Implicit conversion from float -20.756972111553793 to int loses precision