A 187 Elo point gain over previous version ...

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

A 187 Elo point gain over previous version ...

Post by MikeB »

ok, I got cute with the headline, but the story here is that the next version of Honey will have 187 Elo point gain with games to a one depth search... this is primarily due to the new SF net and a few tweaks i have made to advantage of this newly found one depth strength - this will be a very helpful for human players playing on the android devices, as about 60% of the chess players will now have competitive games with just a one depth search, thus saving their device battery for more games or whatever else they use their device for..

Code: Select all

Search_Depth set to 1
Time Control: Time Control-> base+inc: 10+0.100
Games: 10000
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/24/21 : 21:10:43

Projected-> Time: 2h:33m:36s
     Run -> Time: 0h:4m:14s 

10000 game(s) loaded
Rank Name            Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Honey-v13.2-SE   3594   0.0    7    7 10000 7450.0  74.5 7147 2247  606  71.5   6.1  3406
   2 Honey-v13.1b     3406 187.6    7    7 10000 2550.0  25.5 2247 7147  606  22.5   6.1  3594
---------------------------------------------------------------------------------------------------------

LOS:
                Ho Ho
Honey-v13.2-SE    100
Honey-v13.1b     0

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
more to come ...
Image
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: A 187 Elo point gain over previous version ...

Post by MikeB »

this is the one ply Honey Engine vs current SF with the Elo set 1550 - which is a solid club level chess player setting, not a master, not an expert - but a very solid club level player level .. will now use the new variety to see the impact of my variety setting - hopefully it's very minor (variety turned off once the 4th piece comes off)

Code: Select all

#############################################################################################################################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05242137.pgn
Time Control: Time Control-> base+inc: 10+0.100
Games: 600
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/24/21 : 21:41:40

Projected-> Time: 0h:10m:3s
     Run -> Time: 0h:4m:3s

600 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Honey-v13.2-SE     3505   0.0   27   27   600  309.0  51.5  301  283   16  50.2   2.7  3495
   2 Stockfish-230521   3495  10.7   27   27   600  291.0  48.5  283  301   16  47.2   2.7  3505
---------------------------------------------------------------------------------------------------------

  Δ = delta from the next higher rated opponent
  # = number of games played
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
                  Ho St
Honey-v13.2-SE       77
Stockfish-230521  22

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
Image
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: A 187 Elo point gain over previous version ...

Post by MikeB »

so the new variety algorithm is costing abut 75 Elo , as compared to the previous version, variety was costing to 200 to 300 Elo...
Current-Stockfish set to 1550 Elo
Honey set to depth 1 search with variety, variety auto turned off once the total piece counts drops to 29 ( board must have 30 or more pieces for variety to be active)

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05242148.pgn
Time Control: Time Control-> base+inc: 10+0.100
Games: 600
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/24/21 : 21:52:32

Projected-> Time: 0h:10m:3s
     Run -> Time: 0h:4m:9s

600 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Stockfish-230521   3533   0.0   28   28   600  356.0  59.3  348  236   16  58.0   2.7  3467
   2 Honey-v13.2-SE     3467  65.8   28   28   600  244.0  40.7  236  348   16  39.3   2.7  3533
---------------------------------------------------------------------------------------------------------



LOS:
                  St Ho
Stockfish-230521     99
Honey-v13.2-SE     0

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
Image
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: A 187 Elo point gain over previous version ...

Post by MikeB »

this run is new Honey set to search depth 9 with variety "on" against Stockfish with an Elo setting of 2200 ( probably about FIDE 2300 to 2400)

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05242202.pgn
Time Control: Time Control-> base+inc: 10+0.100
Games: 600
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/24/21 : 22:07:49

Projected-> Time: 0h:10m:3s
     Run -> Time: 0h:5m:11s

600 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Honey-v13.2-SE     3532   0.0   26   26   600  354.5  59.1  306  197   97  51.0  16.2  3468
   2 Stockfish-230521   3468  63.6   26   26   600  245.5  40.9  197  306   97  32.8  16.2  3532
---------------------------------------------------------------------------------------------------------

  Δ = delta from the next higher rated opponent
  # = number of games played
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
                  Ho St
Honey-v13.2-SE       99
Stockfish-230521   0

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
Image
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: A 187 Elo point gain over previous version ...

Post by MikeB »

