Implicit conversion from float -24.64285714285714 to int loses precision