O O-O Bc1-f4 Nb8-c6 Ra1-d1 Nc6-a5 Qb3-a4 Bb7-c6 Qa4-c2 Bc6xf3
Bg2xf3 c7-c6) +1.53/12 3} a6 {(a7-a6 Bc1-f4 h7-h6 c4-c5 O-O Rf1-e1 g6-g5
Bf4-e5 Nb8-d7 Ra1-d1 Ra8-b8 Qb3-c2 Bb7-a8 Be5xf6) -1.08/14 16} 12. Re1
{(Rf1-e1 O-O Bc1-f4 e7-e6 Ra1-d1 Nb8-d7 d4-d5 e6xd5 c4xd5 Nd7-c5 Qb3-c4
a6-a5 Nf3-e5 Nf6-e8 Ne5-c6) +2.34/15 6} h6 {(h7-h6 c4-c5 O-O Nf3-e5 Bb7xg2
Kg1xg2 e7-e6 Bc1-e3 b6xc5 Ra1-d1 Nf6-d7 Ne5xd7 Nb8xd7) -1.17/13 5} 13. Bf4
{(Bc1-f4 O-O Ra1-d1 Nb8-d7 Nf3-e5 Bb7xg2 Kg1xg2 e7-e6 Ne5-c6 Qd8-c8 Nc6-e7+
Kg8-h7) +3.04/12 8} g5 {(g6-g5 Bf4-g3 Nb8-d7 Ra1-d1 Ra8-b8 d4-d5 Nd7-c5
Qb3-c2 Nf6-d7 Qc2-f5 Bg7xc3 b2xc3 Ke8-f8) -1.49/13 4} 14. Bg3 {(Bf4-g3 O-O
Ra1-d1 Nb8-d7 c4-c5 e7-e6 Nf3-e5 Bb7xg2 Kg1xg2 Nd7xe5 d4xe5 Nf6-d7 c5-c6)
+3.87/13 4} Nbd7 {(Nb8-d7 Ra1-d1 h6-h5 c4-c5 Bb7xf3 Bg2xf3 b6xc5 Bf3-d5
Nf6xd5 Nc3xd5 Ra8-b8 Qb3-c2 c5xd4) -1.52/13 5} 15. Qa3 {(Qb3-a3 c7-c5
Ra1-d1 O-O d4xc5 b6xc5 Nf3-e5 Bb7xg2 Kg1xg2 Qd8-c8 Ne5xd7 Nf6xd7 Nc3-d5
Qc8-b7 Qa3-f3) +4.06/15 7} Rc8 {(Ra8-c8 Nf3-h4 Bb7xg2 Nh4-f5 Bg7-f8 Kg1xg2
e7-e6 Qa3xa6 h6-h5 g4xh5 Rh8xh5 d4-d5 Nd7-c5 Qa6-b5+) -1.80/14 6} 16. Rad1
{(Ra1-d1 c7-c5 d4xc5 Rc8xc5 Nc3-a4 Rc5xc4 Nf3-e5 Bb7xg2 Ne5xc4 Bg2-d5
Nc4-d6+) +4.84/11 6} h5 {(h6-h5 Nf3xg5 Bb7xg2 Kg1xg2 h5xg4 h3-h4 Rh8-h5
Re1-e3 Nd7-f8 Rd1-e1 Nf8-g6 Kg2-g1 Bg7-f8 Qa3-a4+ c7-c6) -1.51/15 6} 17. d5
{(d4-d5 h5xg4 h3xg4 Nf6xg4 Nc3-e4 Ng4-e5 Nf3xe5 Nd7xe5 Ne4xg5 Ne5xc4 Qa3-f3
Nc4-d6) +4.32/12 10} hxg4 {(h5xg4 Nf3xg5 g4xh3 Ng5xh3 Bg7-f8 Qa3-a4 Rh8-g8
Nh3-f4 Rg8-h8 Nf4-e6 f7xe6 d5xe6 Bb7xg2 e6xd7+ Nf6xd7) -1.03/15 5} 18. hxg4
{(h3xg4 Nf6xg4 Nc3-e4 Ng4-e5 Nf3xe5 Nd7xe5 Ne4xg5 Ne5xc4 Qa3-f3 Nc4-d6)
+4.48/10 2} Nxg4 {(Nf6xg4 Nf3xg5 Nd7-e5 Qa3-a4+ Ke8-f8 Qa4-c2 Qd8-d6 Qc2-f5
Qd6-f6 Ng5-h7+ Rh8xh7 Qf5xh7 Ne5-f3+ Bg2xf3 Qf6xf3) -1.83/15 7} 19. Ne4
{(Nc3-e4 Nd7-e5 Ne4xg5 Ne5xc4 Qa3-b4 Nc4-d6 Qb4xg4 Bg7-f6 Nf3-e5 Rh8-g8
Ne5xf7) +4.68/11 4} Nge5 {(Ng4-e5 Nf3xe5 Bg7xe5 Ne4xg5 Rh8-h5 Ng5xf7 Ke8xf7
Qa3-f3+ Kf7-g6 Bg3xe5 Rh5xe5 Qf3-f4 Re5xe1+) -2.27/13 10} 20. Nxe5 {(Nf3xe5
Nd7xe5 Qa3-c3 f7-f6 Ne4xg5 f6xg5 Bg3xe5 Bg7xe5 Qc3xe5 Qd8-d6 Qe5xh8+)
+4.15/11 11} Bxe5 {(Bg7xe5 d5-d6 Bb7xe4 Bg3xe5 Nd7xe5 Re1xe4 c7xd6 Re4xe5
Rc8xc4 Qa3xa6 Rh8-h4 Re5xg5 Ke8-f8 Qa6-a8 Qd8xa8 Bg2xa8 Kf8-e8) -4.80/17 8}
21. Nxg5 {(Ne4xg5 Be5xg3 Qa3xg3 Nd7-f6 d5-d6 c7xd6 Bg2xb7 Rc8xc4 Rd1xd6
