The problem is that a positive result would only show that a specific sequence of attacks worked well. It wouldn't show that "chi" or other unusual models were required to explain it; there could be perfectly normal explanations for why a series of attacks was effective.