have been working on a 4x4 solver that uses a time budget on every CPU core; repo - https://github.com/aviralgoyal42-hash/QuadSolver
This is not a new method, and build on Shuang Chen’s TPR three-phase reduction. The Stocksolver used to take the first reduction it finds, while mine searches for a lot of reductions in parallel and keeps the one whose solution is the shortest.
(I ran the stock solver again as the numbers may differ per scramble and my PC is slower than the one Shuang Chen’s repo mentions)
This solver depends on the machine it is run one, and is nowhere near the most optimal solution a person could create for a 4x4
Feel free to check my repo out and test it yourself, claude code helped me in making this btw
This is not a new method, and build on Shuang Chen’s TPR three-phase reduction. The Stocksolver used to take the first reduction it finds, while mine searches for a lot of reductions in parallel and keeps the one whose solution is the shortest.
| time per solve | average moves | stock TPR on the same scrambles |
|---|---|---|
| 0.3 s | 43.0 | 44.4 |
| 1.5 s | 42.0 | 44.4 |
| 3 s | 41.6 | 44.2 |
This solver depends on the machine it is run one, and is nowhere near the most optimal solution a person could create for a 4x4
Feel free to check my repo out and test it yourself, claude code helped me in making this btw
