Implicit conversion from float -10.384615384615387 to int loses precision