Implicit conversion from float -98.98678414096915 to int loses precision