Implicit conversion from float -17.608695652173907 to int loses precision