Implicit conversion from float -20.86956521739131 to int loses precision