here's a little teaser, and note- there is no fortress deduction code being used

[d]3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 1
54 +7.84! 8.45G 2:21.18 Rxg4!
54 +7.62! 7.72G 2:08.46 Rxg4!
53 +7.41 6.32G 1:44.46 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Qg6 Kc7 h6 Rfh7 e6 Kb7 Kb2 Ka6 Rh5 g3 Qxg3 Bxe6 Qg6 Bd5 Rh2 Kb6 Rh3 Ka6 Qf6 Kb6 Qd6 Ka6 Rh4 Kb7 Rh5 Ka6 Qxc5 Rxh6 Rg5 Re6 Rxf5 Rg6 Re5 Rg7 Qd6 Kb7 Rf5 Ra8 Rf6 Rh7 Qc5 Rd7 Qxd4 Rdd8 Qc5 Rg8 Qe7+ K
52 +7.60 5.60G 1:32.45 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Kb2 Kc7 h6 Rfh7 Qg6 Kb6 e6 g3 Qxg3 Bxe6 Qg6 Bd5 Rh3 Kb7 Qf6 Ka6 Qd6 Kb6 Rh5 Ka6 Rh4 Kb7 Rh2 Bf3 Qxc5 Rxh6 Rf2 Bd5 Qxa5 R8h7 Qxb4+ Ka8 Qc5 Re6 Qf8+ Kb7 Qxf5 Reh6 Re2 Rh2 Rxh2 Rxh2 Qd7+ Ka8 a5 Rh8
52 +7.61! 5.12G 1:24.56 Rxg4!
51 +7.39 4.77G 1:18.57 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Rfh7 e6 Kb7 Kb2 Kb6 Rh5 g3 Qxg3 Bxe6 Qg5 Bf7 Rh3 Bd5 Qg6 Ka6 Qd6 Kb6 Rh5 Ka6 Qxc5 Rxh6 Rg5 Ra8 Rxf5 Rg8 Re5 Rh7 Qd6 Rgh8 Rg5 Kb7 Qc5 Ra8 Qxd4 Ka6 Rg6 Rd7 Qc5 Kb7 Rg5 Bf7 Rf5 Be6 Re5 Bg
51 +7.50? 4.44G 1:13.00 Rxg4 hxg4?
50 +7.72 4.36G 1:11.83 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Rfh7 Kb1 Kb6 Kb2 Kc7 Rh5 Kb6 Rh4 Ka6 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Kb7 Rh5 Kb6 Qd6 Ka6 Qxc5 Rxh6 Rg5 Rb8 Rxf5 Rg8 Re5 Rh7 Qd6 Rgh8 Rg5 Kb7 Qc5 Ra8 Re5 Ka6 Rf5 Kb7 Rf6 Rhh8 f5 Rhd8 Qe7+ K
49 +7.50 3.61G 0:59.39 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Rh4 Kb6 Kb2 Rfh7 Qg6 Bf3 e6 Bd5 Rh2 g3 Qxg3 Bxe6 Qg6 Bd5 Rh3 Ka6 Qd6 Kb6 Rh5 Ka6 Qxc5 Rxh6 Rxf5 Rg6 Re5 Rg7 Qd6 Kb7 Rf5 Rgh7 Rg5 Ra8 Qc5 Rd7 Re5 Rh7 Rf5 Rd7 Rg5 Rh7 Re5 Ka6 Rf5 Kb7 Rf6 Rc7
49 +7.54! 3.44G 0:56.60 Rxg4!
48 +7.32 3.23G 0:53.10 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Qg6 Kc7 h6 Rfh7 Kb2 Kb7 e6 Kb6 Rh5 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Ka6 Qg6 Kb7 Qd6 Kb6 Kb1 Kb7 Kc1 Kb6 Kb2 Ka6 Qxc5 Rxh6 Rxf5 Rg6 Re5 Rg7 Qd6 Kb7 Rf5 Rgh7 Rg5 Ra8 Qc5 Rc7 Rf5 Rh7 Rf6 Rd7 f5 Kc7 Rg6 Rf7
47 +7.46 3.11G 0:51.01 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Qg6 Kc7 h6 Rfh7 Kb1 Kb6 Kb2 Ka7 e6 Kb6 Rh5 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Ka6 Rh2 Bf3 Rh3 Bd5 Rh5 Kb6 Qd6 Ka6 Qxc5 Rxh6 Rg5 Rb8 Rxf5 Rg8 Re5 Rh7 Qd6 Rgh8 Rg5 Rh2 Qc5 Rg2 Rf5 Rg6 Qxd4 Rh1 Re5 Re6 Rg5 Rh
47 +7.59! 2.86G 0:46.76 Rxg4!
