Implicit conversion from float -23.42857142857143 to int loses precision