Implicit conversion from float -2.682926829268297 to int loses precision