[awm]

I'm not sure my point about the statistics has been understood. It's not the sample size that bothers me -- I have seen similar statistics on okbridge and it would surprise me if the results were much different if Ben were to extend this research to a period of years or hundreds of thousands of hands. Let me try to state the problems more clearly. Here are some conclusions that one might think would follow from the statistics.

Myth: You will always do badly when you have a weak hand with a six card major
Reasoning: Virtually all actions seem to lose imps.
Reality: Not every possible action was listed here. On any given hand, the total net imps for all openings MUST sum to zero. Thus alternate, nonlisted actions must be winning rather large number of imps. Probably most of these are multi 2♦ openings on the hands surveyed.

Myth: Weak twos are an ineffective convention
Reasoning: They are net negative over a large number of boards
Reality: Opening multi 2♦ appears to be really effective, partly because it is played only by reasonably good players and people don't know how to defend it. Because of this, all non-multi openings appear to be worse (since net imps must add to zero and multi is a huge winner).

Myth: You should open 3♠ on six card suits frequently
Reasoning: People who bid 3♠ tend to win imps whereas 2♠ bidders lose on average.
Reality: Opening 3♠ with a six card suit is indeed a winner on SOME HANDS. The people opening 3♠ are not doing so randomly, they are backing their judgement. The statistic indicates only that people who occasionally bid 3♠ on six card suits seem to have good judgement and/or be good players. It's not clear that opening 3♠ randomly on all the reasonable 2♠ openings would be effective at all.

Myth: Light openings in the majors are awful
Reasoning: The 1♠ openers (and presumably similarly the 1♥ openers) did awfully on the hands, much worse than any other bid.
Reality: Light openings when partner doesn't expect them/know how to cope with them are pretty lousy. Probably the majority of these hands are not within established partnerships that have methods designed to deal with 1♠ openings on 9 hcp with a shapely hand.

Myth: Artificial methods like transfer openings and multi are huge winners
Reasoning: On many of these hands none of the natural bids did well. Thus people who opened something other than spades or pass when holding spades are doing better.
Reality: People who have good agreements tend to do well. This includes some people playing natural systems, but includes virtually all people playing artificial systems. Thus people who open hands with spades routinely with a moscito 1♥ or a multi 2♦ or a transfer preempt 2♥ will do well because they have agreements. It doesn't necessarily mean that these are "good bids" against people who have well-prepared defenses.

Myth: Passing on hands with six card suits is a terrible idea
Reasoning: The people who pass do routinely extremely badly.
Reality: A lot of the people passing are those who have ridiculously strict criteria for weak twos, or who don't play weak twos (and I'd bet more of those are beginners than are playing some hyper-modern method), or people who don't trust partner enough to open a weak two, or people who are not paying enough attention to their hands to open a weak two. Most of these people are not going to do well in general, and so the net imps for pass will be poor regardless of the actual merits of passing these hands.

Anyways, I'm not arguing with the statistics themselves or the way in which they were collected, or saying that a larger sample would make much of a difference. The point is that these types of statistics appear at first glance to be much more useful in analyzing effectiveness of bidding methods than they actually are. Table results have a lot more to do with the players' skill levels than the methods they use, and since methods and skill level are not independent, this sort of analysis of methods has a number of flaws.