lex-game 0.1.4

LEX is Martin Gardner's Hexapawn where matchboxes are Excel files

LEX --- Learning EX-a-pawn

A game written by Mattia Monga for a 'Coding for lawyers' course. Copyright 2020 - Free to distribute, use, and modify according to the terms of GPLv3.

The code is simplistic but this version uses an OOP style; see branch master for an even simpler approach.

Installation

pip install lex_game

lex_game -h gives help on command line arguments.

Rules of the game

Pawns move and capture as in chess, but there are neither two-step moves nor en-passant captures. Players win by reaching the last row or by blocking the opponent. Moves are given by two letters: the starting column and the ending one.

Machine learning

Automatic players can learn to play better by memorizing its experience in Excel files (the format was chosen as an example of using Python with spreadsheets).

With the command:

lex_game -e2 exapawn-empty.xlsx exapawn-001.xlsx

player 2 plays with the experience found in exapawn-empty.xlsx, at the end of the match, all the learned experience is saved in exapawn-001.xlsx.

It is possibile to make two automatic player play one against the other, see learn1vs2.sh for a match with several games. After a dozen of games, player2 becomes unbeatable. This is a typical trend for a match with 30 games.

game	Player 1	Player 2
1	1
2	2
3	3
4	3	1
5	3	2
6	3	3
7	3	4
8	3	5
9	4	5
10	4	6
11	5	6
12	5	7
13	6	7
14	6	8
15	6	9
16	6	10
17	6	11
18	6	12
19	6	13
20	6	14
21	6	15
22	6	16
23	6	17
24	7	17
25	7	18
26	7	19
27	7	20
28	7	21
29	7	22
30	7	23

The game can also be solved deterministically. See tree-solved (cyan is player 1, magenta is player 2) for a spoiler of the three pawns version. With four or five paws for each player the game tree is much bigger, but nevertheless it can be produced by the program in a few minutes.

References

See: https://www.gwern.net/docs/rl/1991-gardner-ch8amatchboxgamelearningmachine.pdf