Implicit conversion from float -1.6562107904642431 to int loses precision