Implicit conversion from float -3.103448275862064 to int loses precision