Honey-X5i and Crystal-Honey-X5i Dual Release!

[Ovyron]

One thing that would give Honey the edge over other branches would be an option for showing the scores of alternative moves when using the Tactical setting, since there's no other way to mix excludemoves and MultiPV (I continue to use McCain X with my change when I need a 5th alternative after the main 4 moves from SF have fallen, so I can get the 5th, 6th, 7th, and 8th moves quickly. I couldn't find a way for later versions to do this, so that's why I'm stuck using an old patched version.)

[MikeB]

One thing that would give Honey the edge over other branches would be an option for showing the scores of alternative moves when using the Tactical setting, since there's no other way to mix excludemoves and MultiPV (I continue to use McCain X with my change when I need a 5th alternative after the main 4 moves from SF have fallen, so I can get the 5th, 6th, 7th, and 8th moves quickly. I couldn't find a way for later versions to do this, so that's why I'm stuck using an old patched version.)

You can use mulitipv - which shows all the moves and say of you want is multipv setting of 16, but show only 4, set tactical to 4 and and multipv to 4.
That is what it was designed for, less clutter when using multipv at a high value and you just want see the few best candidates.

tactical =4 and multipv = 4.
dep score nodes time (not shown: tbhits knps seldep)
17 +54.64 9.75M 0:09.09 e4 e6 d4 d5 Nc3 Bb4 exd5 exd5 Be3 Nc6 Nf3 Nf6 Bd3 O-O O-O h6 h3 Be6 Ne2 Bd6 Nf4 Bxf4 Bxf4
17 +53.94 10.4M 0:09.64 c4 e5 Nc3 Nc6 a3 Nf6 Nf3 Be7 e3 d5 d4 exd4 exd4 a6 Be3 O-O Bd3 Be6 Ng5 dxc4 Nxe6 fxe6
17 +53.29 10.1M 0:09.45 d4 e6 e4 d5 Nc3 Bb4 exd5 exd5 Be3 Nc6 Nf3 Nf6 Bd3 O-O O-O Bd6 Nb5 Nb4 Nxd6 Qxd6 Ne5 Nxd3 Qxd3 Be6 Rab1 Ne4 Rfe1 Rad8
17 +52.84 10.5M 0:09.76 e3 d5 Nf3 e6 c4 Nf6 Nc3 Be7 d4 O-O Be2 c5 O-O dxc4 Bxc4 Nc6 Bd3 cxd4 exd4 Qb6 Rb1 Nxd4 Bxh7+ Nxh7
17 +52.71 9.86M 0:09.20 Nf3 Nf6 c4 c5 e3 e6 d4 d5 Nc3 Nc6 cxd5 exd5 Bb5 a6 Bxc6+ bxc6 Ne5 cxd4 Nxc6
16 +53.16 6.74M 0:06.25 e4 e6 d4 d5 e5 c5 c3 Nc6 Nf3 Nge7 Bd3 cxd4 cxd4 Nf5 Bxf5 exf5 Nc3 Be7 Bd2 Be6
16 +52.43 7.12M 0:06.59 Nf3 Nf6 c4 e6 d4 d5 Nc3 Bb4 e3 O-O Bd3 dxc4 Bxc4 c5 O-O cxd4 exd4 Nc6 Bf4 Qb6 Rc1 Rd8
16 +52.30 7.24M 0:06.69 e3 d5 c4 e6 Nf3 Nf6 d4 Be7 Nc3 O-O Bd2 c5 dxc5 Nc6 Be2 dxc4 O-O Bxc5 Bxc4 Bd7 Rc1
16 +52.30 6.67M 0:06.19 d4 e6 Nf3 d5 Bg5 Nf6 e3 Be7 Bd3 c5 O-O c4 Be2 O-O Bxf6 Bxf6 b3
15 +56.79 3.56M 0:03.35 e4 e6 d4 d5 exd5 exd5 Nf3 Nf6 Bd3 Nc6 O-O Bd6 c4 O-O c5 Be7 Nc3 Be6
15 +52.22 3.90M 0:03.66 d4 Nf6 Bf4 e6 e3 c5 Nf3 cxd4 exd4 Nc6 Nc3 Qb6 Nb5 Nd5 Bd2
15 +52.06 4.15M 0:03.88 e3 d5 c4 e6 Nf3 Be7 Nc3 Nf6 d4 O-O Bd3 c5 O-O Nc6 Ne5 cxd4 exd4
15 +51.48 4.07M 0:03.81 Nf3 d5 d4 e6 Bg5 Nf6 e3 c5 c3 Nc6 Nbd2 cxd4 exd4 h6 Bxf6 Qxf6 Bd3
14 +56.06 3.22M 0:03.00 e4 e5 Nf3 Nc6 Bb5 Nf6 O-O Be7 d3 d6 Be3 O-O Nbd2 Bg4 h3 Bxf3 Qxf3
14 +53.25 3.22M 0:03.00 Nf3 d5 d4 e6 Bg5 Be7 Bxe7 Nxe7 e3 c5 Bd3 c4 Be2 Nd7 O-O O-O Nbd2 h6 c3 Qb6
14 +52.84 3.22M 0:03.00 d4 d5 Bf4 Nf6 e3 e6 Nd2 Bd6 Bxd6 cxd6 Ngf3 Bd7 Be2 O-O O-O Nc6
14 +52.67 3.22M 0:03.00 Nc3 d5 d4 Nf6 Bg5 e6 e3 c5 Bxf6 Qxf6 Nf3 Nc6 Bb5

