Implicit conversion from float -4.186046511627907 to int loses precision