Implicit conversion from float -20.133843212237096 to int loses precision