- Joined
- 28 December 2013
- Posts
- 6,390
- Reactions
- 24,317
So did you use the 2008/2009 period as an out of sample data set or as an in data set?
My issue [which may not exist] is that when seeking to build efficiency into a system [any system] that system has no redundancies with which to weather a storm [financial] as seems to afflict markets on a rather regular basis.
Efficiency, cuts all redundancies, as redundancies cost money [and hence reduce] efficiency.
jog on
duc
ducati916, thanks for asking this question "So did you use the 2008/2009 period as an out of sample data set"
Out of sample data set
Out of sample data set takes many forms, testing over many years, testing over clumps of years, spread over different indexes & finally running my strategy over market that I don't trade. I've tested my strategy over many markets around the world & surprisingly get very similar results, meaning the strategy is sound & safe to trade, heck it better be, I've got a lot of money ring on it.
I don't have to tell you how to suck eggs
Out of sample data is simple data the strategy hasn't seen before or optimised against.
This topic
This topic is worthy of its own thread as its a very important topic, getting it right can mean if you trade a good strategy or not. Imagine having a great strategy thrown away because of not testing it against 'Out of Sample' data, where changing a few parameters could make it tradable.
Too complex & time consuming
It would takes many posts for me to explain what I do & how I do it. Strategy development is such a complex subject I've decided to post about the subject so to form a template for other to follow. (in its most basic form)
All the tests
Backtesting, forward testing, optimization & Monte Carlo results singularly have importance, but in the combination is where you find the 'Goldilocks area'. Its important to stress, we are not after the best the strategy has to offer, but the best the strategy can achieve over a variety of trading conditions.
Skate.
So the short answer is yes.
The data represents [well actually is] different economic/political/regulatory/monetary/etc environments.
Are you:
(a) the past, as an echo will re-emerge at some point; or
(b) the current environment is the most important [for the model] type of chap; and
(c) do you model volatility in any shape or form as 'out of sample' or 'in sample'?
The impression I get, is that your system is LONG only. Would that be correct?
jog on
duc
Thank you @Skate I was unaware of Barbara Star and her passion for technical analysis.
A PDF of Barbara Star's article on her CAM indicator is in this link.
Published in Technical Analysis of Stocks & Commodities Jan 2018.
https://www.omnitrader.com/currentclients/proforum/get-attachment.asp?attachmentid=8090
There are Traders Tips to assist traders to code the CAM indicator in many trading platforms if you're interested. Search for Traders Tips Jan 18.
View attachment 91727
I'll be posting some backtest reports for an Amibroker Strategy I'm prepared to give away.
I have commitments for the next few hours, I'll post the results this afternoon.
Skate.
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?