Implicit conversion from float -20.994263862332687 to int loses precision