Define your expected optimization outcomes in plain human language, and verify them extensively in trials. I can write some useful formal properties right here

Define your expected optimization outcomes in plain human language, and verify them extensively in trials. I can write some useful formal properties right here: - there are n openings and m qualified applicants. At the end of assignment, there should be (m-n) applicants without a role. - if (m>n), all roles should be filled - if there are n identical average candidates and m identical ideal candidates, if there are p roles where (p>m), all ideal candidates should get a role - every candidate not selected should be worse than every candidate selected - every candidate not selected should have participated in n trials where n is the number of openings - etc