Intrade trading (1) – is there a system to be gamed?

Well, those of you who know me well have already heard my humble endorsement for this year's race: I am all for Hillary, despite not being allowed to vote as a non-citizen-permanent-resident.

Despite the fact I can't vote, I still think I am entitled to an opinion; after all, I pay plenty taxes and the President of the United States of America is the Prime Mover of all tax expenses.

According to my humble computations (aided by my faithful laptop), I personally spent tens of thousands of dollars on the War in Iraq. That's computed with the formula:

  {moszoomimglink:marco_iraq_taxes}

In this particular formula, the variables have the following meaning:

  • di,m is the money (dollars, d) spent for Iraq (i) by Marco (m)
  • tm is the taxes paid by Marco
  • di is the amount of money spent on Iraq
  • df is the federal budget

Seeing that all fritter away for the purpose of bringing democracy to a people that doesn't seem to appreciate it that much makes me interested in the election beyond my non-voting status.

Now, I noticed how Intrade was really good at predicting the outcome of the 2004 election, and I am checking it these days for the presidential election. Right now, the leading individual is Hillary Clinton, with a predicted probability of 37.9% of becoming President.

Then I noticed that you could bet on a bunch of different things. This included the two interesting bets of (a) who was going to be the Democratic presidential candidate, and (b) which party was going to win the elections.

Now, barring sudden death or incredibly weird developments between nomination and election, the probability of someone becoming President of the United States should be the probability of their becoming their party's candidate, times the probability their party is going to win.

Not so, it turns out. There is significant variance between this expected outcome and the actual prediction, as exemplified in the chart below:

{moszoomimglink:intrade_accuracy_clinton_pre}

Hm… An imbalance in trading. Shouldn't that be a way to make money? One would think so: after all, the chart is not like a regular stock graph (although it looks a lot like one); instead, it tells you with precision that something is overvalued. If the chart is in positive territory, then people believe it is more likely that Hillary is going to be the next US President than that she is going to win the nomination and the Democratic contender to win the Presidency. If it is in negative territory, it's the opposite.

So, one would think, you can automatically make money by buying the undervalued contract and selling the overvalued one. You stay financially neutral, and you can only gain money.

Let's see how this works. Let's look at all the times the graph spiked up or down, and let's buy and sell accordingly:

Date  Nom Dem Hill Buy Sell
5/20/07 51.2 56 42 Nom/Dem Hill
7/30/07 48 56.6 26.4 Hill Nom/Dem

 Let's try with just this combination for now. The idea is to buy and sell at local maxima and minima, and you automatically get  money. The trick, here, is that you have two different contracts to deal with on one side, and only one on the other. When I say, buy Nom/Dem on 5/20/07, which one should you buy and in what quantity?

In hindsight, looking at the values for our local minimum on 7/30/07, the contract for Democrats has barely moved, and all the motion was in Hillary. Well, that makes perfect sense, since the probability of a Democrat winning the race is tied to factors bigger than one single candidate's faux pas. So we can assume that that contract is more stable and less volatile. We can either use it to anchor our bets, or we can go for the higher volatility. 

Let's take, for the purpose of illustration, the Dem contract. On 5/20/07 we detect that Hillary is totally oversold as President. So we sell Hillary and buy Democrats in equal value. According to the rules, each point is a tenth of a dollar, so if we sell $1000 of Hillary for President, we short 238 … certificates. By buying in equal amount, we get 178 certificates of Democrats winning the White House.

When we see the trend reverse, on 7/30/07, we decide it's time to cash in. We sell our 178 certificates of Dems, cashing in $1007, and we buy the 238 certificates of Hillary we shorted before, for $629. That's a neat and automatic profit of $378.

(By the way, if we had used the volatile contract instead of the stable one, we would have had a buy-sell cycle netting $937, or $70 less than with the stable contract, reducing the profit to $308 – but it could just as easily go the other way around.)