![]() |
This website will be down for maintenance from Friday May 24th at 6PM MDT until Saturday May 25th at 11AM MDT. We apologize for the inconvenience. If you need assistance during this time, please email sales@ninjatrader.com
|
|||||||
| Strategy Analyzer Support for automated system backtesting and optimization using the NinjaTrader Strategy Analyzer. |
![]() |
|
|
Thread Tools | Display Modes |
|
|
#1 |
|
Senior Member
Join Date: Mar 2008
Posts: 205
Thanks: 0
Thanked 1 time in 1 post
|
Here I am again- struggling to understand the benefits of Walk~Forwarding a strategy.
As far as I understand it if I have a strategy and backtest a chunk of data- say 3 or 4 months it may look nice but if I was to take a period of time - say 2 weeks for the WF to use as a "guinea pig" then shoot those best parameters over to the next chunk of time- say 3 weeks- then the last 2 weeks of that future 3 week period will be used as a "guinea pig " for the next chunk of time and so forth. The idea being that any variations in overall market movement can be accounted for during this process. Is this how it is? I am concerned about that because really the test data is going to be old anyway. Who's to say that the parameters for the last two weeks are going to be tweaked perfectly for the next three? SO here is what I did. ![]() I used the Whitmark Genetic Optimizer and set the parametres to be tweaked and had it run on 32 months data. It was a very positive result. So I ran it on the previous 32 months data with those same parameters so as to achieve ONE set of parameters that would work in all market conditions... even if there were times of lower profitability. I think I now have a strategy that (as far as I know) is robust enough for my liking~though not specifically as fruitful as when tweaked for the last 2 months. What I want is something I can use with confidence knowing I don't have to fiddle with it month after month. Have I missed the point of WalkForward or have I achieved something better? raycam |
|
|
|
|
|
#2 |
|
NinjaTrader Product Manager
Join Date: May 2007
Location: Denver, CO
Posts: 17,458
Thanks: 1
Thanked 106 times in 70 posts
|
raycam,
In statistics, one of the most common problems with models is data fitting. This means you optimized the strategy so well to fit the data set present, but this is unrealistic in the real world. It is equivalent to looking at a chart and saying I would have traded here at 9:30AM. I would have exited here at 10:30AM and had a perfect trade. Walk forward optimizations corrects this by providing you results of the parameters that are not data fitted.
Josh
NinjaTrader Customer Service |
|
|
|
|
|
#3 |
|
Senior Member
Join Date: Mar 2008
Posts: 205
Thanks: 0
Thanked 1 time in 1 post
|
WOW. ok...
Thanks raycam |
|
|
|
![]() |
| Thread Tools | |
| Display Modes | |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| MoGo - new genetic optimiser for NT 6.5 | -Swig- | Strategy Analyzer | 34 | 11-23-2010 09:33 AM |
| walkforward | tamas | Strategy Analyzer | 2 | 08-20-2009 05:54 AM |
| Walkforward Question | Buttermilk | Miscellaneous Support | 6 | 08-18-2009 12:35 PM |
| MoGo - new genetic optimiser for NT 6.5 | -Swig- | Strategy Development | 0 | 06-23-2009 12:31 AM |
| Optimiser Type | dgregor5 | Strategy Analyzer | 22 | 01-20-2008 08:51 AM |