[Tradewinds]

tradestation 9 has Walk-Forward testing to analyze strategies. They recently added the Walk-Forward Optimizer. The tradestation Walk-Forward Optimizer (WFO) is now available to all customers with the recent release of version 8880.

https://www.tradestation.com/support...tion_0900.aspx

• First a Strategy needs to inserted into the Chart
• You will either need to program a strategy, or choose one of the example strategies.
• Once a strategy is inserted into the chart, you will need to format it.

For a video of the process, open this link:

https://www.tradestation.com/support...alkforward.flv

Has anyone used the Walk-Forward testing and the Optimizer? I'd be interested to know what you have learned.

[Tradewinds]

The Walk-Forward Analysis claims to detect if a strategy has been custom fit to just one set of data, and won't work well under other circumstances. The tradestation help states:

"It detects if a system has been overfit or improperly optimized. (Studies have shown that a randomly chosen, poor, or overfit system can make money in one or two walk-forward test, but will not make money over a large number of walk-forward tests) A system that makes profit over a large number of walk-forward tests, is most likely to be successful in future."

reference: TradeStation Walk-Forward Optimizer Help

[Tradewinds]

tradestation recommends that the strategy being tested generates at least 450 trades. The WFO (Walk-Forward-Optimization) uses an Out-Of-Sample size of 20%, so out of 150 trades, the number of Out-Of-Sample trades would only be 30 trades.

The minimum of 450 trades is to make sure that a "10 run walk-forward analysis" can be done.

Then there is the Cluster Analysis (CA) which may need even more initial trades.

[equtrader]

Walk-forward analysis is good but the problem is finding an edge. It's like someone offering you a fancy car audio system as a present but you have no car.

I would like TS to incorporate some data mining capability (not NNs or anything else along that line that fits to the data). The concept behind Price Action Lab for example would be a good addition and walk-forward testing can complement that to screen patterns for robust behavior. What good is it giving a fancy tool to people when their main problem is that they have hard time finding any good strategies manually?

[electroniclocal]

The WFO is a great tool. Problem is the usual: RIRO (rubbish in, rubbish out). The criteria that is used to both enter and exit trades is of great importance. Just using your favourrite indicator with optimised inputs, rigid targets and stops won't in most cases provide a robust model, even if one set of parameters passes WFO. It's usually just a fluke. You got lucky, as Clint would say. This is easily seen as you try and get the same "chosen" inputs after using different lengths of data both for your IS and OS testing.

The way to use the WFO is to use criteris that relates to market activity so that the algo will self adjust to the volatility and to use the WFO to fine tune its sensitivity.

Having said that, TS's WFO is the state of the art after having purchased the Grail and having the guy that owned it integrate it into TS. But its really slow when you compare it to MultiCharts but what you get from it is much better.

EL

[Do Or Die]

Which algo does tradestation uses for optimization?

Amibroker has a WFO engine since quite some time. It's CMAE engine is blazing fast: Amibroker- How to optimize trading system I checked a year back and the same strategy took 30-40% less time in Amibroker for optimization when compare with tradestation.

Would also like to hear how fast is tradestation's optimization algo.

[Do Or Die]

WFO is available in Amibroker since early 2008: Walk-forward testing and optimization

[equtrader]

Originally Posted by Do Or Die »

WFO is available in Amibroker since early 2008: Walk-forward testing and optimization

Amibroker is a great tool. Why even compare to other bulky, overloaded expensive programs?