Implicit conversion from float -51.21673003802283 to int loses precision