Implicit conversion from float -33.36734693877551 to int loses precision