Introduction

This is the home page for the pgn-extract program, which is a command-line program for searching, manipulating and formatting chess games recorded in the Portable Game Notation (PGN) or something close. It is capable of handling files containing millions of games. It also recognises Chess960 encodings.

A full description of pgn-extract's functionality is available and included with the sources.

Here you can find the C source code and win32 binary for the current version. pgn-extract compiles and runs under Windows, Linux and Mac OS X. This program is made available under the terms of the GNU General Public License (Version 1).

Getting-started video for Windows users

For Windows users who are really only interested in getting the binary working, here is a short introductory video:

Blog post about data mining with pgn-extract

In October 2018 I wrote blog post about using pgn-extract to mine a PGN database. As an example it looks at the effect of having a bishop pair versus a knight pair.

Overview

The program is designed to make it easy to extract and format selected games from a PGN format data file based on a wide variety of criteria. The criteria include:

  • textual move sequences;

  • the position reached after a sequence of moves;

  • information in the tag fields;

  • fuzzy board position;

  • and material balance in the ending.

Over the on-going 20+ year course of its development, it has also added lots of features for controlling what is output (e.g., different algebraic formats, EPD, no move numbers, restricting game length, etc.)

The program includes a semantic analyser which will report errors in game scores and it is also able to detect duplicate games found in its input files.

The range of input move formats accepted is fairly wide and includes recognition of lower-case piece letters for English and upper-case piece letters for Dutch and German. The output is normally in English Standard Algebraic Notation (SAN), but this can be varied.

Extracted games may be written out either including or excluding comments, NAGs, variations, move numbers, tags and/or results. Games may be given ECO classifications derived from the accompanying file eco.pgn, or a customised version provided by the user.

The program is designed to be relatively memory-friendly, so it does not retain a game's moves in memory once it has been processed. This also makes it suitable for bulk processing very large collections of games - it can efficiently process files containing several millions of games.

Use the --help argument to the program to get the full lists of arguments.

New in recent versions

These are the main changes in the most recent versions:

  • Accept Z0 as an alternative to --.

  • Added -y as an alternative to -z for colour-specific material matches.

  • Added --dropbefore to drop ply occuring before a particular comment pattern.

  • With --allownullmoves uppress the warning for null moves.

  • Added --dropply.

  • Added --nestedcomments to allow non-standard nested comments.

  • Corrected failure to OR rather than AND the pseudo Elo tag used with -t.

  • Fixed a memory leak with -z.

  • Made it possible to identify which -z pattern was matched in a game.

Available Files

You can take a copy of the full source and documentation as either pgn-extract.tgz or pgn-extract.zip. Alternatively, a win32 binary is available.

Name Description Size Date
pgn-extract.tgz
GZipped tar file of the complete source of the latest version of the program.
Includes usage documentation, Makefile for compilation and eco.pgn file for ECO classification.
393K bytes  14 Oct 2018
pgn-extract.zip Zipped file of the complete source of the latest version of the program.
Includes usage documentation, Makefile for compilation and eco.pgn file for ECO classification.
547K bytes  14 Oct 2018
pgn-extract.exe win32 binary of the latest version of the program. 687K bytes  14 Oct 2018
eco.zip Zipped version of eco.pgn. 32K bytes  May 24 2013
eco.pgn File of openings with PGN classification.
This file is already included in the source archives.
254K bytes  May 24 2013
COPYING GNU General Public License (version 3). 35K bytes   

Feedback

Feedback and suggestions for further features are always welcome, although I can't always promise to undertake significant development work.