Image by longan drink via Flickr
A few weeks back, I was emailing with Woodshedder about the problem of volatility targeting in tradings systems. This cause him to write this better sumary of my point in this post:
Finally, Damian from over at Skill Analytics reminded me that traders can get hit with larger than expected losses when using ATR stops. This would occur when volatility has died down, which would cause the ATR stop to get tighter and tighter. If a system is using a percent-risk formula for position-sizing, as the stop gets tigher, it is buying bigger and bigger positions. If volatility suddenly returns, the system may experience some large losses, until the ATR stop has time to catch up with the market volatility.
Well, I was a bit amused today to read via AllAboutAlpha, that the folks over at the major quant fund AQR recognized this as one of the main reasons for the Quant Bloodbath in August 2007.
According to Asness and Berger, quant managers often target constant volatility not constant dollar exposure. As a result, they lever up in times of low market volatility and lever down in times of higher volatility. The problem last August, they say, was that many quant managers were using higher leverage than usual at that point in the cycle. As a result, a pure Fama/French model with no leverage would have actually performed much better that month than most quants.
It’s important to note that they are talking about using volatility not as a stop in this case, but as a way of judging how much leverage one should have. This is why I believe you need to have minimums in volatility-based position sizing to insure that in low-vol environments you don’t get taken out on a stretcher. So it is rather shocking that these guys would let volatility alone determine the amount of leverage. The fact that these guys overused leverage, however, isn’t really all that shocking – it is the opinion of this writer that many hedge funds are selling alpha while delivering nothing more than beta * leverage.
![Reblog this post [with Zemanta]](http://img.zemanta.com/reblog_e.png?x-id=0e352ec1-0dc7-4b93-8af3-1b65dd3a7b9f)
I gave up on the idea of ATR-based stops and percentage-risk position sizing well over a year ago (two years ago?).
The opposite response happens when markets recover from a downturn – ATRs are wide, so are stops, and positions are small. This implies many, many positions. As the market calms significantly, one has to decide what to do with the multiple positions … which to keep as stops shrink, which to sell in order to buy more of others, etc.
One could always go the math-turbational route that Barra et al suggest with covariance matrices and the like, but that ass-u-me-s the covariances are constant (or forces a resizing).
Much more elegant to just keep it simple. Equal position sizes by dollars, with cash for positions where qualifiers can’t be found.
I’ve been using equal position sizes in dollars for my longer term systems, and I’ve been looking at ATR as a factor in position size – but bounded and again, just one factor. So I’d have a minimum size and a maximum size, and ATR alone would not determine the size of the position.
Bill, for short term systems that expect to stay in a trade no longer than a couple of weeks, ATR stops can work very well.
Due to constant rotation, the system can begin to catch up with the market conditions often on the 2nd or 3rd battery of trades after the turn has occured.
The other consideration is of a psychological one concerning the need to be right, i.e., trading a system with a high winning percentage. ATR stops will often be wide enough to increase the winning percentage on swing trade system, but as you said, they may keep positions small enough that this benefit is not worth the smaller wins.