Implicit conversion from float -48.970588235294116 to int loses precision