Feature ranking is completely wrong.
Re-read Kantardzic p68: Implement this properly in the trial runner.
Fix Wiki Report, restate properly how features are selected for removal in the round-robin trial.
Also, get rid of existing workstation results in the report.
Fix other wording as encountered.