Implicit conversion from float -50.86956521739131 to int loses precision