For the top four US presidential candidates from each party, InTrade.com gives these chances of nomination, becoming president, and the implied conditional chance of winning if nominated: Al Gore’s conditional chances seem way out of line to me. And Ron Paul’s conditional chances are suspiciously large compared to the other Republicans.
I was tipped off to this play, and arbitraged this (with a small amount of money) many months ago.
Note that the margin requirements are sort of crappy, you have to put up your full potential loss a few months before the event. So the shorting half of the play has a pretty significant opportunity cost.
Don't read too much into what amounts to random statistical noise. Markets tend to towards accuracy, but at any given moment, they're wildly inaccurate. As Benjamin Graham noted a long time ago: In the short-term, a market is a voting machine. In the long run, it's a weighing machine.
I've looked at these lines before, in particular the Gore one since it seems so out of whack. One problem is that when the numbers get towards one end or the other, you have to invest a lot of money for a fairly long period of time for a low predicted return.
For instance, suppose you knew that Gore was guaranteed to lose. You can sell contracts, but intrade requires that you keep enough money on hand to cover those if you lose. So you have to keep that much cash on the site in order to get a 5.3% return over a year.
Not terrible, but as the rate drops it gets less and less appetizing. So people stop selling below a certain amount unless the event is really close, and stop buying if it's over a certain amount for the same reason.
So I think you can think of intrade predictions as being bounded from 5%-95% or something as long as the event is far away.
Sorry -- InTrade charges a 10 cent expiry fee on a $10 contract, so there's still a bit of money to be made in shorting. The fees are a high percentage of your expected earnings in shorting low probability events, but don't eat up all of it.
InTrade charges a 5 cent in-the-running expiration fee, right? That makes me worry a bit about the low-odds candidates: there's no money in shorting them at current odds given the expiration fee. Correct me if I'm wrong...
I don't think using a gambling site as a predictor of anything is a good idea. If gamblers are so good at predicting the future, then why are the casinos repeatedly taking in so much more money than they give out?By the way, owners of casinos are not gamblers-- ask one.
The argument that the same sort of good news that would allow a dark horse to win the nomination would also help in the general election makes a certain amount of sense, but that doesn't seem to be what's happening here. Gore's chance of winning the primary is so low largely because he isn't running.
Jason, these are fine measures of the conditional probability stated, but they are not measures of other conditional probabilities you might find more interesting.
I put a close-italics tag ("lessthan,slash,letter-i,greaterthan") between the words. Seems to have done the job.
Doh! I should have viewed source the first time. In attempting to close, I left out only the slash, so I opened italics twice. In the first attempt to redeem myself, I closed one, got confused and gave up, leaving it to you to close my second. Thank you, and apologies to all. We need blogs where all comments are editable by their original authors.
gore at 90% isn't all that crazy of an idea. the only way he would win the nomination is if he builds up a full head of steam and becomes the dominate force heading into the convention... this "momentum" would imply his odds to win the election are very high. if he is just a likable candidate with a middling chance of winning a general election, he won't win the primary because he is way behind in $.
if you disagree with me, trade the spread and make yourself a free lunch.
gore at 90% isn't all that crazy of an idea. the only way he would win the nomination is if he builds up a full head of steam and becomes the dominate force heading into the convention... this "momentum" would imply his odds to win the election are very high. if he is just a likable candidate with a middling chance of winning a general election, he won't win the primary because he is way behind in $.
if you disagree with me, trade the spread and make yourself a free lunch.
So, are you using this market as evidence to update your beliefs about various candidates' chances, or pointing out a situation where a market is giving bad information that you intend to ignore?
It would appear that there are divergences far enough from MY expectations, that if transaction costs were low, I'd be willing to take some of the middles (buy gore nominations and sell gore wins for instance). I presume you're in the same boat.
The question on my mind is: What good is a biased market? Many (heck, most) prediction/wagering markets are constrained in ways that bias the pricing (by allowing partisans to exert more influence than if the market were robust, by not allowing interested parties to include themselves, and likely more subtle effects). How much information can you really get from a thinly-traded market on an emotional/political topic? Put a different way, how can you measure the information efficiency of a yet-to-be-resolved set of wagers?
I was tipped off to this play, and arbitraged this (with a small amount of money) many months ago.
