Implicit conversion from float -0.4137931034482847 to int loses precision