Implicit conversion from float -1.538461538461533 to int loses precision