Implicit conversion from float -137.74647887323943 to int loses precision