Qd8xd6 Qg3xd6 Rc4-g4+ Bb7-g2 Rg4xg2+) +4.68/14 4} Bf6 {(Be5-f6 Ng5-e6 f7xe6
d5xe6 Bb7xg2 e6xd7+ Ke8-f7 d7xc8Q Qd8xc8 Kg1xg2 Qc8-h3+ Kg2-f3 Qh3-f5+
Kf3-e2 Qf5-g4+ Qa3-f3 Qg4xc4+) -6.83/17 13} 22. Ne6 {(Ng5-e6 f7xe6 d5xe6
Bb7xg2 Kg1xg2 Nd7-c5 Rd1xd8+ Rc8xd8 b2-b4 Rd8-d3 Qa3-c1 Rd3-c3 Qc1-f4)
+5.59/13 11} fxe6 {(f7xe6 d5xe6 Bb7xg2 Kg1xg2 Ke8-f8 e6xd7 Rc8-a8 Bg3-e5
Kf8-f7 Qa3-f3 b6-b5 Be5xf6 e7xf6 Qf3-d5+ Kf7-g7 c4xb5 a6xb5 Re1-h1 Rh8-g8
Kg2-f1) -6.94/20 9} 23. dxe6 {(d5xe6 Bb7xg2 Kg1xg2 Nd7-c5 Rd1xd8+ Rc8xd8
b2-b4 Rd8-d3 Qa3-c1 Rd3-c3 Qc1-f4) +5.76/11 1} Bxg2 {(Bb7xg2 Kg1xg2 Ke8-f8
Qa3-f3 Rh8-g8 e6xd7 Rc8-b8 Kg2-f1 Rg8-g6 Qf3-c6 Rg6xg3 f2xg3 Bf6xb2 Re1-e2
Bb2-f6 Kf1-f2 Bf6-g7 Qc6-d5 b6-b5 Qd5-f5+ Bg7-f6) -8.08/21 19} 24. Kxg2
{(Kg1xg2 Nd7-c5 Rd1xd8+ Rc8xd8 Qa3-f3 Ke8-f8 Bg3xc7 Rh8-g8+) +6.16/8 4} Kf8
{(Ke8-f8 Qa3-f3 Rh8-g8 e6xd7 Rc8-b8 Kg2-f1 Rg8-g6 Qf3-c6 Rg6xg3 f2xg3
Bf6xb2 Re1-e2 Bb2-f6 Kf1-f2 Bf6-g7 Qc6-g6 Rb8-a8 Kf2-g2) -8.41/21 7} 25.
exd7 {(e6xd7 Rc8-a8 Qa3-f3 Ra8-a7 Re1-e6 Kf8-f7 Qf3-d5 Kf7-g7 Re6-e4
Kg7-f8) +6.70/10 6} Ra8 {(Rc8-a8 Bg3-e5 Kf8-f7 Qa3-f3 e7-e6 Kg2-f1 b6-b5
c4-c5 Rh8-f8 Qf3-h5+ Kf7-g7 Rd1-d3 Bf6xe5 Re1xe5 Rf8-f6 Rd3-g3+ Kg7-f8
Qh5-h7) -9.04/18 6} 26. Qf3 {(Qa3-f3 Ra8-a7 Re1-e6 Kf8-f7 Qf3-d5 Kf7-g7
Re6-e4 Kg7-f8 Re4-g4) +6.53/9 13} Kf7 {(Kf8-f7 Re1-h1 c7-c6 Rh1xh8 Qd8xh8
Bg3-c7 Ra8-g8+ Kg2-f1 Qh8-h4 d7-d8Q Rg8xd8 Rd1xd8 Qh4xc4+ Kf1-g2 Kf7-g6
Qf3-d3+ Qc4xd3 Rd8xd3 Bf6xb2 Bc7xb6) -11.12/20 25} 27. Qd5+ {(Qf3-d5+
Kf7-g7 Re1-e4 Ra8-a7 Re4-g4+ Kg7-f8 Qd5-e6 c7-c5 Rd1-d5 Ra7-b7) +6.64/10
13} Kg7 {(Kf7-g7 Bg3-e5 Rh8-h6 Be5xc7 Qd8xc7 Qd5xa8 Qc7-f4 Kg2-f1 Qf4xc4+
Re1-e2 e7-e5 b2-b3 Qc4-c5 d7-d8Q Bf6xd8 Qa8xd8 Rh6-h1+ Kf1-g2 Qc5-c6+)
-11.50/19 12} 28. Re4 {(Re1-e4 Rh8-h6 Re4-g4+ Rh6-g6 Rg4xg6+ Kg7xg6 Bg3xc7
Qd8xc7 Qd5xa8 Qc7xc4 d7-d8Q Qc4-g4+) +6.70/12 13} c6 {(c7-c6 Qd5xc6 Ra8-a7
Bg3-e5 Qd8-a8 Qc6xa8 Ra7xa8 Be5-c7 e7-e5 Re4-g4+ Kg7-f7 Bc7xb6 Bf6-d8 c4-c5
Rh8-h6 Rd1-h1 Rh6xh1 Kg2xh1 Kf7-e7 Bb6xd8+ Ra8xd8) -14.95/21 14} 29. Qxc6
{(Qd5xc6 Ra8-a7 Re4-g4+ Kg7-f7 Qc6-d5+ e7-e6 Qd5-e4 Ra7xd7 Rd1xd7+ Qd8xd7
Qe4-g6+ Kf7-e7 Rg4-f4 Qd7-c6+) +6.92/14 10} Ra7 {(Ra8-a7 Bg3-e5 Qd8-a8
Qc6xa8 Ra7xa8 Be5-c7 e7-e5 Re4-g4+ Kg7-f7 Kg2-f3 b6-b5 c4-c5 Ra8-a7 d7-d8Q
Bf6xd8 Bc7xd8 Rh8-h5 Kf3-e4 Kf7-f8 Ke4-f3) -14.46/21 20 Black resigns} 1-0
[/pgn]

