Implicit conversion from float -24.538043478260875 to int loses precision