Implicit conversion from float -50.449438202247194 to int loses precision