multipv 16
dep score nodes time (not shown: tbhits knps seldep)
17 +54.64 9.75M 0:09.21 e4 e6 d4 d5 Nc3 Bb4 exd5 exd5 Be3 Nc6 Nf3 Nf6 Bd3 O-O O-O h6 h3 Be6 Ne2 Bd6 Nf4 Bxf4 Bxf4
17 +53.94 10.4M 0:09.77 c4 e5 Nc3 Nc6 a3 Nf6 Nf3 Be7 e3 d5 d4 exd4 exd4 a6 Be3 O-O Bd3 Be6 Ng5 dxc4 Nxe6 fxe6
17 +53.29 10.1M 0:09.58 d4 e6 e4 d5 Nc3 Bb4 exd5 exd5 Be3 Nc6 Nf3 Nf6 Bd3 O-O O-O Bd6 Nb5 Nb4 Nxd6 Qxd6 Ne5 Nxd3 Qxd3 Be6 Rab1 Ne4 Rfe1 Rad8
17 +52.84 10.5M 0:09.88 e3 d5 Nf3 e6 c4 Nf6 Nc3 Be7 d4 O-O Be2 c5 O-O dxc4 Bxc4 Nc6 Bd3 cxd4 exd4 Qb6 Rb1 Nxd4 Bxh7+ Nxh7
17 +52.71 9.86M 0:09.31 Nf3 Nf6 c4 c5 e3 e6 d4 d5 Nc3 Nc6 cxd5 exd5 Bb5 a6 Bxc6+ bxc6 Ne5 cxd4 Nxc6
17 +51.93 11.5M 0:10.81 a3 e5 c4 Nf6 Nc3 Nc6 Nf3 Be7 e4 O-O Be2 d6 d4 Bg4 d5 Nd4 Nxd4 exd4 Qxd4
17 +51.23 10.8M 0:10.14 Nc3 d5 d4 Nf6 Nf3 e6 Bg5 Be7 e3 O-O Bd3 c5 O-O h6 Bf4 c4 Be2 Nc6 Rb1 Bd7 Ne5 Rc8 h3 Nxe5
17 +50.91 12.2M 0:11.42 h3 e5 e4 Nf6 Nf3 Nc6 Nc3 Bc5 Bc4 d6 O-O O-O d3 Be6 Nd5 Bxd5 exd5 Nd4 Bd2
17 +50.58 13.0M 0:12.08 b3 d5 Nf3 e6 e3 Nf6 Be2 c5 d4 Bd6 dxc5 Bxc5 Bb2 Bd7 Nbd2 O-O Bd3
17 +48.97 12.6M 0:11.72 c3 e5 d4 e4 Bf4 d5 e3 Nf6 Ne2 Nc6 Nd2 h6 h4 Be7 Ng3 O-O Be2 Be6 O-O Bd6 Bxd6 cxd6 c4 dxc4
17 +48.89 12.4M 0:11.55 g3 d5 Nf3 c5 d4 Nc6 c3 Nf6 Bg2 e6 O-O Be7 Na3 O-O Bf4 Qb6 Rb1 cxd4 cxd4 h6 e3
17 +48.77 12.6M 0:11.79 d3 e5 Nf3 Nc6 e4 Nf6 Be2 d5 exd5 Nxd5 O-O Bc5 c3 O-O d4 exd4 cxd4 Bb6 Nc3 Re8 Bd3 Nf6
17 +47.25 14.2M 0:13.24 b4 Nf6 Nf3 e6 a3 a5 b5 Be7 e3 d5 d4 Ne4 Bd3 Nd7 c4 O-O cxd5 exd5 O-O Bd6
17 +46.88 13.5M 0:12.63 f4 d5 e3 Bf5 Nf3 e6 Be2 Be7 O-O Nf6 Nh4 O-O d3 c5 Nd2 Nc6 Nxf5 exf5 Nb3 d4 Bd2 dxe3 Bxe3
