Implicit conversion from float -2.4671052631578902 to int loses precision