Implicit conversion from float -20.715789473684225 to int loses precision