Unfortunately, the current version of Classic Ara is no stronger than version 0.9.5. According to the author's tests, an increase over the previous version on CPU only occurs with extremely short thinking times (30 ms..), otherwise the increase is just 6.9 Elo. Even with a thinking time of 4 min + 2 sec, both the CPU and GPU versions are tactically bad.

Dio wrote:

Unfortunately, the current version of Classic Ara is no stronger than version 0.9.5. According to the author's tests, an increase over the previous version on CPU only occurs with extremely short thinking times (30 ms..), otherwise the increase is just 6.9 Elo. Even with a thinking time of 4 min + 2 sec, both the CPU and GPU versions are tactically bad.

In my games (so far) version 1.0.5 performs much better than 0.9.5. I don't agree self-play or matches between versions of the same engine.

since you haven't tested version 0.9.5 against Madchess, there is no need to comment.

when I see selfplay results of 100 Elo, I expect an increase of maybe 50 to 70 Elo, never more.

R_Silvian wrote:

ClasicAra 1.0.5 (CPU version)
==================
I wonder how well the GPU version performs (I have only Intel integrated graphics on my PC destined to computer chess).

Broke off  already after 45 games, getting too much difference. Maybe Booot simply was too strong with 16 threads of 16x3.5GHz CPU, UHO 6mvs. 110-119 making the difference even bigger, and maybe the Ara- setting with Batch Size 64 wasn't the best for the STC of 3'+1" neither. GPU was 3070ti Nvidia, 2 threads of CPU, cutechess- GUI, ponder off:

Score of ClassicAra_1.0.5_Win_TensorRT vs booot72_avx2_pext: 2 - 28 - 15 [0.211]
Elo difference: -229.0 +/- 91.9, LOS: 0.0 %, DrawRatio: 33.3 %
46 of 100 games finished.

