# Mean reverting strategies

I would like to take advantage of a volatile market by selling highs and buying lows. As we all know the RSI indicator is very bad and I want to create a superior strategy for this purpose.
I have tried to model the price using a time varying ARMA process, with no success for now.

Any other ideas?

-
Not quite what you want, but have you considered a "scale" strategy? Buy and sell at regular price intervals, but always sell one interval higher than where you buy? –  barrycarter Feb 1 '11 at 3:52
I'm not sure asking for strategy ideas is going to garner an appropriate answer. That's like posting on Hacker News, Ask HN: What should my business model be? –  chrisaycock Feb 1 '11 at 4:24
"Those who say don't know. Those who know don't say. Lao-tzu, Tao Te Ching" –  Shane Feb 1 '11 at 4:26
Added a question to meta to discuss this: meta.quant.stackexchange.com/questions/11/… –  Shane Feb 1 '11 at 15:06
This link is quite interesting I think arxiv.org/abs/0808.1710 –  RockScience Feb 2 '11 at 5:54

Your question's title suggests the market prices are mean reverting. I strongly suggest verifying that assumption via one of the usual tests, such as the Augmented Dickey-Fuller test (implemented in the tseries package of R by the adf.test function, and in other R packages, too).

If the market is truly mean reverting, a possible strategy is

1. Detrend the data.
2. Monitor the market for an extreme high or extreme low, based on its historical range.
3. Buy or sell-short the market at those extremes.
4. Cover at a logical point: at the mean or at the half-way point, for example.
5. Repeat.

Detrending is useful to eliminate the long-term trend (in stocks) or eliminate the effects of carry (in futures). "Extreme highs" and "extreme lows" must really be extreme: I look for prices in the upper 90 to 95th percentile or lower 10th to 5th percentile, based on a few years of history.

Buying or selling-short at the extremes is fine ... unless the market decides to exceed its historical limits, in which case you'll experience drawdown, potentially large. I use a momentum filter and that helps but it's not perfect.