Implicit conversion from float -20.89494163424125 to int loses precision