Implicit conversion from float -37.83825816485225 to int loses precision