Implicit conversion from float -0.7438016528925715 to int loses precision