Implicit conversion from float -12.723735408560316 to int loses precision