Implicit conversion from float -37.678571428571416 to int loses precision