Implicit conversion from float -31.730769230769226 to int loses precision