Implicit conversion from float -31.06109324758843 to int loses precision