Implicit conversion from float -20.32258064516128 to int loses precision