Implicit conversion from float -48.46153846153848 to int loses precision