Implicit conversion from float -23.63636363636364 to int loses precision