17 +46.14 15.3M 0:14.31 a4 e5 e4 Nf6 Nf3 Nxe4 d3 Nf6 Nxe5 d5 d4 Bd6 Bb5+ c6 Be2 O-O Nc3 c5 Nb5 Nc6 Nxc6 bxc6 Nxd6 Qxd6 O-O Re8
17 +46.10 14.8M 0:13.82 h4 e5 e3 d5 d4 exd4 exd4 Nf6 Nf3 Be7 Be2 O-O O-O Nc6 c4 Bg4 Nc3 dxc4 Bxc4 Bxf3 Qxf3
16 +53.16 6.74M 0:06.31 e4 e6 d4 d5 e5 c5 c3 Nc6 Nf3 Nge7 Bd3 cxd4 cxd4 Nf5 Bxf5 exf5 Nc3 Be7 Bd2 Be6
16 +52.43 7.12M 0:06.64 Nf3 Nf6 c4 e6 d4 d5 Nc3 Bb4 e3 O-O Bd3 dxc4 Bxc4 c5 O-O cxd4 exd4 Nc6 Bf4 Qb6 Rc1 Rd8
16 +52.30 7.24M 0:06.75 e3 d5 c4 e6 Nf3 Nf6 d4 Be7 Nc3 O-O Bd2 c5 dxc5 Nc6 Be2 dxc4 O-O Bxc5 Bxc4 Bd7 Rc1
16 +52.30 6.67M 0:06.24 d4 e6 Nf3 d5 Bg5 Nf6 e3 Be7 Bd3 c5 O-O c4 Be2 O-O Bxf6 Bxf6 b3
16 +51.93 7.83M 0:07.33 c4 e5 Nc3 Nc6 a3 Nf6 Nf3 Be7 e4 O-O Be2 d6 d4 Bg4 d5 Nd4 Nxd4 exd4 Qxd4 Bxe2 Nxe2
16 +50.74 7.56M 0:07.06 Nc3 d5 d4 Nf6 Bg5 c5 e3 e6 Nf3 Be7 Be2 O-O Bf4 h6 O-O Bd7 Ne5 Nc6 Nb5 Qb6 c4 Nxe5 Bxe5 dxc4
16 +50.70 8.55M 0:08.06 g3 e5 d3 d5 Nf3 Nc6 Bg2 Nf6 d4 e4 Ng5 b5 O-O b4 c3 Be7
16 +50.66 7.85M 0:07.35 c3 e5 d4 e4 Bf4 d5 e3 Nf6 Nd2 Nc6 Ne2 Be7 Ng3 h5 h4 Ng4 Be2 Bxh4 Bxc7 Qxc7 Rxh4
16 +48.89 8.58M 0:08.09 d3 d5 Nf3 Nf6 Bf4 Nc6 Nbd2 Bf5 g3 e6 Bg2 Bd6 Bxd6 cxd6 O-O h6 b4 O-O b5 Ne5 Nd4 Bg6 c4 dxc4 dxc4
16 +48.85 8.18M 0:07.71 h3 e5 e4 Nf6 Nf3 d5 Nxe5 Bd6 d4 Nxe4 Bd3 Qe7 Bxe4 Bxe5 dxe5 dxe4 Qd4 Nc6 Qxe4 Qxe5
16 +48.85 8.08M 0:07.60 a3 d5 d4 Nf6 e3 Bf5 Bd3 e6 Bxf5 exf5 Ne2 Bd6 O-O Nbd7 Ng3 Bxg3 fxg3
16 +48.68 8.28M 0:07.81 h4 e5 e4 Nc6 Nf3 Nf6 Nc3 Be7 d4 exd4 Nxd4 O-O a3 d6 Nxc6 bxc6 Be2 c5
16 +47.37 8.86M 0:08.36 b3 e5 Bb2 Nc6 e4 Nf6 Nf3 Nxe4 Qe2 d5 d3 Nd6 Nxe5 Nxe5 Bxe5 Qe7 Nd2 f6
