Implicit conversion from float -15.89488636363636 to int loses precision