Implicit conversion from float -20.60606060606061 to int loses precision