Implicit conversion from float -20.564516129032242 to int loses precision