46 +7.37 2.59G 0:42.41 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 h6 Kc7 Qg6 Rfh7 Kb1 Kb6 Kb2 Ka7 e6 Kb6 Rh5 g3 Qxg3 Bxe6 Qg5 Bf7 Rh3 Be6 Qf6 Bd5 Rh5 Ka6 Rh2 Kb6 Rh3 Ka6 Qd6 Kb6 Rh5 Ka6 Qxc5 Rxh6 Rxf5 Rg8 Re5 Rh7 Qd6 Rgh8 Rg5 Kb7 Qc5 Ra8 Re5 Ka6 Rf5 Kb7 Rg5 Ka6
46 +7.69! 2.51G 0:40.98 Rxg4!
46 +7.46? 2.36G 0:38.59 Rxg4 hxg4?
45 +7.69 2.28G 0:37.34 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Rfh7 Kb2 Kb6 Rh4 Bf3 e6 Bd5 Rh5 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Ka6 Rh2 Bf3 Rh3 Bd5 Qd6 Kb6 Rh5 Ka6 Qxc5 Rxh6 Rxf5 Rg8 Re5 Rh7 Qd6 Rgh8 Rg5 Rh2 Qc5 Rg2 Rf5 Rd2 Re5 Rhh2 Re8 Rxc2+ Qxc2 Rxc2+ K
45 +7.25? 2.17G 0:35.57 Rxg4 hxg4?
45 +7.48? 2.16G 0:35.38 Rxg4 hxg4?
44 +7.70 2.16G 0:35.36 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Kb2 Kc8 Qg5 Kc7 Re2 Be6 Qg6 Re7 Qf6 Rhh7 Rh2 Kb6 Rh5 Re8 Rg5 Rf7 Qg6 Rfe7 Qh5 Rh8 Qh4 Reh7 Rg6 Bf7 Rf6 Bd5 Qg5 g3 Qxg3 Rxh6 Rxh6 Rxh6 Qg5 Rh7 Qd8+ Ka6 Qc8+ Rb7 Qxf5 Bxb3 cxb3 Kb6 Qc8 Rf7
44 +7.63! 1.69G 0:27.93 Rxg4!
43 +7.41? 1.48G 0:24.47 Rxg4 hxg4?
42 +7.64 1.47G 0:24.34 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Kb6 Kb2 Rfh7 Rh4 Rf7 Rh5 Rfh7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Ka6 Rh2 Kb6 Rh3 Ka6 Qg5 Rf8 Rg3 Kb6 Qg6 Rd7 Rh3 Bg8 h7 Rxh7 Rxh7 Bxh7 Qxh7 c4 bxc4 Re8 Qxf5 Re1 Qc8 c5 f5 Rf1 Qe6+ Ka7 Qd7+ Kb
42 +7.61! 1.29G 0:21.38 Rxg4!
41 +7.39 992.6M 0:16.60 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Rfh7 Kb2 Kb6 Rh4 Ra7 Rh5 Rah7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Rh3 Rf7 Qg6 Rff8 Qd6 Rd8 Qf6 Rdf8 Qe7 Re8 Qd7 Ref8 Rg3 Rd8 Qxf5 Rxh6 Rg7 Rg8 Rxg8 Bxg8 Qf8 Rg6 Qb8+ Ka6 Qe8 Bh7 Qa8+ Kb6 f5
40 +7.41 917.0M 0:15.39 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qh5 Kc7 Qg6 Rfh7 Rh5 Kb6 e6 Re7 Qg5 Rxe6 h7 Ree8 Kb2 Be6 Rh6 Ref8 Qe7 Re8 Qf6 g3 Qg7 Bd5 Qxg3 Re7 Qg6 Be6 Rh3 Kb7 Rh5 Kc7 Qf6 Rhxh7 Rxh7 Rxh7 Qxe6 Rd7 Qe5+ Kb6 Qxf5 Kc7 Qe5+ Rd6 f5 Kd7 Qg7+ Kc8 f
39 +7.20 696.5M 0:11.76 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qh5 Kc7 Qg6 Rfh7 Rh5 Kb6 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Rh3 Bf3 Rxf3 Rxh6 Rg3 Rh1+ Kb2 Re8 Rg1 Rxg1 Qxg1 Kc7 Qg7+ Kb6 Qh7 Re1 Qxf5 Rf1 Qc8 Re1 Qb8+ Ka6 Qc7 Rf1 Qxc6+ Ka7 Qc7+ Ka6 f5 Rf2 Qxc5 Kb7
38 +7.19 625.6M 0:10.62 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Kb6 Kb2 Rff8 h7 Rf7 e6 Re7 Qf6 Ree8 Qg5 Bxe6 Rh6 Bg8 Qg6 Be6 Qf6 Kb7 Qg7+ Kb6 Rf6 Bd5 Rxf5 g3 Re5 Rxe5 fxe5 Rxh7 Qxh7 g2 Qg7 Ka6
38 +7.16! 598.2M 0:10.18 Rxg4!
