It's not simple in my opinion.
Black wants to create a passed pawn on the queenside which would decoy White's pieces so that Black can then play ...Bxe4 and take on f5 and g6 as well, but it needs to be subtle. The immediate 1...c4?! allows 2.b4 and it's hard to see how Black can win.
So if I was Black I would bide my time for a bit, maybe play ...Bb7 and ...Ba6 intending ...a4 at the right moment, see how things go.
I may well be completely wrong. How would other people try to win this?
Black wants to create a passed pawn on the queenside which would decoy White's pieces so that Black can then play ...Bxe4 and take on f5 and g6 as well, but it needs to be subtle. The immediate 1...c4?! allows 2.b4 and it's hard to see how Black can win.
So if I was Black I would bide my time for a bit, maybe play ...Bb7 and ...Ba6 intending ...a4 at the right moment, see how things go.
I may well be completely wrong. How would other people try to win this?
It's black to play. Flip the chessboard if you must to understand the flow of the pawns.
If pawns are dead locked and cannot be captured by the bishops, than they either get remove from the position or aim to capture them by the king. I tend to simplify a position to make it easier on an engine to give a solution.
It could look like this and become a draw:
lichess.org/analysis/8/8/2b5/ppp1k3/4P3/PP2K3/8/1B6_b_-_-_0_1?color=white
If pawns are dead locked and cannot be captured by the bishops, than they either get remove from the position or aim to capture them by the king. I tend to simplify a position to make it easier on an engine to give a solution.
It could look like this and become a draw:
lichess.org/analysis/8/8/2b5/ppp1k3/4P3/PP2K3/8/1B6_b_-_-_0_1?color=white
<Comment deleted by user>
We must also imagine if the game position was simplified like this. (Pattern recognition)
lichess.org/analysis/8/8/8/ppp1k3/8/PP2K3/8/8_b_-_-_0_1?color=black
lichess.org/analysis/8/8/8/ppp1k3/8/PP2K3/8/8_b_-_-_0_1?color=black
Not the best solution, but driving the Bishop to a6 and then advancing the pawn to c4 is to me easier to understand and still a winning move. Not that difficult, as Black has 3 vs 2 in the Queen side.
Notice that Black Bishop and King are threatening the center White pawn chain at its base, but Black chain is virtually unbreakable, so the material balance of 5 pawns is not really working for White.
Notice that Black Bishop and King are threatening the center White pawn chain at its base, but Black chain is virtually unbreakable, so the material balance of 5 pawns is not really working for White.
Candidate moves played by kibitzing. (Picked only top candidate move)
[White "Maia-1900"]
[Black "Maia-1900"]
[Result "*"]
[FEN "8/6p1/2b2pP1/ppp1kP2/4P3/PP2K3/8/1B6 b - - 1 45"]
[SetUp "1"]
[PlyCount "31"]
45... b4 46.axb4 cxb4 47.Bc2 Bb5 48.Kf3 Kd4 49.Kf4 Be8 50.Bd1 Bc6 51.e5 fxe5+
52.Kg4 Be8 53.f6 gxf6 54.g7 Bf7 55.Kf5 e4 56.Bh5 Bxb3 57.Kxf6 e3 58.Bf7 Bxf7
59.Kxf7 e2 60.g8=Q e1=Q
[White "Maia-1900"]
[Black "Maia-1900"]
[Result "*"]
[FEN "8/6p1/2b2pP1/ppp1kP2/4P3/PP2K3/8/1B6 b - - 1 45"]
[SetUp "1"]
[PlyCount "31"]
45... b4 46.axb4 cxb4 47.Bc2 Bb5 48.Kf3 Kd4 49.Kf4 Be8 50.Bd1 Bc6 51.e5 fxe5+
52.Kg4 Be8 53.f6 gxf6 54.g7 Bf7 55.Kf5 e4 56.Bh5 Bxb3 57.Kxf6 e3 58.Bf7 Bxf7
59.Kxf7 e2 60.g8=Q e1=Q
Used Nibbler self-play, but used the engine that was located ...\LucasChessR\bin\OS\win32\Engines\maia
This one was done using Nibbler 2.4.3 in Electron 9.4.4.
github.com/rooklift/nibbler/releases
Engine used: ...\lc0-v0.30.0-windows-gpu-opengl\lc0.exe
github.com/LeelaChessZero/lc0/releases
Settings Used: Self-play, 100000 nodes, maia-1900.pb.gz
There are newer versions for both of them. I should put them in my RSS feed GUI. That way I know when they get updated.
lichess.org/CHanR2iR/
github.com/rooklift/nibbler/releases
Engine used: ...\lc0-v0.30.0-windows-gpu-opengl\lc0.exe
github.com/LeelaChessZero/lc0/releases
Settings Used: Self-play, 100000 nodes, maia-1900.pb.gz
There are newer versions for both of them. I should put them in my RSS feed GUI. That way I know when they get updated.
lichess.org/CHanR2iR/
Updated GUI: nibbler-2.4.4-windows
Updated engine: ...\lc0-v0.31.0-rc1-windows-gpu-opencl\lc0.exe
Set: Threads 3, Hash 1G.
