Implicit conversion from float -0.9316770186335361 to int loses precision