38 +6.95! 454.2M 0:07.79 Rxg4!
37 +6.74 363.4M 0:06.32 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qg5 Kc7 Qg6 Kb6 Rh5 Rfh7 e6 Re7 Qg5 Rxe6 h7 Ree8 Kb2 Rd8 Qg7 Be6 Qf6 Rde8 Rh6 Rhf8 Qg6 Bd5 Qg7 Rh8 Rg6 Rhf8 Rg5 Rd8 Qh6 Rh8 Rg7 Rdf8 Qg6 Rf7 Rxf7 Bxf7 Qg7
36 +6.71 322.9M 0:05.65 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 h6 Kc7 Qg6 Kb6 Rh5 Rfh7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Rh3 Bf3 Rxf3 Rxh6 Rg3 Rh1+ Kb2 Re8 Rg1 Rxg1 Qxg1 Re7 Qg5 Re1 Qxf5 Kc7 Qh7+ Kc8 Qg8+ Kd7 Qg4+ Kd8 Qh4+ Re7 Qh5 Kc7 Qxc5 Re8 f5 Re1
35 +6.54 283.4M 0:05.01 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 h6 Kc7 Rh5 Rfh7 Qg6 Kb6 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Rh3 Bf3 Rxf3 Rxh6 Rg3 Rh1+ Kb2 Re8 Rg1 Rxg1 Qxg1 Re7 Qg5 Re1 Qxf5 Kc7 Qh7+ Kd8 Qh4+ Re7 f5 c4 f6 Rf7 Qxd4+ Ke8 dxc4 c5 Qe3+ Kf8 Qxc5+ Kg8
34 +6.51 260.2M 0:04.61 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 h6 Kc7 Rh5 Kb6 Qg6 Rfh7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Rh3 Bf3 Rxf3 Rxh6 Rg3 Rh1+ Kb2 Re8 Rg1 Rxg1 Qxg1 Ra8 Qg7 Re8 Qf7 Re1 Qf8 Rh1 Qb8+ Ka6 Qc8+ Kb6 Qxf5 Rf1 Qf8 Kb7 Qe7+ Kc8 Qa7 Rxf4 Qa8+ Kc7
33 +6.51 253.2M 0:04.50 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qh5 Kc8 Qg6 Kc7 Rh5 Kb6 Kb2 Rfh7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Bf3 Rg5 Rxh6 Qxf5 Bd5 Rg7 Rg8 Rxg8 Bxg8 Qf8 Rh2 Qxg8 c4 bxc4 Kb7 Qf7+ Kb6 c5+ Ka6
32 +6.42 227.1M 0:04.06 Rxg4 hxg4 h5 Rh7 Qh4 Kd7 Qf6 Rf7 Qg5 Kc7 h6 Rh8 Rh5 Kb6 Qg6 Rfh7 e6 Re7 Qg5 Reh7 e7 g3 Qxg3 Rxe7 Qg5 Reh7 Rh3 Bf3 Rxf3 Rxh6 Kb2 Rh1 Rg3 Re1 Rg1
32 +6.39! 221.3M 0:03.96 Rxg4!
32 +6.19! 195.6M 0:03.51 Rxg4!
31 +6.00 169.5M 0:03.04 Rxg4 hxg4 h5 Rh7 Qh4 Kd7 Qf6 Be6 h6 Rg8 Rh5 g3 Rg5 Rf7 h7 Rxh7 Rxg8 Bxg8 Qd6+ Kc8 Qxc6+ Kb8 Qg6 Kc7 Qxg8 Rh1+ Kb2 c4 bxc4 g2 Qxg2 Re1
30 +4.95 103.1M 0:01.83 Re2 Re8 Qf2 Rg8 Re1 Rdg7 Rgg1 Rg6 Kb2 Rf8 Kb1 Rh8 Qd2 Rhg8 Rg3 Rb8 Qf2 Rbg8 Re4 R6g7 Rxd4 cxd4 Qxd4 Rd7 Qb6 Ra8 Qc5 Rdd8 Kc1 Rd7 Kd2 Rdd8 Rg1
29 +5.17 87.5M 0:01.56 Re2 Re8 Qf2 Rg7 Kb1 Reg8 Kb2 Rg6 Re1 Rf8 Rgg1 Rfg8 Kb1 g3 Qe2 Rg4 Rg2 R8g6 Kb2 Bxg2 Qxg2 Ke7 Qf3 Kf7 Rg1 c4 dxc4 g2 Qf2 c5 Qf3
28 +5.32 63.3M 0:01.09 Re2 Rg7 Kb2 Rdg8 Rg1 Re8 Qg3 Rg6 Rge1 Rc8 Rf2 Rf8 Rff1 Re8 Kb1 Rd8 Kc1 Re8 Re2 Rf8 Rg1 Re8 Kb2 Reg8 Ree1 Re8 Kb1 Rf8
27 +5.44 43.7M 0:00.73 Kb2 Rc7 Rg1 Re7 Qg3 Rg7 Rf2 Re8 Rff1 Rg6 Re1 Rf8 Kc1 Re8 Re2 Reg8 Kb1 Rf8 Rge1 Re8 Kb2 Bf3 Rf2 Bd5 Rff1 Reg8 Re2 Rf8 Ree1
