Implicit conversion from float -137.67857142857142 to int loses precision