Implicit conversion from float -17.913669064748206 to int loses precision