Implicit conversion from float -23.868613138686143 to int loses precision