Implicit conversion from float -8.148148148148138 to int loses precision