Newsmax “Political Insider” Dick Morris writes:

An analysis of the published polling data on 80 House races indicates that there are 54 districts now represented by Democrats in which Republicans are now ahead and another 19 where they are within five points and where the Democratic incumbent is under 50 percent of the vote. That's 73 likely wins. (The undecided vote always goes against the incumbent, so if a congressman is significantly under 50 percent, even though he may have a lead, he is likely to lose).

Morris doesn't provide any sources for his claims in the first half of the paragraph, so let's just skip ahead to the parenthetical where he explains that Republicans are likely to win all 19 districts in which they are within 5 points of a Democratic incumbent who is under 50 because "[t]he undecided vote always goes against the incumbent." That part is, like much of what Morris writes, false. The “incumbent rule” was never that undecided always break against the incumbent; it was that they usually do so. And there are indications that the rule is increasingly unreliable (as President Kerry could tell you, if he existed.)

But even if it was true that "[t]he undecided vote always goes against the incumbent," that does not mean, as Morris seems to think, that all undecided voters go against the incumbent. Take a hypothetical race that meets Morris's criteria: a 46-43 lead for the Democratic incumbent, with 11 percent undecided. For the Republican to win, s/he must* do more than win a majority of undecided -- s/he must win a whopping 64 percent of the undecided vote.

So, basically, Morris's “analysis” is a crock. And his analogies aren't any better:

We hope that polling will show that the likes of House Majority Leader Steny Hoyer and Banking Committee Chairman Barney Frank are within our reach. We'll take the pawns. But we'd like to snare a few bishops, rooks, castles and knights in our sweep. (We will dethrone the queen, but she'll still be in the House. Next cycle, we win checkmate and take the king). [Emphasis added]

Rooks and castles are the same thing.

* This assumes, of course, that the size of the electorate is static, and that neither candidate sheds support s/he already has.