When benchmarking, the main.py redefines some variables for the game, namely the board size, and number of mines on the board. In the solver, currently the lsqr solver is initialized with an x0= (naive estimate of bomb probability in informed cells)
But this naive x0 is wrong, since it uses the MINES variable from conf.py to create it, when main.py changes the number of MINES for each benchmark case.
When benchmarking, the main.py redefines some variables for the game, namely the board size, and number of mines on the board. In the solver, currently the lsqr solver is initialized with an x0= (naive estimate of bomb probability in informed cells)
But this naive x0 is wrong, since it uses the MINES variable from conf.py to create it, when main.py changes the number of MINES for each benchmark case.