An experimental tool to find fastest winning moves using bruteforce for the board game Splendor. It was born after a question on boardgames.stackexchange.com:
Q: What is the fastest possible game in Splendor?
The goal of the board game Splendor is to reach 15 points by either taking chips (gems) from the pool or buying cards with gems that give a permament gem bonus - a 1 gem discount for later buys.
- Create a virtual environment and install dependencies:
python -m venv venv . venv/Scripts/activate pip install -r requirements.txt - Run the script:
python splendor_fastest_win.py
The tool uses breadth-first search to greedily check all possible move sequences. Optionally you can use a heuristic that limits search space of BFS by using only the most promising game states.
Data for all cards is stored in cards.csv. Cards are referenced by a string consisting of a card's point value, one-letter color and a sorted list of its non-zero cost values.
Before the first run the tool caches (pickles) a dictionary that lists all cards that can be bought with each set of gems (possible buys).
Unoptimized bruteforce takes way too much memory and is unfeasible beyond 8 moves, but with a -u key it can get to 15 moves needed to reach the goal. Further optimization is required.
usage: splendor_fastest_win.py [-h] [-u] [-b] [-e] [goal_pts]
An experimental tool to find fastest winning moves using bruteforce for the board game Splendor.
positional arguments:
goal_pts target amount of points
options:
-h, --help show this help message and exit
-u, --use_heuristic use a heuristic formula to limit the search space of BFS
-b, --buys regenerate and store all possible buys
-e, --export export possible buys to a .txt file