Implicit conversion from float -4.428969359331475 to int loses precision