Implicit conversion from float -21.103151862464188 to int loses precision