Page 1 of 1

To Detrend or not to Detrend? That is my question.

Posted: Sun Sep 05, 2004 10:02 pm
by flex
I'm very confused regarding the subject of detrending.
I've read & re read the "Essay on computed contracts" that CSI supply with the Unfair Advantage software & understand that Detrending removes upward or downward biases. In order to pin down my problem I'll use the 10Yr Note as an example which I've already back adjusted.
I trade using a long term trend following system (variation of Aberration)
I realise that Detrending "removes a bias that that may give the false impression that buying is always better than selling" BUT in the case of the 10Yr Note (or 5Yr Note or ED for that matter) since inception it has predominately been in a bull mkt, so of course the results are going to show a bias toward buying over selling.
The question is; Do I assume my un Detrended results are flawed just because there's a bias towards long trades OR are my results sound?
The difference in results are large enough to be a major concern.
Undetrended = 34K profit/lot since 1988 while Detrended = 15K profit/lot over the same period.

Posted: Sun Sep 05, 2004 10:41 pm
by Tim Arnold
I've wondered the same thing myself. I've concluded to not detrend.

My current thinking is that the prices the simluation uses for entries and exits given my average trade duration of 2-4 weeks are the actual price variations that I would have traded. If the contract rolls, I would have done that at about the spread indicated in the back adjustment. So if there is a bias over a 10-20 year period it should not affect the results of my individual trades.

Put in some slippage for contract rolling and you will best simulate what you need to emulate.

Test what you trade, and trade what you test I think is the philosophy.

Good luck,


Posted: Mon Sep 06, 2004 12:46 am
by TC
Back-adjusted continuous contracts (BACC) provide the best simulation of what your system would experience when rolling contracts in real trading. These back adjustments can lead to negative prices and therefore the absolute numbers clearly do not represent actual prices at that time. You must look elsewhere if you need actual historical prices.

However, if your want to find out how your system is likely to perform in the future, the performance stats you derive using BACC will, in my opinion, provide the best approximation for performance going forward.