16 +46.67 8.73M 0:08.22 a4 d5 d4 Bf5 e3 e6 Bd3 Bd6 Bxf5 exf5 Ne2 Nf6 Ng3 Bxg3 hxg3 Nc6
16 +46.34 8.94M 0:08.44 f4 d5 e3 Nf6 c4 d4 Nf3 dxe3 Nc3 exd2+ Qxd2 Qxd2+ Bxd2 e6 O-O-O Bc5 Be2 Nbd7 Nb5 Bb6
16 +44.71 9.54M 0:09.00 b4 e5 Bb2 Nf6 a3 d6 e3 Be7 c4 O-O Ne2 d5 Bxe5 Nc6 Bb2 dxc4 Nbc3 a6
15 +56.79 3.56M 0:03.41 e4 e6 d4 d5 exd5 exd5 Nf3 Nf6 Bd3 Nc6 O-O Bd6 c4 O-O c5 Be7 Nc3 Be6
15 +52.22 3.90M 0:03.71 d4 Nf6 Bf4 e6 e3 c5 Nf3 cxd4 exd4 Nc6 Nc3 Qb6 Nb5 Nd5 Bd2
15 +52.06 4.15M 0:03.94 e3 d5 c4 e6 Nf3 Be7 Nc3 Nf6 d4 O-O Bd3 c5 O-O Nc6 Ne5 cxd4 exd4
15 +51.48 4.07M 0:03.87 Nf3 d5 d4 e6 Bg5 Nf6 e3 c5 c3 Nc6 Nbd2 cxd4 exd4 h6 Bxf6 Qxf6 Bd3
15 +51.28 4.27M 0:04.08 Nc3 d5 d4 Nf6 Bg5 h6 Bxf6 exf6 e3 Be6 Bd3 Bd6 Nge2 O-O O-O c6 Nf4 Nd7 Nxe6
15 +50.33 4.32M 0:04.12 c3 e5 d4 exd4 cxd4 d5 Nf3 Nf6 e3 Be7 Bd2 O-O Bd3 c5 dxc5 Bxc5
15 +49.42 4.91M 0:04.62 a3 e5 c4 Nf6 e3 Be7 d4 e4 Nc3 O-O d5 d6 Nge2 c6 Ng3 cxd5
15 +49.09 5.01M 0:04.72 h3 d5 Nf3 e6 e3 c5 d4 Nf6 c4 cxd4 exd4 Bb4+ Bd2 Bxd2+ Nbxd2 dxc4 Bxc4 O-O O-O Nc6
15 +48.68 4.61M 0:04.36 c4 e5 e3 Nf6 a3 d5 cxd5 Nxd5 d4 exd4 Qxd4 Nc6 Bb5 a6 Bxc6+ bxc6 Nc3 Nxc3
15 +48.19 5.21M 0:04.91 h4 e5 e4 Nf6 Nf3 d5 Nxe5 Nxe4 d4 c6 Nd2 Nf6 c3 Bd6 Bd3 O-O O-O c5
15 +48.15 4.75M 0:04.48 d3 e5 Nf3 Nc6 e4 Nf6 Bg5 d5 Nbd2 h6 Bxf6 gxf6 Be2 Rg8 g3 dxe4
15 +46.67 5.71M 0:05.40 g3 e5 c4 Nf6 Nc3 Nc6 Bg2 Bb4 Nf3 Bxc3 dxc3 O-O Be3 h6 O-O d6 Nd2 Be6
15 +45.85 5.80M 0:05.49 a4 Nf6 d4 e6 e3 d5 Nf3 Bd6 Bd3 O-O O-O c5 c4 Nc6 Nc3 Nb4 dxc5 Bxc5 cxd5 Nxd3 Qxd3
15 +45.61 5.93M 0:05.60 b3 e5 e3 d5 Bb2 e4 c4 c6 cxd5 cxd5 Qc2 Bd7 Nc3 Nf6 Nge2 Nc6