Boot with 16 threads is of course much too strong even against Ara on GPU. Even Booot with 1 thread should be stronger....

This time I broke off after 75 games, because Texel 1.09 better had played SMP, yet this time it ran single threaded (3.5GHz). Ara again with 3070ti,  3'+1", UHO 6mvs. 110-119:

Score of texel64-avx2-bmi vs ClassicAra: 10 - 35 - 30 [0.333]
Elo difference: -120.4 +/- 62.7, LOS: 0.0 %, DrawRatio: 40.0 %
76 of 100 games finished.

Clover.6.0-avx2:
==========

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases for both engines
-NNs: defaults
-Ponder: off
-OS: Windows 10 Home.

The final result:





+11 -8 =81

+14 Elo points in 100 games for Clover.6.0-avx2

For Clover 5 (the previous version) and Caissa 1.11 ratings - see, please, CCRL Blitz & Stefan Pohl rating lists (both from 13.08.2023):





Maybe a little progress.

Lc0-v0.30.0-dnnl
===========

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases (only for Lc0-v0.30.0-dnnl)
  -DCNN & NNUE: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result in my match:





+27 -13 =60

+49 Elo points in 100 games for Lc0-v0.30.0-dnnl

Mantissa 3.7.2 avx2 has 3281 Elo points (Stefan Pohl rating list).

Lc0 0.30 dnll 791556 has 3269 Elo points (Stefan Pohl rating list).

Arasan-24.0-x64-avx2-bmi2
=================

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases (only for Arasan-24.0-x64-avx2-bmi2)
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result in my match:





+23 -8 =69

+56 Elo points in 100 games for Arasan-24.0-x64-avx2-bmi2

Rice 7.0.0 x64 has 3474 Elo points (CCRL Blitz rating list).

Arasan 23.5 x64 has 3460 Elo points (CCRL Blitz rating list).

====================================================
3 time forfeits for Arasan-24.0-x64-avx2-bmi2 (installed like UCI engine).
====================================================

Velvet-v5.3.0-x86_64-avx2
=================

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases (only for Velvet-v5.3.0-x86_64-avx2)
-NNs: defaults
-Ponder: off
-OS: Windows 10 Home.

The final result in my match:





+9 -28 =63

-63 Elo points in 100 games

Black Marlin 8.0 avx2 has 3543 Elo points (Stefan Pohl rating list).

Velvet 5.3.0 avx2 has 3494 Elo points (Stefan Pohl rating list).

StockDory-Starfish-0.1-x86-64-v3
=====================

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases (only for Marvin_6.2.0_x86_64_avx2)
-NNs: defaults
-Ponder: off
-OS: Windows 10 Home.

The final result in my match:





+9 -27 =64

-63 Elo points in 100 games

Marvin 6.2.0 x64 has 3401 Elo points (in CCRL Blitz rating list) & 3419 Elo points (in Stefan Pohl rating list).

StockNemo 5.7.0.0 x64 has 3188 Elo points (in CCRL 40/15 rating list).

StockDory 06072023 x64 has 3248 Elo points (in CCRL 40/15 rating list).

Stock family (Nemo & Dory) wasn't tested by Stefan Pohl (despite the very good results).

For sure an over 3300 Elo chess engine (and a significant progress appears with each version).

REBEL-EAS
=======

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases (only for Clover.6.0-avx2)
-NNs: defaults
-Ponder: off
-OS: Windows 10 Home.

The final result:





+16 -5 =79

+42 Elo points in 100 games for REBEL-EAS

Clover 6 avx2 has 3651 Elo points (Stefan Pohl rating list).

Rebel EAS avx2 has 3662 Elo points (Stefan Pohl rating list).

Nalwald-18-modern
============

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: none
-NNs: default (for StockNemo-5.7.0.0_x64); Nalwald-18-modern-is an HCE chess engine
-Ponder: off
-OS: Windows 10 Home.

The final result in my match:





+30 -17 =53

+49 Elo points in 100 games for Nalwald-18-modern

StockNemo 5.7.0.0 x64 has 3271 Elo points (CCRL Blitz rating list).

Nalwald 18 x64 has 3273 Elo points (CCRL Blitz rating list).

