Implicit conversion from float -10.071428571428555 to int loses precision