Implicit conversion from float -52.96296296296296 to int loses precision