Implicit conversion from float -6.724137931034477 to int loses precision