walk-forward testing
Posted: Fri Jun 23, 2006 4:50 pm
jstockman wrote:Dear TRO,
Thank-you for the statistics not radar code. I am digesting it.
I have a couple of questions.
1. What is walk-forward out of sample testing? And do you use it.
Jimmy,
TRO doesn't like backtesting. He trades successfully in a semi-discretionary environment with a set of his renowed indicators and with his smart, experienced traderbrain.
So you should never ever utter words like "backtesting", "walk-forward testing" etc. in front of Avery
In fact, I am a backtesting type of guy instead That's why we are teasing each other with Avery all the time
Walk-forward testing is best descibed in Rober Pardo's book "Design, Testing and optimization of Trading Systems". You can find it here:
http://www.amazon.com/gp/product/0471554464/
Very simply put, wall-forward testing is in fact looking at what your strategy would do after it has been optimized and put to the real, live trading.
The basic walk-forward testing is done by assessing the equity curve gradient and other factors compared accross the in-sample and out-of-sample trading period.
Advanced walk-forward methods involve computing the WF efficiency by dividing the averaged WF objective value (for example, netprofit) by the averaged in-sample objective value. The walk-forward may be done in several time windows, each representing a brand new "live" trading outcome.
Read the book I've recommended above and you'll become more versant in the subject.
Michal