Note that the margin requirements are sort of crappy, you have to put up your full potential loss a few months before the event. So the shorting half of the play has a pretty significant opportunity cost.
Don't read too much into what amounts to random statistical noise. Markets tend to towards accuracy, but at any given moment, they're wildly inaccurate. As Benjamin Graham noted a long time ago: In the short-term, a market is a voting machine. In the long run, it's a weighing machine.
I've looked at these lines before, in particular the Gore one since it seems so out of whack. One problem is that when the numbers get towards one end or the other, you have to invest a lot of money for a fairly long period of time for a low predicted return.
For instance, suppose you knew that Gore was guaranteed to lose. You can sell contracts, but intrade requires that you keep enough money on hand to cover those if you lose. So you have to keep that much cash on the site in order to get a 5.3% return over a year.
Not terrible, but as the rate drops it gets less and less appetizing. So people stop selling below a certain amount unless the event is really close, and stop buying if it's over a certain amount for the same reason.
So I think you can think of intrade predictions as being bounded from 5%-95% or something as long as the event is far away.
Hey, Intrade was way off on the economics Nobel. They did not even have any of the winners up on the board.
Sorry -- InTrade charges a 10 cent expiry fee on a $10 contract, so there's still a bit of money to be made in shorting. The fees are a high percentage of your expected earnings in shorting low probability events, but don't eat up all of it.
With the trade fee, the foregone interest, and the point spread, it is difficult to arbitrage these imperfections out of the Intrade market.
A more interesting measure would be the probability a given candidate would win _if his/her nomination was forced_.
InTrade charges a 5 cent in-the-running expiration fee, right? That makes me worry a bit about the low-odds candidates: there's no money in shorting them at current odds given the expiration fee. Correct me if I'm wrong...
I don't think using a gambling site as a predictor of anything is a good idea. If gamblers are so good at predicting the future, then why are the casinos repeatedly taking in so much more money than they give out?By the way, owners of casinos are not gamblers-- ask one.
The argument that the same sort of good news that would allow a dark horse to win the nomination would also help in the general election makes a certain amount of sense, but that doesn't seem to be what's happening here. Gore's chance of winning the primary is so low largely because he isn't running.
Jason, these are fine measures of the conditional probability stated, but they are not measures of other conditional probabilities you might find more interesting.
Tiiba, that is time to arbitrage.
Dagon, see my added to the post.
I put a close-italics tag ("lessthan,slash,letter-i,greaterthan") between the words. Seems to have done the job.
Doh! I should have viewed source the first time. In attempting to close, I left out only the slash, so I opened italics twice. In the first attempt to redeem myself, I closed one, got confused and gave up, leaving it to you to close my second. Thank you, and apologies to all. We need blogs where all comments are editable by their original authors.
gore at 90% isn't all that crazy of an idea. the only way he would win the nomination is if he builds up a full head of steam and becomes the dominate force heading into the convention... this "momentum" would imply his odds to win the election are very high. if he is just a likable candidate with a middling chance of winning a general election, he won't win the primary because he is way behind in $.
if you disagree with me, trade the spread and make yourself a free lunch.
gore at 90% isn't all that crazy of an idea. the only way he would win the nomination is if he builds up a full head of steam and becomes the dominate force heading into the convention... this "momentum" would imply his odds to win the election are very high. if he is just a likable candidate with a middling chance of winning a general election, he won't win the primary because he is way behind in $.
if you disagree with me, trade the spread and make yourself a free lunch.
So, are you using this market as evidence to update your beliefs about various candidates' chances, or pointing out a situation where a market is giving bad information that you intend to ignore?
It would appear that there are divergences far enough from MY expectations, that if transaction costs were low, I'd be willing to take some of the middles (buy gore nominations and sell gore wins for instance). I presume you're in the same boat.
The question on my mind is: What good is a biased market? Many (heck, most) prediction/wagering markets are constrained in ways that bias the pricing (by allowing partisans to exert more influence than if the market were robust, by not allowing interested parties to include themselves, and likely more subtle effects). How much information can you really get from a thinly-traded market on an emotional/political topic? Put a different way, how can you measure the information efficiency of a yet-to-be-resolved set of wagers?
These are only naive measures of the conditional probability, as David Schneider-Joseph explains here.