I've been reading The Direction of Play, by Kajiwara, and a particular comment of his has gotten me thinking about joseki and game databases in general. The specific comment, for those with the book, is on page 49, the commentary on diagram 31.
Given this board position in an amateur game:
His suggestion is that black respond to
with the three space high pincer at 
, since it balances well with the other corner and leads to a good result for black if white takes the corner at the 3-3 point. Playing the two space high pincer instead, as happened in the game, gives black problems later in the top left.Normally, as I understand it, the 3-space high pincer isn't played very often in professional games because it's too soft on the approaching stone, but in this case he feels it is exactly the right move. Of course, black has made several mistakes in the game already to lead him to this point, and a professional would not generally be in a situation where this came up in the first place.
This makes me wonder, how many other uncommon joseki or nonstandard responses are discounted via things like pro game database searches because professionals avoid the situations that would make them the best response? Situations, of course, that amateur players are much more likely to play into. And by extension, are we shortchanging ourselves by looking at joseki dictionaries or pro game databases and deciding that a move isn't good because it isn't listed or common?
IMO, to try to choose among pro moves based upon frequency is straining at gnats. Now, if a move isn't listed in a large database, that usually is because it is not a good move. And if a move is listed only a handful of times, it may be a mistake. Pros do make mistakes.