26 +5.82 37.0M 0:00.61 Kb2 Rg8 Re2 Rdg7 Qf2 Rf8 Rg1 Rg6 Ree1 Rfg8 Qg3 Rd8 Re2 Re8 Kc1 Rgg8 Qf2 Rg6 Kb1 Re7 Ree1 Kd7 e6+ Rgxe6 Rxe6 Bxe6
25 +5.27 33.2M 0:00.55 Kb2 Re7 Re2 Rg8 Qf2 Reg7 Re1 Rg6 Ka1 R6g7 Kb1 Rg6 Kb2 R6g7 Kc1 Rg6 Re4 R6g7 Rxd4 cxd4 Qxd4 Rd7 Qc5 Ra8 Kd2 Kf7 Rg1 Ra6 e6+ Bxe6 Qe5
24 +5.23 30.2M 0:00.50 Kb2 Re8 Re2 Rde7 Qf2 Rg8 Re1 Reg7 Ka1 Rg6 Kb1 R6g7 Kb2 Rg6 Qd2 Re8 Kb1 Rg7 Qh2 Rgg8 Qf2 Rg6 Kc1 Kd7 e6+ Rgxe6
23 +6.22 20.6M 0:00.34 Kb2 Re7 Re2 Rf8 Qd2 Kd7 e6+ Kc7 Rg1 Rf6 Rge1 Rh6 Re5 Rf6 Qf2 Kd6 Kb1 Rexe6 Kb2 Ke7 Kc1 Kd7 Rxe6 Rxe6 Rxe6 Kxe6
22 +5.46 11.0M 0:00.19 Rg1 Re7 Kb1 Kd7 Qd2 Kc8 Rf2 Kb7 Rff1 Kc7 Rg3 Rde8 Kb2 Kb6 Rgg1 Rd8 Re1 Rf8 Rg3 Kc7 Qh2 Kb6 Rf1
21 +5.41 7.68M 0:00.14 Kd1 Re7 Ke2 Kd7 Kf2 Rde8 Rg1 Kc7 Kg3 Kb7 Qd2 Kb6 Rf2 Rf8 Re2 Bf3 Rge1 Rfe8 e6 Bxe2 Qxe2 Kb7 Qe5
20 +5.51 6.88M 0:00.12 Kd1 Re7 Ke2 Rde8 Kf2 Kd7 Rg1 Kc7 Kg3 Kb6 Qd2 Bf3 Rf2 Bd5 Re2 Bf3 Ree1 Kc7 e6 Kd6 Re5 Rxe6
19 +5.47 6.01M 0:00.11 Kd1 Re7 Ke2 Rde8 Rg1 Kd7 Kf2 Kc7 Qd2 Kb6 Kg3 Kc7 Rf2 Kb7 Re2 Bf3 Ree1 Bd5 e6 Kc7 Re5 Rxe6
18 +5.47 5.88M 0:00.10 Kd1 Re7 Ke2 Kd7 Kf2 Kc7 Rg1 Rde8 Kg3 Kb6 Qd2 Kc7 Rf2 Kb6 Re2 Bf3 Ree1 Bd5 e6 Kc7 Re5
17 +5.33 5.69M 0:00.10 Rg1 Re7 Kd2 Kd7 Ke2 Kc8 Kf2 Kb7 Kg3 Rde8 Re2 Bf3 Qf2 Bxe2 Qxe2 Kc8 Re1 Kd7 Qg2 Kc7
16 +5.25 5.08M 0:00.09 Re2 Re7 Qf2 Rf8 Kd2 Kd7 Rg1 Kc7 Ree1 Kb6 Qh2 Bf3 Rgf1 Rd8 Rxf3 gxf3 Qg3
15 +5.43 3.35M 0:00.06 Re2 Re7 Qf1 Kd7 e6+ Kc7 Re5 Rf8 Rxd5 cxd5 Qe1 Kd6 Qe5+ Kc6
14 +5.61 2.13M 0:00.04 Rg1 Re7 Qg3 Kd7 Kb1 Kc7 Rd2 Kb6 Rc1 Bf3 Qe1 Kb7 Kb2 Kc7 Rh2 Kb6 Ra1
13 +5.67 1.69M 0:00.03 Rg1 Re7 Qg3 Kd7 Kb1 Kc7 Rd2 Kb6 Qf2 Kc7 Re1 Kc8 Rc1 Kb7 Qg3 Kb6
12 +6.25 891097 0:00.02 Rg1 Re7 Kb2 Kd7 Qg3 Kc8 Rc1 Kc7 Re1 Bf3 Rf2 Kb6 Rxf3 gxf3
11 +6.62 589365 0:00.01 Rg1 Re7 Qg3 Kd7 Re2 Kc8 Kb2 Kc7 Rc1 Bf3 Ree1
10 +7.03 345468 0:00.01 Kb1 Re7 Qg1 Kd7 Rd2 Kc8 Qf2 Kb7 Rg1
9 +6.91 258537 0:00.01 Rg1 Re7 Kb2 Bf3 Rf2 c4 dxc4 d3 cxd3
8 +6.91 150691 0:00.00 Rg1 Rb7 Kb1 Bf3 Ka1 Rg8 Qg3 c4 dxc4
7 +7.11 110121 0:00.00 Qf2 Rh8 Rg1 Re7 Kb1 Kf7 Ka1
6 +7.20 68514 0:00.00 Kb2 Rc7 Rg1 Bf3 Ka1 c4
5 +7.67 62224 0:00.00 Qg1 Rc7 Kb2 Ra7 Ka1
4 +9.29 17295 0:00.00 Rf2 Bf3 Rfxf3
3 +7.30 10935 0:00.00 Kb2 Bf3 Rd2
2 +7.20 7013 0:00.00 Kb2 c4
1 NNUE evaluation using nn-7756374aaed3.nnue enabled
1 +6.54 4232 0:00.00 Kb2
0 #
Image
User avatar
Nordlandia
Posts: 2821
Joined: Fri Sep 25, 2015 9:38 pm
Location: Sortland, Norway

