Implicit conversion from float -20.294117647058826 to int loses precision