Implicit conversion from float -25.17857142857143 to int loses precision