====================

Next:

Stockfish_23090307_x64_bmi2
====================

This is the latest version:



Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases for both engines
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result in my match:





+2 -4 =94

-7 Elo points in 100 games

Stockfish_16_x64_bmi2 has 3702 Elo points (CCRL Blitz rating list).

Dio wrote:

since you haven't tested version 0.9.5 against Madchess, there is no need to comment.

Tested ! ..........and still extensively.





+47 -22 =31

+92 Elo points in 100 games for ClassicAra 1.0.5





+59 -13 =28

+173 Elo points in 100 games for ClassicAra 0.9.5.post0 x64

MadChess 3.1 x64 has 2699 Elo points (CCRL Blitz rating list).

ClassicAra 0.9.5.post0 x64 (the last version tested by CCRL) has 2655 Elo points (CCRL Blitz rating list).

ClassicAra 0.9.5.post0 x64 rests the best ClassicAra chess engine !

One question remains: how did CCRL tested this engine (ClassicAra 0.9.5.post0 x64)? +173 Elo points face to MadChess 3.1 x64 in my 100 games match .... -44 Elo points in CCRL Blitz rating list ! Again, I tested it extensively. Most chess engines between ClassicAra 0.9.5 and MadChess 3.1 (from CCRL Blitz rating list) are easily beaten by ClassicAra 0.9.5!

???



Conditions in my matches:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases (only for ClassicAra 1.0.5 & ClassicAra 0.9.5.post0 x64)
  -NNs: MadChess 3.1 x64 is a HCE chess engine; ClassicAra 1.0.5 & ClassicAra 0.9.5.post0 x64 use DCNNs & Monte-CarloGraph Search based on a directed acyclic
            graph (instead of MCTS)
  -Ponder: off
  -OS: Windows 10 Home.

Stockfish_23091120_x64_bmi2
====================

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases for both engines
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result:





+4 -0 =96

+14 Elo points in 100 games for Stockfish_23091120_x64_bmi2

Stockfish_16_x64_bmi2 has 3706 Elo points (CCRL Blitz rating list).

Stockfish_23091120_x64_bmi2 is the best chess engine at the date on my PC. Exactly like in Stefan Pohl's rating list (https://www.sp-cc.de/).

This is exactly the version that played in my match:

Carp-v3.0.1-win-x86_64-V3
=================

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases (only for Berserk-7-x64-avx2-pext)
-NNs: defaults
-Ponder: off
-OS: Windows 10 Home.

The final result in my match:





+8 -9 =83

+/- 0 Elo points in 100 games

Berserk 7.0 x64 has 3497 Elo points (CCRL Blitz rating list).

Carp 3.0.1 x64 has 3501 Elo points (CCRL Blitz rating list).

RubiChess-230918_x86-64-bmi2
====================

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases for both engines
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result:





+3 -15 =82

-42 Elo points in 100 games

Stefan Pohl rating list:

Sounds likw we have a new champion in C# world.....let's wait a bit.....

Pedantic 0.5.0 x64
============

Match conditions:

-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases (only for Pedantic 0.5.0 x64)
-NNs: both engines are HCE
-Ponder: off
-OS: Windows 10 Home.

The final result in my match:





+54 -22 =24

+115 Elo points in 100 games for Pedantic 0.5.0 x64

Leorik 2.5 x64 has 2938 Elo points (CCRL Blitz rating list).

Pedantic 0.4.1 x64 has 2886 Elo points (CCRL Blitz rating list).

Impressive this new version of Pedantic !

=============

Next:

Pawn_2.0_win_haswell
===============

Match conditions:

  -TC=4'+2"
  -Hash=256 MB
  -GUI: Arena 3.5.1
  -Book: SuperGM_4mvs.abk (8 plies book) for both engines
  -1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
  -TBS: 6-men Syzygy bases for both engines
  -NNs: defaults
  -Ponder: off
  -OS: Windows 10 Home.

The final result in this match:





+6 -28 =66

-78 Elo points in 100 games

PowerFritz 18 avx2 has 3474 Elo points (Stefan Pohl SPCC rating list).

Pawn 1.0 x64 has 3276 Elo points (Stefan Pohl SPCC rating list).

--------------------------------------------------------------------------------------------------

Ginkgo Vienna x64 (PowerFritz 18 AVX2) has 3469 Elo points (CCRL Blitz rating list).

Pawn 1.0 x64 has 3300 Elo points (CCRL Blitz rating list).

An improvement of around 100 Elo points for the new Pawn_2.0_win_haswell.

==========

Next:

The post (with the correct display) has been reinserted below.

This is spam, or ?!