8/6p1/2b2pP1/ppp1kP2/4P3/PP2K3/8/1B6 b - - 1 45
stopped, halt, ucinewgame
Lc0 v0.31.0-rc1 built Mar 26 2024
Found pb network file: C:\Program Files\Chess\engines\lc0-v0.31.0-rc1-windows-gpu-opencl/753723.pb.gz
Creating backend [opencl]...
OpenCL, maximum batch size set to 16.
Initializing OpenCL.
Detected 2 OpenCL platforms.
Platform version: OpenCL 1.2
Platform profile: FULL_PROFILE
Platform name: Intel(R) OpenCL
Platform vendor: Intel(R) Corporation
Device ID: 0
Device name: Intel(R) HD Graphics 4400
Device type: GPU
Device vendor: Intel(R) Corporation
Device driver: 20.19.15.4624
Device speed: 1100 MHZ
Device cores: 20 CU
Device score: 612
Device ID: 1
Device name: Intel(R) Core(TM) i7-4500U CPU @ 1.80GHz
Device type: CPU
Device vendor: Intel(R) Corporation
Device driver: 5.2.0.10094
Device speed: 1800 MHZ
Device cores: 4 CU
Device score: 512
Platform version: OpenCL 1.2 CUDA 9.1.126
Platform profile: FULL_PROFILE
Platform name: NVIDIA CUDA
Platform vendor: NVIDIA Corporation
Device ID: 2
Device name: GeForce 820M
Device type: GPU
Device vendor: NVIDIA Corporation
Device driver: 388.73
Device speed: 1250 MHZ
Device cores: 2 CU
Device score: 1112
Selected platform: NVIDIA CUDA
Selected device: GeForce 820M
with OpenCL 1.2 capability.
Loaded existing SGEMM tuning for batch size 16.
Wavefront/Warp size: 32
Max workgroup size: 1024
Max workgroup dimensions: 1024 1024 64
running, self_play, go nodes 100000
Final Pgn Results:
Lc0 v0.31.0-rc1
1 inaccuracy
0 mistakes
0 blunders
23 Average centipawn loss
98% Accuracy
Lc0 v0.31.0-rc1
0 inaccuracies
0 mistakes
0 blunders
7 Average centipawn loss
99% Accuracy
If another engine can do better post it. I'm done ...
Updated engine: ...\lc0-v0.31.0-rc1-windows-gpu-opencl\lc0.exe
Set: Threads 3, Hash 1G.
8/6p1/2b2pP1/ppp1kP2/4P3/PP2K3/8/1B6 b - - 1 45
stopped, halt, ucinewgame
Lc0 v0.31.0-rc1 built Mar 26 2024
Found pb network file: C:\Program Files\Chess\engines\lc0-v0.31.0-rc1-windows-gpu-opencl/753723.pb.gz
Creating backend [opencl]...
OpenCL, maximum batch size set to 16.
Initializing OpenCL.
Detected 2 OpenCL platforms.
Platform version: OpenCL 1.2
Platform profile: FULL_PROFILE
Platform name: Intel(R) OpenCL
Platform vendor: Intel(R) Corporation
Device ID: 0
Device name: Intel(R) HD Graphics 4400
Device type: GPU
Device vendor: Intel(R) Corporation
Device driver: 20.19.15.4624
Device speed: 1100 MHZ
Device cores: 20 CU
Device score: 612
Device ID: 1
Device name: Intel(R) Core(TM) i7-4500U CPU @ 1.80GHz
Device type: CPU
Device vendor: Intel(R) Corporation
Device driver: 5.2.0.10094
Device speed: 1800 MHZ
Device cores: 4 CU
Device score: 512
Platform version: OpenCL 1.2 CUDA 9.1.126
Platform profile: FULL_PROFILE
Platform name: NVIDIA CUDA
Platform vendor: NVIDIA Corporation
Device ID: 2
Device name: GeForce 820M
Device type: GPU
Device vendor: NVIDIA Corporation
Device driver: 388.73
Device speed: 1250 MHZ
Device cores: 2 CU
Device score: 1112
Selected platform: NVIDIA CUDA
Selected device: GeForce 820M
with OpenCL 1.2 capability.
Loaded existing SGEMM tuning for batch size 16.
Wavefront/Warp size: 32
Max workgroup size: 1024
Max workgroup dimensions: 1024 1024 64
running, self_play, go nodes 100000
Final Pgn Results:
Lc0 v0.31.0-rc1
1 inaccuracy
0 mistakes
0 blunders
23 Average centipawn loss
98% Accuracy
Lc0 v0.31.0-rc1
0 inaccuracies
0 mistakes
0 blunders
7 Average centipawn loss
99% Accuracy
If another engine can do better post it. I'm done ...