Implicit conversion from float -30.384615384615387 to int loses precision