Implicit conversion from float -104.82758620689657 to int loses precision