Implicit conversion from float -49.919354838709666 to int loses precision