Re: A 187 Elo point gain over previous version ...

Post by Nordlandia »

Is Rxg4 winning or a fortress in the end ?
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: A 187 Elo point gain over previous version ...

Post by peter »

MikeB wrote: Tue May 25, 2021 6:44 am here's a little teaser, and note- there is no fortress deduction code being used

[d]3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 1
54 +7.84! 8.45G 2:21.18 Rxg4!
54 +7.62! 7.72G 2:08.46 Rxg4!
53 +7.41 6.32G 1:44.46 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Qg6 Kc7 h6 Rfh7 e6 Kb7 Kb2 Ka6 Rh5 g3 Qxg3 Bxe6 Qg6 Bd5 Rh2 Kb6 Rh3 Ka6 Qf6 Kb6 Qd6 Ka6 Rh4 Kb7 Rh5 Ka6 Qxc5 Rxh6 Rg5 Re6 Rxf5 Rg6 Re5 Rg7 Qd6 Kb7 Rf5 Ra8 Rf6 Rh7 Qc5 Rd7 Qxd4 Rdd8 Qc5 Rg8 Qe7+ K
52 +7.60 5.60G 1:32.45 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 Qg5 Kd7 Kb2 Kc7 h6 Rfh7 Qg6 Kb6 e6 g3 Qxg3 Bxe6 Qg6 Bd5 Rh3 Kb7 Qf6 Ka6 Qd6 Kb6

...

33 +6.51 253.2M 0:04.50 Rxg4 hxg4 h5 Rh8 Qh4 Rf7 h6 Kd7 Qh5 Kc8 Qg6 Kc7 Rh5 Kb6 Kb2 Rfh7 e6 g3 Qxg3 Bxe6 Qg5 Bd5 Qf6 Bf3 Rg5 Rxh6 Qxf5 Bd5 Rg7 Rg8 Rxg8 Bxg8 Qf8 Rh2 Qxg8 c4 bxc4 Kb7 Qf7+ Kb6 c5+ Ka6
32 +6.42 227.1M 0:04.06 Rxg4 hxg4 h5 Rh7 Qh4 Kd7 Qf6 Rf7 Qg5 Kc7 h6 Rh8 Rh5 Kb6 Qg6 Rfh7 e6 Re7 Qg5 Reh7 e7 g3 Qxg3 Rxe7 Qg5 Reh7 Rh3 Bf3 Rxf3 Rxh6 Kb2 Rh1 Rg3 Re1 Rg1
32 +6.39! 221.3M 0:03.96 Rxg4!
32 +6.19! 195.6M 0:03.51 Rxg4!
31 +6.00 169.5M 0:03.04 Rxg4 hxg4 h5 Rh7 Qh4 Kd7 Qf6 Be6 h6 Rg8 Rh5 g3 Rg5 Rf7 h7 Rxh7 Rxg8 Bxg8 Qd6+ Kc8 Qxc6+ Kb8 Qg6 Kc7 Qxg8 Rh1+ Kb2 c4 bxc4 g2 Qxg2 Re1
30 +4.95 103.1M 0:01.83 Re2 Re8 Qf2 Rg8 Re1 Rdg7 Rgg1 Rg6 Kb2 Rf8 Kb1 Rh8 Qd2 Rhg8 Rg3 Rb8 Qf2 Rbg8 Re4 R6g7 Rxd4 cxd4 Qxd4 Rd7 Qb6 Ra8 Qc5 Rdd8 Kc1 Rd7 Kd2 Rdd8 Rg1

...
But SF dev. isn't really bad at this one neither, is it?

[d]3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 1

Analysis by Stockfish 190521:

