Implicit conversion from float -10.649350649350652 to int loses precision