Implicit conversion from float -20.192307692307693 to int loses precision