Code: Select all

1.Kb1 Rc7 2.Ka1 Rf7 3.Rg1 Bf3 
  +-  (6.10)   Depth: 7/7   00:00:01  20kN
...
1.Kd1 Rg7 2.Rd2 Bf3+ 3.Kc1 Bd5 4.Qf2 Rg6 5.Re2 Re8 6.Re4 Rg7 7.Re1 Rg6 8.Rgg1 Kd7 9.e6+ Rexe6 10.Re5 Rxe5 11.fxe5 Be6 12.Kd2 Kc7 13.Ke2 Kb7 14.Qf4 Kc7 15.Kf2 Kb6 
  +-  (4.88)   Depth: 23/30   00:00:02  13800kN
1.Re2 Rg7 2.Kb1 Re8 3.Qd2 Rge7 4.Qc1 Rf8 5.Kb2 Rb8 6.Re1 Rg8 7.Ka1 Rb8 8.Rgg1 Bf3 9.Qd2 Rg8 10.Kb1 Bd5 11.Qf2 Rg6 12.Rg3 Rg8 13.Reg1 Kd7 
  +-  (4.63)   Depth: 24/31   00:00:03  28629kN
1.Re2 Re7 2.Kb2 Rb8 3.Qf2 Rd8 4.Re1 Rg8 5.Qd2 Rgg7 6.Qc1 Rg8 7.Rgg1 Bf3 8.Ka2 Bd5 9.Re2 Rb8 10.Rg3 Rg8 11.c4 bxc3 12.Qa3 Rb8 13.Qxc5 
  +-  (4.59)   Depth: 25/31   00:00:03  34374kN
...
1.Re2 Rg7 
  +-  (4.31 --)   Depth: 36/38   00:00:27  365MN
1.Rxg4 
  +-  (4.39 ++)   Depth: 36/41   00:00:40  473MN
...
1.Rxg4 hxg4 2.h5 Rh7 3.Qh4 Kd7 4.Qf6 Rf7 5.Qg6 Rh8 6.Kb2 Kc7 7.e6 Rff8 8.h6 Rhg8 9.Qh7+ Kb6 10.Qd7 g3 11.Re2 Rd8 12.h7 Rh8 13.Qg7 g2 14.e7 Rde8 15.Re1 Kb7 16.Qg6 Bxb3 17.Kxb3 g1R 18.Qxg1 Rxh7 19.Qg5 
  +-  (5.48)   Depth: 36/49   00:00:40  536MN
1.Rxg4 
  +-  (5.56 ++)   Depth: 37/17   00:00:41  573MN
1.Rxg4 hxg4 2.h5 Rh8 3.Qh4 Rf7 4.h6 Kd7 5.Kb2 Rfh7 6.Qg5 Kc7 7.Rh4 Kb6 8.Qg6 Re7 9.Rh5 Reh7 10.e6 g3 11.Qxg3 Bxe6 12.Qg5 Bd5 13.Qf6 Ka6 14.Qg6 Ka7 15.Qd6 Kb6 16.Kc1 Ka6 17.Qxc5 Rxh6 18.Rxf5 Rg8 19.Kb2 Rg1 20.Rxd5 cxd5 21.Qb5+ Ka7 22.Qxa5+ Kb7 23.Qb5+ Rb6 
  +-  (5.49)   Depth: 37/47   00:00:41  578MN
...
1.Rxg4 hxg4 2.h5 Rh8 3.Qh4 Rf7 4.h6 Kd7 5.Kb2 Rfh7 6.Qg5 Kc7 7.Rh4 Rf7 8.Qg6 Kb6 9.Rh5 Rfh7 10.e6 g3 11.Qxg3 Bxe6 12.Qg5 Bd5 13.Rh3 Ka6 14.Qf6 Kb6 15.Qd6 Rg8 16.Rh5 Rgh8 17.Kc1 Ka6 18.Qxc5 Rxh6 19.Rxf5 Rh1+ 20.Kb2 R1h2 21.Rxd5 Rxc2+ 22.Kxc2 Rh2+ 23.Kd1 cxd5 24.Qd6+ Ka7 25.f5 Rg2 
...
1.Rxg4 hxg4 2.h5 Rh8 3.Qh4 Rf7 4.h6 Kd7 5.Kb2 Rfh7 6.Qg5 Kc7 7.Qg6 Kb7 8.e6 g3 9.Qxg3 Bxe6 10.Qg6 Bd5 11.Rh3 Ka7 12.Qf6 Kb6 13.Qd6 Ka6 14.Rh5 Kb6 15.Kc1 Kb7 16.Rg5 Bxb3 17.Qxc5 Bxa4 18.Kb2 Bb5 19.Rxf5 Ka6 20.Rg5 a4 21.Rg7 a3+ 22.Kb3 Rxg7 23.hxg7 
  +-  (5.88)   Depth: 39/60   00:01:03  917MN
...
1.Rxg4 hxg4 2.h5 Rh8 3.Qh4 Rf7 4.h6 Kd7 5.Kb2 Kc7 6.e6 Rfh7 7.e7 Re8 8.Re2 g3 9.Qxg3 Rxh6 10.Qg7 Rd6 11.Qf8 Kd7 12.Qxf5+ Re6 13.Rxe6 Bxe6 14.Qxc5 Rxe7 15.Qxa5 Bd5 16.Qa8 Rf7 17.Qb8 c5 18.a5 c4 19.bxc4 Bxc4 20.Qe5 Rxf4 21.Qxf4 Bd5 22.Qe5 Bc6 23.Qxd4+ Kc8 24.Qxb4 
  +-  (6.17)   Depth: 40/52   00:01:28  1300MN
Latest but one green net (first bigger one), 12x3GHz SSE4.1 popcnt 8G hash.
BTW newest CorChess with latest green net was a little faster even, took 1 time 11 and 2 times 17 seconds to solution at three triasl with same hardware.

@ Jon: the position is a changed ERET nr. 75 fortress, changed to White wins, that's the line I've stored in my database:

3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 6

6. Rxg4 hxg4 7. h5 Rh7 8. Qh4 Kd7 9. e6+ Bxe6 10. Re2 Re8 11. Re5 Rf7 12. Rxc5
Rg8 13. h6 Kd6 14. Rxa5 g3 15. Kb2 1-0
Peter.
Paloma
Posts: 1167
Joined: Thu Dec 25, 2008 9:07 pm
Full name: Herbert L

Re: A 187 Elo point gain over previous version ...

Post by Paloma »

peter wrote: Tue May 25, 2021 10:19 am ...
Latest but one green net (first bigger one), 12x3GHz SSE4.1 popcnt 8G hash.
BTW newest CorChess with latest green net was a little faster even, took 1 time 11 and 2 times 17 seconds to solution at three triasl with same hardware.

@ Jon: the position is a changed ERET nr. 75 fortress, changed to White wins, that's the line I've stored in my database:

3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 6

6. Rxg4 hxg4 7. h5 Rh7 8. Qh4 Kd7 9. e6+ Bxe6 10. Re2 Re8 11. Re5 Rf7 12. Rxc5
Rg8 13. h6 Kd6 14. Rxa5 g3 15. Kb2 1-0
?? What are the first 5 moves from

how did you discover that this is a modified Eret 75?
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: A 187 Elo point gain over previous version ...

