Implicit conversion from float -1.3636363636363598 to int loses precision