Hedge-Fund Phrase: Unprecedented, unique circumstances
Translation: Stuff happens. But we had no clue.
Anyone who read the best seller The Black Swan knows that random geopolitical, financial, and economic events can cause the prices of assets to move in ways that defy history and sophisticated computer models. But it comes as a shock to the brightest minds on Wall Street, especially those who run quantitative-based funds. "Wednesday is the type of day people will remember in quant-land for a very long time," Matthew Rothman, head of quantitative equity strategies for Lehman Brothers told the Wall Street Journal last week. "Events that models only predicted would happen once in 10,000 years happened every day for three days." Strangely, these same models failed to predict the once-in-10,000-year events that roiled the markets in 1997, 1998, 2001, and 2002.
RichardCWilson Dud?
Joined: 08 Feb 2008
Posts: 20
Location: Cambridge, MA
It is true, markets move in new ways that can't always be predicted. I recently sat in on a hedge fund manager pitch by a private equity firm trying to make their foray into the space. I cringed at listening to them confidently say that while they have no real performance yet the backtesting of their quantiative models shows that they should be able to return 40-50% every year...it just isn't true and anyone who has been around the block more than 1/2 a time would AT LEAST want to do some serious homework before moving forward with a black box group that is relying solely on backtesting and a few feathers in a trader's hat.
Backtesting should come with a warning label. It amazes me how many otherwise intelligent people don't seem to realize that once you execute a strategy in the market, it changes the market. The observer is not separate from the subject.
Further, once a strategy is discovered, it spreads, eventually pricing itself out. Even if you think you've had perfect skill in keeping your strategy under wraps, you probably aren't the only person who thought of it. We all get to the next level by standing on the shoulders of the same giants...
Backtesting should come with a warning label. It amazes me how many otherwise intelligent people don't seem to realize that once you execute a strategy in the market, it changes the market. The observer is not separate from the subject.
The problem with the quantitative approach is that you can run a backtest to find positive and negative correlations across equities, fixed income, and commodities, go long or short depending on the correlation coefficient and juice your alpha with leverage (convergence trades). However, these time series happen only once, and you have no assurances that the correlations will hold going forward.
You can always build a hypothesis regressing empirical data, in the long run it has no predictive power because you have not established a model of how these asset prices change over time and how the input weights for asset prices change over time.
There will be a variable with no correlation in your model that will correlate precisely at the wrong time, like selling off liquid assets to cover a margin call on illiquid assets last summer. Or your bid asked spreads widen beyond what was modeled, to where you now lost money on the trade because that asset has become illiquid (Long Term Capital Management).
_________________ "Government is the great fiction, through which everybody endeavors to live at the expense of everybody else."
"Ah, you miserable creatures! You who think that you are so great! You who judge humanity to be so small! You who wish to reform everything! Why don't you reform yourselves? That task would be sufficient enough."
— Frédéric Bastiat
Backtesting should come with a warning label. It amazes me how many otherwise intelligent people don't seem to realize that once you execute a strategy in the market, it changes the market. The observer is not separate from the subject.
The problem with the quantitative approach is that you can run a backtest to find positive and negative correlations across equities, fixed income, and commodities, go long or short depending on the correlation coefficient and juice your alpha with leverage (convergence trades). However, these time series happen only once, and you have no assurances that the correlations will hold going forward.
You can always build a hypothesis regressing empirical data, in the long run it has no predictive power because you have not established a model of how these asset prices change over time and how the input weights for asset prices change over time.
There will be a variable with no correlation in your model that will correlate precisely at the wrong time, like selling off liquid assets to cover a margin call on illiquid assets last summer. Or your bid asked spreads widen beyond what was modeled, to where you now lost money on the trade because that asset has become illiquid (Long Term Capital Management).
In other words bull***t baffles brains occasionally? Or rather, models can not and will not predict radical functions.
RichardCWilson Dud?
Joined: 08 Feb 2008
Posts: 20
Location: Cambridge, MA
There was another article just today in the WSJ about how many instruments that used to move in tandem are now playing their own tune. I wouldn't be shocked if a few more quant funds get hurt this next quarter.
View next topic View previous topic
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum