Implicit conversion from float -1.9444444444444429 to int loses precision