Deep Position Analysis or DPA

Introduction

When practicing openings I thought it would be cool to be able to generate a tree consisting of best moves from any given positions.

And doing that recursively would allow a deeper analysis.

I searched a bit and only found DeA from Fritz which sadly wasn't free. I didn't find any open source alternative.

And so I created my own script which allows you to do exactly that !

Overview

• Analyze any fen or epd position with any installed engine.
• Analyze pgn files, analysis will start at the last position from main line.
• Customizable depth (be carefull it grows exponentially)
• Customizable number of best moves to explore (MultiPV)
• You can choose between exploring each position for X nodes or X seconds
• Customizable threshold to stop exploring after a side gained the upper hand.
• Supports engine configuration
• Supports multiple games/positions per file
• Supports multiple files

How to install

It's really easy : just download the .py and it's dependencies

Install dependencies

pip install python-chess==0.23.10

Clone repository

git clone XXX

How to use

Basics

It's easy, you only need to know about the command line parameters needed :

• --depth : tree depth in plies. --depth=4 will explore position 2 moves deep
• --pv : pick the top-pv moves from each position. Equivalent to multiPV in UCI settings
• -p/--engine : path to engine [mandatory]
• --nodes : stop exploring each node after set amount of nodes
• --time : stop exploring each node after set amount of time in seconds
• --threshold : stop exploring node further if the score exceeds the given threshold for either side (centipawns)
• --appending : append foreshadowed continuation to last nodes. Off by default.
• a file in epd or fen format OR a pgn (the analysis will start from the last node of the mainline)

To edit engine uci config, edit the .cfg created in the directory after the first use of the said engine.

About PGN

If a pgn is feed to the script the output pgn file will contain all the content from the original file.

The analysis (starting from the last node of the main line) will be appended at the end if --appending is set.

Important

Total number of positions to analyze is given by formula below :

The growth is exponential with depth so take care

Example

python3 dpa.py -p "C:/Engines/stockfish 10/stockfish_10_x64.exe" --pv=2 --depth 3 --nodes 1000000 sicilian.epd

Will generate a .pgn of 7 nodes where each node will be the best move selected after 1M nodes by stockfish 10.

Advanced

For more commands use --help

It is possible to process multiple fen at once by passing multiple files or by appending each fen to one file (one position per line) or both.

You can export the raw tree using --tree if you want to process it.

I found a bug

Are you using python 2 ?

Then use python 3 and don't forget the dependancies.

It's taking too long !

Sadly you have to blame the maths.

The total amount of nodes calculated is going to be , where npm is the number of nodes per move.

The growth is exponential so try lowering the nodes count, the pv, or the depth.

It's a bug I know it

Submit an issue, I will try to fix it :)