Implicit conversion from float -9.74226804123711 to int loses precision