15 +45.57 5.45M 0:05.16 f4 d5 e3 Bf5 Nf3 Nf6 Be2 e6 O-O Be7 c4 dxc4 Bxc4 O-O Nc3 c6 d4 Nbd7 Qb3
15 +44.83 6.09M 0:05.75 b4 d5 e3 e5 Bb2 Nd7 a3 Ngf6 c4 c5 bxc5 Bxc5 Nf3 O-O cxd5 Nxd5 Nxe5 Nxe5 Bxe5
14 +56.06 3.14M 0:03.02 e4 e5 Nf3 Nc6 Bb5 Nf6 O-O Be7 d3 d6 Be3 O-O Nbd2 Bg4 h3 Bxf3 Qxf3
14 +53.25 3.14M 0:03.02 Nf3 d5 d4 e6 Bg5 Be7 Bxe7 Nxe7 e3 c5 Bd3 c4 Be2 Nd7 O-O O-O Nbd2 h6 c3 Qb6
14 +52.84 3.14M 0:03.02 d4 d5 Bf4 Nf6 e3 e6 Nd2 Bd6 Bxd6 cxd6 Ngf3 Bd7 Be2 O-O O-O Nc6
14 +52.67 3.14M 0:03.02 Nc3 d5 d4 Nf6 Bg5 e6 e3 c5 Bxf6 Qxf6 Nf3 Nc6 Bb5
14 +51.73 3.14M 0:03.02 e3 d5 c4 e6 Nf3 c5 d4 Nf6 Nc3 Nc6 cxd5 exd5 Bb5 Bg4 dxc5 Bxc5
14 +49.22 3.14M 0:03.02 h3 d5 Nf3 e6 e3 c5 d4 Nf6 c4 cxd4 exd4 Bb4+ Nc3 O-O Bg5 Nc6
14 +49.14 3.14M 0:03.02 c3 e5 d4 exd4 cxd4 d5 Nc3 Nf6 Bg5 Bb4 a3 Bxc3+ bxc3 O-O e3 Nbd7 Nf3 h6
14 +48.68 3.14M 0:03.02 a3 e5 e4 Nc6 d3 Nf6 Bg5 Bc5 Nf3 O-O Bxf6 Qxf6 Nc3 Bb6 Nd5 Qg6 Nxb6
14 +48.19 3.14M 0:03.02 h4 e5 c4 Nc6 Nc3 Nf6 e3 Bc5 Nf3 O-O d4 exd4 exd4 d5 dxc5 d4
14 +48.15 3.14M 0:03.02 c4 e5 Nc3 Nf6 e3 Bb4 Nge2 O-O a3 Be7 d4 exd4 Nxd4 d6 Qc2 Nc6 Nxc6 bxc6
14 +47.12 3.14M 0:03.02 d3 e5 Nf3 Nc6 Nbd2 d5 e4 Nf6 Be2 Bc5 O-O O-O c3 h6 b4 Bb6 exd5 Nxd5 Ne4 f5
14 +46.51 3.14M 0:03.02 f4 d5 e3 e6 Nf3 c5 Be2 Nf6 O-O Bd6 c4 O-O Nc3 Nc6
14 +45.40 3.32M 0:03.19 b3 e5 e4 Nf6 Bb2 Bc5 d4 exd4 Bxd4 Bxd4 Qxd4 Nc6 Qe3
14 +45.12 3.23M 0:03.11 g3 e5 d4 exd4 c3 dxc3 Nxc3 Nc6 Bg2 Nf6 Nf3 d5 O-O Bb4 Bg5 Bxc3 bxc3
14 +44.92 3.22M 0:03.09 a4 e5 e4 Nf6 Nf3 d5 d4 Nxe4 Nxe5 c6 c3 Nd7 Bf4 Be7 Nd2 O-O Nxe4 dxe4
14 +44.51 3.29M 0:03.17 b4 d5 e3 Nf6 Bb2 e6 Nf3 Nbd7 c3 c6 d4 Be7 Bd3 Bd6

