Implicit conversion from float -31.184210526315795 to int loses precision