Implicit conversion from float -26.51785714285714 to int loses precision