I recall an interview of somone that had been polled 6 times and said he had intentionally misled the pollsters so as to throw off those changing their politics based on those polls. With this in mind, I can think of a few things that would make such a device a better predictor. First, it's voluntary. Second, if one is playing with their own money, there's a real incentive to contribute honestly. Third, as there must surely be a continum of reasoning behind one's vote, a poll could tend to oversimplify responses subject to the poll maker's presuppositions. Forth, the lure of making money could tend to increase participation.

Now that said, at least one disadvantage is this does not measure the tendency for someone to vote a certain way, but rather how he thinks everyone will vote and is thus a measure of our ability taken in the aggregate as analysts.