Implicit conversion from float -12.328767123287676 to int loses precision