Implicit conversion from float -10.264550264550266 to int loses precision