Implicit conversion from float -7.714285714285708 to int loses precision