Predictit.org is a prediction market run by a university as a nonprofit enterprise. They say, "Our fees are necessary to cover the costs related to running this site." In order to have a positive expectation value for your bets on such a site, you either have to know something someone else doesn't know (Trump has cancer) or exploit the fact that other people act irrationally despite having good information (mispricing). It seems to me that a lot of the long-shot bets on this site are mispriced, including the following probabilities:
- .15 that Pence is not the vice-presidential nominee
- .11 that Biden drops out
- .11 that Trump drops out
- .07 that Hillary Clinton runs
Other people have made the same observation about long-shot bias on predictit.
I was thinking it would make sense to make a predictit account and take the underpriced high-probability side of some of these bets, making a small but significant profit with almost no effort. Then I looked at predictit's fee structure. They take 10% of your profits, and they charge you a 5% fee to cash out your money. That makes this type of bet a lot less attractive. For example, if I spend $93 on the Hillary bet, win, and then cash out, they pay me $94.33 after their fees. However, I could make many such bets at once (reducing my risk by diversifying), and keep reinvesting my winnings in more such bets. After, say, 10 cycles of this I would have made a 100% profit, and then I would cash out.
Does my analysis make sense, assuming I'm correct in thinking that these long shots really are irrationally mispriced? Are there other transaction costs that I'm not aware of?