Implicit conversion from float -18.81502890173411 to int loses precision