Implicit conversion from float -49.13043478260872 to int loses precision