Arbitrage in the InTrade Dem VP Market
Koleman Strumpf August 19th, 2008
There has been an unexploited arbitrage opportunity in the Intrade Democratic VP market (”2008 Democratic VP Nominee (others upon request)”). As the attachment shows, you can sell the slate of candidates for 123.2 (just sum the bids) while you will only have to payout 100. This possibility has existed for at least three weeks, and is particularly puzzling now given that the announcement is likely to occur this week.
What is also a bit odd is that Intrade has another market (”2008 Democratic Vice-Presidential Nominee (with Field contract)”) on the same outcome which includes a catch-all field contract which does not have the same arb–again see the attachment below. It is substantially cheaper to buy the field contract in the second market than the omitted candidates (Kaine, Sebelius, Hagel, Schweitzer, Gephardt, Kerry, and others) in the first market.
Any thoughts on why this is occurring?
attachment: intradedemvp_summedbidsexceed100.pdf
Koleman, have you considered Intrade’s fees in your calculations?
As best I can tell, InTrade calculates your downside risk for each contract individually, so if you sell
40 contracts at 5.0, then your “risk” (and more importantly, frozen margin) is 95.0 x 40 = 3800 . That
means a $20 payoff for a $380 dollar investment where the total is off by 2! (share prices sum to 200)
Since you cannot buy or sell baskets, the payoff on arbitrage is terrible.
-
I did the analysis on the VP market when the sum of the last share trades were 151.2 - sounds like
easy money right? Well if you count how much margin you have to put in, there is only a 1.15%
payoff. That number only gets worse when you consider how big the rake is on expired shares. A
share that drops from say 3 to 0 has a 33% InTrade rake just on the closing fee.
-
There are better trading opportunities out there.
I would argue this is an example of why multi-outcomes markets should be unified rather than split up into separate markets, one for each outcome. There should be easy ways to buy and sell baskets of outcomes. This inneficiency is partly a symptom market design.
[...] The structure of InTrade’s commissions …?… [...]
Thanks to everyone for the reply (and I hope this will be a useful case study for David).
For me the puzzle still remains why one VP market has the arb and the other does not –> the better behaved FIELD market has many contracts so the margin-induced difficulty that Jason pointed out should be a problem for it as well. It would be interesting to look at the dynamics of the two markets: perhaps the FIELD market also had a similar buys summing to >100 arb early in its existence.
Can’t it just be that having fewer contracts (trading for a longer period of time) mitigates the commissions issue? Also we know that (potential) Intrade traders are capital-constrained and I suspect they are relatively uninterested in small % gains, even if they are imminent. Dunno, just doesn’t seem too mysterious.
My first thought was the same as other commenters: the arbitrage opportunity is too small once you factor in transaction fees.
However after thinking about it more, I believe that transaction fees alone cannot really explain this. A trader can get exactly the same expected value (23% return) by flipping a coin and selling one of the candidates at random. It’s not a guaranteed win, but it’s a mathemetically provable positive expected return bet with much smaller transaction costs.
This should be good enough for traders who are close to risk neutral for the amounts they are betting.
An even better strategy is to sell the candidate you think is most overpriced, though then the expected return is not guaranteed to be positive as above.
I also agree with Koleman that it’s strange that the two market groups are different. It might simply be a matter of liquidity, or perhaps some sort of psychological effect of having an explicit “field” contract.
To say it again.. if you adjust all bids by reasonable execution costs (-.6 to -1.3 for example, keeping a floor of zero) it does show that the opportunity is exhausted by those costs, because you can actually end-up walking away with less than 100, and the difference between the two markets is accordingly not very mysterious. Selling a candidate at random is not materially different: 23 divided by the number of candidates (54) minus the commission assumption does not sound like a positive number.
I agree that you want to find those candidates that are overpriced. That was not too hard this past week. Clinton was 16 bid on Thursday and Bayh+Kaine at least 8 bid combined for about a half-hour after the NBC story broke on friday. Largely because of the access and capital constraints and subsequent ease of manipulation, these markets aren’t as “smart” as they could be, and inefficiencies should not be too surprising.
[...] One week later, the debate is still raging under his post. [...]
Thanks Jason for clarifying. But why do you “[divide] by the number of candidates”?
I just meant if someone wanted to attempt to harvest the extra 23% on the bid by selling a candidate at random, they would divide in order to get that random candidate’s share of the 23%. Of course the excess is not distributed evenly and we would rather try to identify the most overpriced candidates.
Thanks Jason. The way I’m currently thinking about it, though I could be wrong, you’d get an expected 23% return on any of the candidates you chose.
Every candidate is on average inflated by 23% — some more, some less. But if you pick one at random, on average you’ll have a 23% return.
Even by that logic there would only be six candidates worth selling for an average post-commission profit of 2ish points, so it’s not surprising that traders didn’t try to exploit that strategy.
What’s the deal with Hillary on Intrade?
http://joshbarro.blogspot.com/.....trade.html
So an unforeseen circumstance (Contract rule 1.9, I believe) only affects that one market? Shouldn’t it affect the basket of markets?
-
Otherwise InTrade is specifically encouraging a non-100 sum of probabilities across markets, because Contract rule 1.9 returns a non-zero payout.
Ok, now I think I see the problem. Suppose the price is inflated from 1 to 1.2 (20% too high). When selling, you have to leverage 98.8 dollars to win an expected 0.2 dollars, only a 0.2 percent return. Is this the root of the problem?
Thanks Jason. I’m confused about your comment. What is contract rule 1.9 and how does it cause non-100 sum markets? This is what I found:
http://www.intrade.com/jsp/int.....rules.html
1.9 0-100 Contract Specifications
General Reduced Tick Size
Minimum Price 0 0
Maximum Price 100 100
Tick Size 1 .1
Contract Value USD $0.10 USD $.01
All Other Contract Specs:
Individual contract information is contained in the contract specifications pop-up window, available by clicking on the contract symbol.
Margin Information
If applicable initial margin will be calculated using the margin rates published on the contract specifications pop-up. Otherwise Initial margin will be calculated as the worst-case possible loss on your open positions and orders for the event.
Order Delay in In-Running Markets
Orders on in-running Financial Markets are held for 0.5 seconds.
When a member cancels an existing order the cancellation is processed immediately.
The rationale behind this industry standard feature is to protect people who enter orders into a market from losing money as a result of access to slower information over which they have no control e.g. a delayed TV signal.
I think the commissions are the main root of the problem here but yes that’s how it manifests itself. I think the other Jason was referring to the old rules that were replaced earlier in the year. I think he meant that if candidate X is the favorite and Y the longshot, if both are immediately settled if X drops-out for any reason, there is no real incentive to bid-up Y. However, if only X is immediately settled, then X’s price does not reflect the outside chance of dropping out — but Y’s does reflect the chance of X dropping out. Thus they would both tend to be too high.. of course you can still sell both if they total more than 100 after commissions.