Implicit conversion from float -1.3043478260869676 to int loses precision