[MikeB]

This final testing before next release- one last shot to make sure playing strength is reasonable - it should never be more than 20 Elo weaker that the SF its as based on. Sometimes it has been, as I have added code that gave it a nice problem solving effect, but ultimately the derivative will get worse, the longer I leave that code in.

500, 5 minute base with 3 second increment games

Results:
https://www.dropbox.com/s/n1jbc24v6w7wc ... a.txt?dl=0

Games:
https://www.dropbox.com/s/75bh5x8cu5m4f4j/all.txt?dl=0

Early Results

Code: Select all

24 of 500 games completed...
Time control: 300 seconds + 3.0000 seconds
Threads = 1
Hash = 256
Date: 10/07/19 : 01:45:49
24 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 061019 64 POPC   1060   0.0   86   86    24   12.5  52.1    6    5   13  25.0  54.2  1048 
   2 Honey-100619               1048  12.7   86   86    24   11.5  47.9    5    6   13  20.8  54.2  1060 
---------------------------------------------------------------------------------------------------------

Results will be updated about every 8 minutes ( every 8 games completed) ,12 games are being run concurrently on a 12 core Mac Pro with ponder off.

This is the final test , I'm fairly confident there will be no surprises, it could finish worse or better than SF - but either way , it will just be noise as 500 games will not prove anything unless something is seriously broken.

A Friday release is looking good, for those who can compile their own, my Github account online should be updated within the next 48 hours or so with all the changes.

[Ovyron]

You can use mulitipv - which shows all the moves and say of you want is multipv setting of 16, but show only 4, set tactical to 4 and and multipv to 4.

But then I can't use exclude-moves. The point of Exclude moves+Tactical setting is to be able to see the 5th, 6th, 7th and 8th best moves on the position, if one already knows the first 4, by excluding them from analysis, so one gets so see them without slowing down the search (because searchmoves doesn't search the top 4 after excluding them.) With my patch it works as it should. Without it it's just a MultiPV that hides the info from non-best moves.

That is what it was designed for, less clutter when using multipv at a high value and you just want see the few best candidates.

It's not less clutter, because if one uses a Tactical setting that is equivalent to MultiPV=4, the 2nd, 3rd, and 4th best moves are getting searched, a PV is built for them, a score is calculated for them, and then the setting hides this info from the user.

Like, in the opening position, the Tactical setting investigates full scores for 1.e4, 1.d4, 1.Nf3 and 1.c4, and their PVs, and their scores, and then HIDES the info searched about the three moves not considered best. How is that clutter?

Since you refuse to fix it for Honey, all I request is that you rename the setting for next version, because "Tactical" makes no sense, if the purpose is to hide the "clutter" of MultiPV (even though the point of MultiPV is getting to see the PVs and scores of the alternative lines, so slowing down the engine to get them and then hiding them is like throwing the baby with the bathwater) call it "Less Clutter".

