Implicit conversion from float -20.781671159029656 to int loses precision