My guess is that most of you recognize the oft-referred to scenario of the "Prisoner's Dilemma." This somewhat over-simplified game theory concept has enjoyed huge utility in just about every academic discipline over the past few years. Below, I apply it to analyze the problem with vote pairing websites. Enjoy.
Here are our two players: a Kerry supporter in Texas (hereafter KT) and a Nader supporter in Wisconsin (NW). KT's goal is an electoral college victory (with a secondary goal of a popular vote victory to guarantee Kerry's mandate). NW's goal is a larger popular vote share for Nader. Both (as in all PD scenarios) have two choices, cooperate or betray (i.e. follow through on their vote trader promise or lie and still vote for their candidate of choice).
Let's look at the outcomes for all possible permutations of choices. If NW betrays (which makes KT's goal 1 fail), KT can succeed somewhat on goal 2 by betraying, or not at all by cooperating. If NW cooperates (goal 1 succeeds), goal 2 with also be helped by KT betraying. Either way NW decides, it is better for KT to betray (this is called a "dominant strategy"). You can see that either way KT decides, NW can also improve Nader's vote share by betraying, so NW's dominant strategy is also betrayal. The equilibrium point (as always in PD), is betrayal by both candidates, even though if they both cooperated, both goals would succeed.
Here's where we break away from that classical over-simplification, however. Unlike PD, we aren't limited to a one-on-one model. There's nothing to prevent a Kerry voter pretending to be a supporter from Texas, and then one from Utah, and then one from Idaho, etc, and pairing up with every Nader supporter that he/she can find. Ditto for a Nader supporter (note that Nader is used only for convenience--holds true for Cobb/Badnarik/what-you-will as well).
I'm not really trying to advocate against using these sites (if you're considering them, you've probably already decided that your vote isn't worth much where you live), I'm just pointing out that you're putting a lot of trust in an anonymous process with no mechanism to prevent betrayal. If you end up deciding that game theory can't tell us much about human nature and that you believe that deep down inside every Nader supporter is a good, honest person (or if you've never considered the potential for spamming/manipulating these sites), it's probably because you yourself are a better human than I.