Joined Nov 2005
|
Status: Trading, writing, conquering.
|721 Posts
Try to cover a timespan that accounts for the usual market conditions (don't think you can cover all possible conditions, that is impossible) ... a nice way for example if you have data from 2000-2008, program it to work say from 2000-2002, then "psuedo" forward test it from 2003-2008, this is done to avoid the obvious curve fitting procedure that is a natural process of doing an EA.
More the better is certainly true though, and depending on the strategy, timeframes can have a part in things, as aspects of the markets work differently between shorter and longer timeframes.
Best of luck!
You can quit and they won't care, but you will always know.
|
Joined Mar 2008
|
Status: using the Force
|445 Posts
Ok so if I have a 4h system that performs well but does not perform well in 2004 should I abandon it.?? Markets change and I guess they change back again but you would never use a 5000 period sma because it would not be relevent. So since I can not exponentially test my results I guess I have to assume that the more recent data should carry more weight???
Joined Feb 2006
|
Status: Blah blah blah
|1,410 Posts
You should not abandon it, what is happening now is what is important, but be aware that it will stop working at some point, so you really need a measure of what constitutes working vs. not-working.
The breaking of a wave cannot explain the whole sea.
Ok so if I have a 4h system that performs well but does not perform well in 2004 should I abandon it.?? Markets change and I guess they change back again but you would never use a 5000 period sma because it would not be relevent. So since I can not exponentially test my results I guess I have to assume that the more recent data should carry more weight???
Thanks for your reply
Max
Ignored
It depends on yourself, do you have a long term view and 1 bad year out of several is acceptable?, or are you in need of more consistent month on month profits where one bad year will be the end of you? Best way to overcome it is with some money management techniques that can limit those bad times (but at times will limit the good years slightly as well ... minimise risk).
You can quit and they won't care, but you will always know.
|
Joined Mar 2008
|
Status: using the Force
|445 Posts
OK, so I have a system that when optimized generates 700% over 18 months. Pretty good. Now if I optimize it every two months and use that data for the next month is that not going to be more relevent than using data from 1.5 years ago. I would have thought that would give me much higher returns and less drawdown.
OK, so I have a system that when optimized generates 700% over 18 months. Pretty good. Now if I optimize it every two months and use that data for the next month is that not going to be more relevent than using data from 1.5 years ago. I would have thought that would give me much higher returns and less drawdown.
I will test it anyway but what do you think?
thanks
max
Ignored
I think it won't last long. But test it on demo (or live micro account) and see.
Personally I am not a big fan of EA's they have a shelf life. I prefer to demo trade under s set trading plan (strategy) that is manually operated. Trading in the present allows you to adapt to the ever changing market. Now I am not saying that money can;t be made with an EA but I have never had consistant growth myself when I used an EA. But I do with the Manual method.
Joined Nov 2005
|
Status: Trading, writing, conquering.
|721 Posts
Essentially, trading with an EA (which I don't, mind you) and manually trading with a set strategy is no different, other than the EA enforcing compliance to that strategy if coded correctly.
Manually trading a strategy or letting an EA do it for you, either way without underlying market structure understanding, adjustments to any strategy will be hard to make, and understanding why you are now losing money, when last week you were gaining it will be just as illusive.
You can quit and they won't care, but you will always know.
I am not to well known on this forum, but I like to glean from the net.
If we were at a table having a coffee or tea I would first tell you to optimize your strategy for stability and reducing risk, after you are sure of that, then optimize your stategy for trade management, (how you move your stops while in a trade), after that you can optimize your stategy for position sizing and how and when you compound.
OK, so I have a system that when optimized generates 700% over 18 months. Pretty good. Now if I optimize it every two months and use that data for the next month is that not going to be more relevent than using data from 1.5 years ago. I would have thought that would give me much higher returns and less drawdown.
I will test it anyway but what do you think?
thanks
max
Ignored
i'm having the same questions as yours. it's look like a tug-o-war between reliability (long-term tests) and optimal settings (closer to present market conditions).
i made a quick test on the monthlies ranging from 6 months to just one month. it seems like that the shorter the length of the backtest, the more profitable it is for the next month...
do you have plans to test it out on the weeklies too?
i'm curious about the results. i'll post some if i see something worthwhile. thanks.