[MikeB]

You can use mulitipv - which shows all the moves and say of you want is multipv setting of 16, but show only 4, set tactical to 4 and and multipv to 4.

But then I can't use exclude-moves. The point of Exclude moves+Tactical setting is to be able to see the 5th, 6th, 7th and 8th best moves on the position, if one already knows the first 4, by excluding them from analysis, so one gets so see them without slowing down the search (because searchmoves doesn't search the top 4 after excluding them.) With my patch it works as it should. Without it it's just a MultiPV that hides the info from non-best moves.

That is what it was designed for, less clutter when using multipv at a high value and you just want see the few best candidates.

It's not less clutter, because if one uses a Tactical setting that is equivalent to MultiPV=4, the 2nd, 3rd, and 4th best moves are getting searched, a PV is built for them, a score is calculated for them, and then the setting hides this info from the user.

Like, in the opening position, the Tactical setting investigates full scores for 1.e4, 1.d4, 1.Nf3 and 1.c4, and their PVs, and their scores, and then HIDES the info searched so about the three moves not considered best. How is that clutter?

Since you refuse to fix it for Honey, all I request is that you rename the setting for next version, because "Tactical" makes no sense, if the purpose is to hide the "clutter" of MultiPV (even though the point of MultiPV is getting to see the PVs and scores of the alternative lines, so slowing down the engine to get them and then hiding them is like throwing the baby with the bathwater) call it "Less Clutter".

I never saw your patch - do you have the actual code changes ? what GUI are you using that enables exclude moves ? is there an exclude move function built in the UCI specs? I personally never used exclude moves ,but I can see why one would. I need more info on how it is implement in the GUI and/or UCI specs.

[Master Om]

Honey X5i

Crystal-Honey X5i

Binaries of all flavors are towards the bottom of each release page, will add more notes in few days headed to my nephew's a wedding at the shore for tomorrow!

One important note - ALL weaker levels ( play b y Elo or adaptive play) should be used with ONLY one core or thread and 16M hash is more than enough.
These weaker levels will make their moves instantly UNLESS you check the UCI_Sleep button, Then they will play exactly the same moves ( except for the random element that may factor in) , but will take the appropriate time for the time control selected since they will fall into a non busy sleep for that time control.

Enjoy - the adaptive levels will tend to play at your level - so if you're having a bad day, the engine will have a bad day as well . If you start to pull ahead, it will pickup it's game and get tougher.

Both engines have the exact same features - ability to use 4 opening books in sequence ( including Brainfish books) , play by Elo , adaptive play , sleep option to stop rapid fire moves etc.

Enjoy! - will check the board remotely for any questions periodically,

Can please tell whats the purpose of Honey ? I mean its specialty over default SF ?

Does SF have adaptive play?
Does SF have FIDE Elo levels?
Does SF have the ability to use 4 openings books?
Does SF have the ability to turn brute force search on?
Does SF have a variety option?
Does SF has the ability to turn off null moves?
Does SF have predefine level settings ( see snapshot below)?
Does SF have the ability to turn off 7 man probing with more than 7 pieces on the board ( Honey performs one probe with 7 man pieces and that's it. If have have most of 7 man syzygy on HDD, you will know exactly what I'm talking about)
If those options above do not interest you, then perhaps Honey is not for you and I'm fine with that and you should be too.

I make Honey for myself - and I'm just willing to share with with others that share my common desire for the features I have added. It's not everybody's cup of tea, and I'm fine with that.

snapshot of UCI options:
https://www.dropbox.com/s/q9b0pexn34351 ... M.png?dl=0
https://www.dropbox.com/s/uzsk3vu1yo0qm ... M.png?dl=0

Edit : Also Honey does more extensions and less pruning - this often causes it to be slightly weaker at micro bullet , but some people may prefer those settings when using Honey for deep analysis , for example, in an ICCF game.

Q:Does SF have the ability to turn brute force search on?
A: This is interesting to use in corr chess analysis but no idea where to use it. Kindly please explain if possible.
QDoes SF have FIDE Elo levels?
A: No and I dont need it

Q:Does SF have the ability to use 4 openings books?
A: I don't need it either as i dont play in engine room like in pcc

Q:Does SF have the ability to turn brute force search on?
A: No. And I have no idea what is it and how to use it. A help will be good.

