Implicit conversion from float -9.574468085106389 to int loses precision