Implicit conversion from float -9.925650557620827 to int loses precision