Implicit conversion from float -33.21428571428571 to int loses precision