For those of you who detrend your data (basic log of price change method) for backtesting, do you use the detrended series to actually look for buy and sell signals? Let's just assume, for example, you're backtesting an algorithmic definition of double bottoms. It seems to me that looking for these on detrended data would not identify the kind of phenomena one is looking for in the original series. Is this correct, or am I missing something? It seems the ideal thing to do would be to look for buy/sell triggers on the original time series, but to measure the P/L of the trade on the detrended series. Opinions?
i believe dtrended data is better for mean reversion or correlation studies. I can't see how it could help in identifying specifc price levels for trades. it could show extremes
You're right -- this is touched on by David Aronson in his book. He says the detrended data is used for calculating expected returns for the rule being tested but that the signals are generated by the actual time series.
Agreed, however what I'm not clear on is the best way to detrend my data, to facilitate analysis, in the first place. Should I create detrended copies of my historic data or detrend the data on the fly as part of the backtest? This leads me to another question. Are there easylanguage functions, or other tools, out there to detrend data? I looked at this a little around New year and then shelved it. This thread has jogged my memory and brought this back to the front of my mind Thx D