Implicit conversion from float -33.88663967611336 to int loses precision