Implicit conversion from float -2.876712328767127 to int loses precision