Be careful with mean reversion trading
Mean reversion.
That term comes up in trading and investing from time to time. It basically means that when the market gets out of line with normal pricing or relationship characteristics it will eventually get back in line. For example, a market gets extremely volatile. Mean reversion would suggest that eventually the volatility drops back down to the long run average.
I hate the term. It’s one of the most dangerous ones in trading, and its not really used properly.
Mean reversion is a statistical term mainly based on populations with a normal distribution, though it’s use in a financial context doesn’t necessarily follow strictly along the main definition. The markets, though, do not follow a normal distribution. They have fat tails, meaning the risk of something extreme happening is higher than what would be the case under normal considerations. A great many folks found that out in this last cycle.
Playing for mean reversion is kind of like being an option seller. It might work very well for long periods of time, but every once in a while something unexpected is going to happen that wipes out all of your prior gains, and then some (the figures saying that most options expire worthless does not imply that only shorts make money trading options, by the way).
Mean reversion is also like relying on correlations to persist. It might work for a while, but when things go wrong they tend to go very, very wrong – and often in a way that hard to avoid or moderate.
OK. Enough venting for now.
Struggling with support & resistance and knowing what the key market levels are? Check out the Price Distribution Analysis methods I use.
2 people have left comments
Posted on April 3, 2009 at 10:38 am
David wrote :
I may be guilty of using the term to describe the successful patterns both myself and subscribers to Trade-Ideas have taken advantage of recently. I agree that taking principles that work to describe phenomenon in other unrelated fields and applying them to capital markets is risky, but 1) terms must be described in order to be understood and 2) applying these strategies a good trader understands there is a ‘black swan’
out there that can end a strategy’s trend. You cant tell yourself I wont exploit a strategy because it may end. Like any trend – find it, ride it, and then get off it.
Posted on April 3, 2009 at 10:45 am
Tim wrote :
While it’s worth pointing out the dangers of reversion to the mean, I believe you made quite a few inaccuracies.
Firstly regression to the mean does not require a normal distribution, it doesn’t even require a symmetric distribution.
Also, it does not apply to volatility, which itself is a statistic measure. For the same reason you can’t apply traditional TA to the VIX which many people try to do.
The problem with it is that it assumes that the distribution of prices remains fixed over the period that you use it. Unfortunately, you have no way of knowing this is the case, unless you knew the future. So unsurprisingly, it all it does is convert the guesswork of forecasting stock prices into guesswork for forecasting distributions.