Implicit conversion from float -86.50375939849624 to int loses precision