Implicit conversion from float -12.391304347826093 to int loses precision