Implicit conversion from float -96.7741935483871 to int loses precision