Implicit conversion from float -18.5427135678392 to int loses precision