Implicit conversion from float -9.931034482758633 to int loses precision