correlation trading

Jesus, a guy with a 401k and a used car could corner BRL/INR...

Regarding the sign on the Euro major, I think it's negative on USD/EUR but I could be wrong.

uh usually correlation trading refers to correlation between CDO tranches

And that's why I said I'd give the FX perspective... I never said it was the only one.

Correlation trading can refer to anything. You're basically trading for a pair to converge. That pair could be KO & PEP, Realized Vol & Implied Vol, UK swap spreads & Germany swap spreads, CDOs, anything.

The thing to think about is that shorting a correlation (i.e., trading for a convergence, and not a divergence)--this is basically what stat-arb is built around BTW--is effectively shorting volatility, i.e., you are naked short an option on the pair. And it trades that way too. 95% of the time, you make the 1-2% consistently on the convergence. Then the rare event comes, and the pair completely blows up; if you're leveraged, it can easily mean bye-bye birdy.

Somehow, people keep doing it anyway. Obviously there are controls to limit the blow-up (like using options as you go), but they quickly and easily can limit or eliminate the profit, since the perceived edge is often miniscule, and they can be less liquid to trade as well.

Further discussion on this?

There's also a lot of equity correlation trading. Index options have an implied underlying correlation, and based on that, you can trade options in individual names and offest that position in option of that index in order to basically go long or short correlation. There is a trend that most correlation traders exploit called the dispersion trading, and it suppposedly works most of the time. I'm not an expert in it, so google would probably offer a better explaination.

In credit land, correlation trading usually refers to CDO Tranches. The random events whose correlation we're measuring are corporate defaults.

Most generally, the lower tranches will usually have to absorb a certain % of losses due to default, then the next senior tranches begin aborbing, and so on.

Now, consider the environment of increased correlation. The lower tranches will actually go up in value, because the probability of reference entities not defaulting as a group has gone up, decreasing the expected loss on this tranche. Conversly, the likelihood of hitting the higher threshold of losses that will effect the senior tranches has increased, reducing the value of the senior tranche.

The opposite logic would work in decreased correlation. The junior tranches loose value, bc the likelihood of at least 1 default has increased. At the same time, the likelihood of hitting the senior threshold has decreased, making the senior tranches go up in value.