Implicit conversion from float -31.19047619047619 to int loses precision