Implicit conversion from float -18.59154929577464 to int loses precision