Q:Does SF have a variety option?
A: No, But how to use it and how ?

Q:Does SF has the ability to turn off null moves?
A: No. But I know where to use this

Q:Also Honey does more extensions and less pruning - this often causes it to be slightly weaker at micro bullet , but some people may prefer those settings when using Honey for deep analysis , for example, in an ICCF game.

A: This the answer i wanted to know. As I used your Honey X Pro previously for that tactical feature only. Hence wanted to know what extra these new ones have for ICCF players like me who do only deep analysis.
Thanks bdw for the engine. Time to test it

[MikeB]

This final testing before next release- one last shot to make sure playing strength is reasonable - it should never be more than 20 Elo weaker that the SF its as based on. Sometimes it has been, as I have added code that gave it a nice problem solving effect, but ultimately the derivative will get worse, the longer I leave that code in.

500, 5 minute base with 3 second increment games

Results:
https://www.dropbox.com/s/n1jbc24v6w7wc ... a.txt?dl=0

Games:
https://www.dropbox.com/s/75bh5x8cu5m4f4j/all.txt?dl=0

Early Results

Code: Select all

24 of 500 games completed...
Time control: 300 seconds + 3.0000 seconds
Threads = 1
Hash = 256
Date: 10/07/19 : 01:45:49
24 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 061019 64 POPC   1060   0.0   86   86    24   12.5  52.1    6    5   13  25.0  54.2  1048 
   2 Honey-100619               1048  12.7   86   86    24   11.5  47.9    5    6   13  20.8  54.2  1060 
---------------------------------------------------------------------------------------------------------

Results will be updated about every 8 minutes ( every 8 games completed) ,12 games are being run concurrently on a 12 core Mac Pro with ponder off.

This is the final test , I'm fairly confident there will be no surprises, it could finish worse or better than SF - but either way , it will just be noise as 500 games will not prove anything unless something is seriously broken.

A Friday release is looking good, for those who can compile their own, my Github account online should be updated within the next 48 hours or so with all the changes.

I had to kill this one early, but there were no surprises seen:

Code: Select all

350 of 500 games completed...
Time control: 300 seconds + 3.0000 seconds
Threads = 1
Hash = 256
Date: 10/07/19 : 10:01:31
350 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 061019 64 POPC   1055   0.0   19   19   350  176.0  50.3   50   48  252  14.3  72.0  1053 
   2 Honey-100619               1053   1.5   19   19   350  174.0  49.7   48   50  252  13.7  72.0  1055 
---------------------------------------------------------------------------------------------------------

[Leo]

Honey is the latest version of the McBrain-McCain branches of Stockfish. Those branches were specifically coded to be useful in analysis of chess positions, with improvements that would allow it to find the best moves before Stockfish dev, other branches, or by using MultiPV internally (called the Tactical Settings).

Around the date McCain X was released I could say it was the definitive Stockfish version to use for analysis of correspondence games (it won me many games that I couldn't have won without it!). But time has passed, there have been important improvements on Stockfish that have made the differences with McCain/Honey less relevant, and that have meshed well with different code changes from other branches, like Eman or CFish (which is still better than Stockfish dev for this.) Otherwise, Honey suggests very similar moves to other branches so it seems its useful times have been over since July or thereabouts (in my experience).

As for the Crystal version, it adds code that makes it a tactical monster in test suites, being the best at finding the most tactics in the most positions the fastest, or something. In practice, I couldn't find an use for it. The problem is that if there's no tactic to find it's really slow, as in, if the best move isn't tactical it's going to take ages to find it. You'd want to use Crystal if there's a tactic on the board and you want to find it ASAP, but the conundrum is that you can't know that in advance, so if you already know about the tactic you don't need Crystal in the first place...

Disclaimer: I haven't tried Sugar NN yet, that's why it's absent from this post.

OK. Thanks. Very helpful.