Evaluation Function

Why is so hard to find a perfect evaluation function? Maybe because the standard is so blur that based on our blur standard, our evaluation function is also not perfect. Evaluation function map a situation to a partial order set, but we can see that not every situation can be "mapped" to an element of the partial order set, or this is impossible. So why not discard the mapping idea completely, so that our optimistic technique based on other theory like probability or some comparable theory other than partial order set.