Implicit conversion from float -91.93069306930693 to int loses precision