Black Hole Solitaire Solver

We’ve ran the solver on the first 1 million PySolFC Black Hole Solitaire deals and one can download the data from that as a compressed textual dump of an SQLite database.

Here’s a description of the tables of the database:

1. bhs_runs - this table contains the results and statistics of running the solver on the boards. Its columns are:

   1. idx - the index of the PySolFC deal.
   2. status - a character representing the verdict - ‘U’ for unsolved, ‘S’ for solved.
   3. num_checked - the number of iterations that were checked (“num_times”). If the game was unsolved, then it will be equal to num_generated (see below).
   4. num_generated - the number of states generated (and not necessarily checked). May only be different if the deal was solved.

2. bhs_solutions - this table contains the compacted solutions for the solved boards. Its columns are:

   1. idx - the index of the deal (that must be ‘S’ on “bhs_runs”).
   2. solution - this is the compacted solution. The 0,2,4 etc. characters are the indices of the stacks moved in the range 1-17. The 1,3,5 etc. characters are the corresponding moved cards after calculating chr(1 + $suits_to_n{$s} * 13 + $ranks_to_n{$r}) where the suits are ordered H,C,D,S. It will total 51*2 characters. Here’s the script that was used to generate them from the more verbose solution files.

29 October 2020: Version 1.10.1 was released featuring a compile-time and run-time option to determine the maximal number of playable cards, which is primarily of interest for impossible deals. This feature was inspired by Jon Wolter’s article titled Experimental Analysis of Black Hole and All-In-A-Row Solitaire over on the Polymorphic Solitaire site (Thanks!).

Version 1.8.0 was released with some fixes and enhancements, and while featuring a new wrapper of the C library for Python which is also open source.

Version 1.4.0 was released with some optimisations and cleanups.

Version 1.0.0 was released with some relatively minor changes and cleanups, but a new major version digit.

Version 0.22.0 was released with hopefully better handling of failed memory allocations, as well as a refactored test suite.

Version 0.20.0 was released with recycling instances support and some speed improvements.

Version 0.18.0 was released with added support for three variants of Golf.

Version 0.16.0 was released with a fix for a crash and some code cleanups.

Version 0.14.0 of Black Hole Solitaire Solver was released. The highlights of this release are setting an SOVERSION to the shared library, some bug fixes and moving the repository to Git and GitHub and enabling Travis-CI.

Version 0.12.0 of Black Hole Solitaire Solver was released. The highlights of this release are an improved API, many cleanups (the code requires “-std=gnu99” from now on), and an ability to be built as an RPM.

Version 0.10.0 of Black Hole Solitaire Solver was released. The highlights of this release are the new --display-iters-step flag, which allows for monitoring the progress of the solutions, and fixes for the build system with version 3.0.0 of CMake.

Version 0.8.0 of Black Hole Solitaire Solver was released. The highlights of this release are the new --version flag, and the --rank-reach-prune flag, which can improve performance.

Version 0.6.0 of Black Hole Solitaire Solver was released. The highlights of this release are the inclusion of the NEWS file, the --help flag and most importantly the --display-boards flag that displays the intermediate board layouts after every move.

Version 0.4.0 of Black Hole Solitaire Solver was released. The highlight of this release is the fact that it can now also solve All in a Row solitaire. As a result of this effort, we have run the solver on the first 1 million PySolFC deals of All in a Row Solitaire and have drawn some conclusions:

• Statistics from the First 1,000,000 PySolFC All in A Row Deals

Version 0.2.0 of Black Hole Solitaire Solver was released. The highlight of this release is the porting of the solver from Perl to C which yields a very dramatic gain in speed and memory consumption. As a result of this effort, we have run the solver on the first 1 million PySolFC deals of Black Hole Solitaire and have drawn some conclusions:

• Solving Statistics for the First 1 Million PySolFC Black Hole Solitaire Deals
• Follow-up with number of unsolved deals with positions span below a certain threshold