Post by peter »

Paloma wrote: Tue May 25, 2021 11:02 am
peter wrote: Tue May 25, 2021 10:19 am ...
Latest but one green net (first bigger one), 12x3GHz SSE4.1 popcnt 8G hash.
BTW newest CorChess with latest green net was a little faster even, took 1 time 11 and 2 times 17 seconds to solution at three triasl with same hardware.

@ Jon: the position is a changed ERET nr. 75 fortress, changed to White wins, that's the line I've stored in my database:

3r4/3r4/2p1k3/p1pbPp1p/Pp1p1PpP/1P1P2R1/2P4R/2K1Q3 w - - 0 6

6. Rxg4 hxg4 7. h5 Rh7 8. Qh4 Kd7 9. e6+ Bxe6 10. Re2 Re8 11. Re5 Rf7 12. Rxc5
Rg8 13. h6 Kd6 14. Rxa5 g3 15. Kb2 1-0
?? What are the first 5 moves from

how did you discover that this is a modified Eret 75?
http://talkchess.com/forum3/viewtopic.p ... 37#p797137
Peter.
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: A 187 Elo point gain over previous version ...

Post by MikeB »

This is the story of the new SF net versus the old SF net at game/search depth 1. A gain of about 280 Elo! ( amazing) That is incredible and it really does change how we can do things for weaker human play. First it is more green , less power is needed.
10,000 game completed in less than 5 minutes

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05260825.pgn
Time Control: Time Control-> base+inc: 60+0.600
Games: 10000
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/26/21 : 08:29:42

Projected-> Time: 15h:33m:52s
     Run -> Time: 0h:4m:23s

10000 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ   %     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Honey-v13.2-SE     3640   0.0    9    9 10000 8317.5  83.2 8088 1453  459  80.9   4.6  3360
   2 Honey-v13.1b     3360 279.2    9    9 10000 1682.5  16.8 1453 8088  459  14.5   4.6  3640
---------------------------------------------------------------------------------------------------------

LOS:
                  Ho St
Honey-v13.2-SE      100
Stockfish-260521   0

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
Next ,we have this same one-ply monster ( pun intended) play current SF with UCI LimitStrength set 1525 ( preliminary testing showed this was about equal.)
Note: Game was set 3 minutes plus 2 seconds - Blitz
This is now, by far my most favorite engine to play against. It feel like you are playing this cocky kid who keeps handing your head back to you on a silver platter. Major kudos to the SF team for this net

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05260919.pgn
Time Control: Time Control-> base+inc: 180+1.500
Games: 1000
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/26/21 : 11:05:13

Projected-> Time: 4h:30m:45s
     Run -> Time: 1h:45m:41s

1000 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Honey-v13.2-SE-SD1     3507   0.0   21   21  1000  519.0  51.9  511  473   16  51.1   1.6  3493
   2 Stockfish-260521-1525   3493  13.1   21   21  1000  481.0  48.1  473  511   16  47.3   1.6  3507
---------------------------------------------------------------------------------------------------------

LOS:
                  Ho St
Honey-v13.2-SE       88
Stockfish-260521  11

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
 
This version of Honey is new rebase, starting with current SF and adding back key functionality one step at a time.
Although it version it is named Honey, Bluefish fans will not be disappointed as many of the BF features ae incorporated, including a tactical setting that is on by default, that is incorporated with a new Tactcial_Depth setting ( idea found in Shahsin and a key component of Deep Profound Analysis found in previous versions). And now even with tactical setting on by default, it plays about equal to SF. I have found Tactical_Depth of 5 best for everyday use when combed with Tactical setting 3, equal to multipv set 8), but you can mix and match any combination you want.

And a new setting Variety setting that can be set to turn off a various game stages that you decide before you start the game. The settings are 0 to 16, Zero represents no variety , a setting of 16 represents that 16 pieces are off the board ( 16 pieces remain on the board.

Below is match of full strength Honey ( with Bluefish features, i.e., using both tactical and tactical depth options set to 3 and 5 respectively ) versus current full strength Stockfish.

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/05260141.pgn
Time Control: Time Control-> base+inc: 30+0.300
Games: 5000
Threads: 1
Hash: 16

Current date : time (EDST)
Date: 05/26/21 : 05:42:49

Projected-> Time: 3h:53m:23s
     Run -> Time: 4h:1m:36s

5000 game(s) loaded
Rank Name              Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Stockfish-260521   3500   0.0    6    6  5000 2501.5  50.0 1125 1122 2753  22.5  55.1  3500
   2 Honey-v13.2-SE     3500   0.3    6    6  5000 2498.5  50.0 1122 1125 2753  22.4  55.1  3500
---------------------------------------------------------------------------------------------------------

  Δ = delta from the next higher rated opponent
  # = number of games played
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
                  St Ho
Stockfish-260521     53
Honey-v13.2-SE    46

#########################################################################################################
###                                                End                                                ###
#########################################################################################################

Image