... that is the question. I have an automated, intraday, mean-reversion strategy (for US stocks) which works best when the market chops and ranges up and down with medium amplitude swings (i.e. of the order of 30/60 minute ATR). If I can find one, I want to add another filter that will switch off the strategy when the probability has increased that price will break from the range. Any suggestions for ideas I could explore? Any comments on the proposed approach? Thanks.
Why switch off? This is when you can make most of the money. This will also be the filter. Switch to position trading when there is a break-out.
Thanks, intradaybill Agreed; other strategies need to take over when there is a break-out. But are you saying "don't worry about trying to switch off the mean reverting strategy in advance; if you get 'hit' with a losing trade just exit at your stop; and then aim to make it up (plus some) on the position trades" ...?
More or less. But this leaves mean-reversion out if done properly at least in the daily time frame. I am working on a dual time frame method now which uses 30 minute bars for mean reversion and daily bars for position trading/filter.
I'm in the middle of reading "Trading Regime Analysis" by Gunn, and I've gotten a few ideas so far (untested) from it about what can help predict a ranging situation: oscillator divergence large round numbers test of major support or resistance a trend that is corrective on a higher time frame may have more ranging behavior bollinger band contraction/expansion cycles downsloping ADX ATR not at either extreme larry williams percentR indicator end of a clearly impulsive elliot wave sequence He also suggested using triangle patterns and squeeze setups to anticipate trending. And a few untested ideas of my own: volume climaxes after a strong move often precede ranging periods High Volume Nodes or vpocs on a volume profile (either long term composite or daily profile) seem to lead to ranging behavior as price rotates around them. large cumulative delta divergence over a period of time And of course there are GARCH, EGARCH, etc which I don't know a lot about but seem to be popular. I think for predicing ranging behavior, though, you'd want to try to use them to predict the ATR rather than the Volatility.