Implicit conversion from float -5.032258064516128 to int loses precision