Ported to rust from this
The goal of this program is to be able to determine whether a given connect 4 position is winning, lost or a draw, and in how many moves the result can be achieved. It makes use of a bitboard representation to be able to check for alignments efficiently using bitwise operations.
You can get information about the program using
> help
To set up a position you can use position which will play the given moves from the starting position:
> position 4 4 5
To make some moves from the current position you can use moves/play/move
> play 7 5 1 2
To get a score for the current position use solve, to get scores for all the columns use analyze.
> play 4 4 5
Played columns: [4, 4, 5]
Current position:
.......
.......
.......
.......
...x...
...oo..
> solve
Searching: alpha -9 beta -8 [min -19, max 20]
Took: 30.753895ms, total nodes 234400, kn/s: 7813
pv: 3 6 7 4 4
Searching: alpha 10 beta 11 [min -8, max 20]
Took: 3.227257ms, total nodes 264222, kn/s: 8006
pv: 3 6 7 4 4 5 5 6 4 4 3 3 3 5 3 3 5 5 2 2
Searching: alpha 5 beta 6 [min -8, max 10]
Took: 108.404003ms, total nodes 1139648, kn/s: 8025
pv: 3 6 7 4 4 5 5 6 4 4 3 5 5 5 2 3 3 3 2 3 2 2 6 6 1 1 1
Searching: alpha -4 beta -3 [min -8, max 5]
Took: 399.142056ms, total nodes 4393337, kn/s: 8120
pv: 3 6 7 4
Searching: alpha 2 beta 3 [min -3, max 5]
Took: 2.058254278s, total nodes 21736819, kn/s: 8363
pv: 3 6 7 3 4 4 4 3
Searching: alpha -1 beta 0 [min -3, max 2]
Took: 2.111754726s, total nodes 39002773, kn/s: 8279
pv: 3 6 7
Searching: alpha 1 beta 2 [min 0, max 2]
Took: 2.39107231s, total nodes 58487426, kn/s: 8235
pv: 3 6 7 3
Score is 2, which means 'x' can win in 19 move(s)
Total number of nodes: 58487426
Took 7.102748709s
To test the performance of the solver you can run a benchmark on one of the test files. The bench is run using the strong or weak solver depending on the current setting. The weak solver only calculates whether the position is a win, draw or loss, which makes it faster. You can toggle it using toggle-weak:
> toggle-weak
Weak set to true
> toggle-weak
Weak set to false
> bench ./benchmark_files/begin_easy
You can specify the maximum number of lines to solve if you want a faster result. Use "all" instead of a file path to run all the benchmarks. This searches for benchmarks in ./benchmark_files/.
bench all 100
In the starting position, it can take a long time to find the best move. For this reason an opening book can be loaded, which knows the best moves in starting positions. An opening book is a file where each line is an entry with three values.
position_key best_move score
By default, the program looks for a book ./opening_book.book, but a custom path can be specified:
> load-book ./my_awesome_openings.book
The generate-book command can be used to generate a book from the current position to a given depth (this can take a long time!):
> generate-book 3
The number of threads can be set using the threads command. The threading is not yet optimal, but still gives a decent improvement.
- Improve the multithreaded search.
- Use
Clapto do the argument parsing instead of doing everything manually. This should make the help messages better, and make the code more maintainable. - Improve the opening book generation. At the moment it doesn't work properly if a (partial) opening book is already loaded. Ideally it should just expand the existing opening book, to the requested depth.
- If you have your own ideas that you want to add, feel free to make a pull request or create an issue.