Optimization Problem

[wuileng]

I have set up my strategy to optimize two parameters. The first parameter has 3 values, whilst the second has 2 values, giving a total of 6 different combinations.

While the results and calculation of the profit factor seem to be consistent for each run and parameter combination, for some reason or other there are combinations which do not appear.

As you can see from the following video capture, there are 6 combinations initially, and every time I click run, some of the combinations are shown, while others are missing. Eventually, I get lucky with another run which happens to show all 6 combinations again.

http://www.screencast.com/t/NqOqqKAePoU

This is a bigger issue when I try to run optimizations with more combinations, and is really risky if the number of combinations exceeds "keep best # results" as I wouldn't be able to tell if all combinations had been tested that run.

[NinjaTrader_Matthew]

I believe you may have posted the wrong video (the one in your post is from your NS editor report). Do you still have your original video regarding the Optimization? You should be able to check your Jing history if it was previously recorded.

[wuileng]

Apologies, please see:

http://www.screencast.com/t/NqOqqKAePoU

[koganam]

Apologies, please see:

http://www.screencast.com/t/NqOqqKAePoU

Which optimizer are you using?

[wuileng]

I'm using the default optimizer for standard optimization.

[NinjaTrader_Matthew]

Thanks - this has been documented to occur if your BarsRequiredToTrade are less than then stable period of a moving average / indicator.

For example if you run an optimization on the SampleMACrossOver, and set the Bars Required to Trade to 0, but the EMA periods require more bars before the stabilized.

To make sure we're talking about the same issue, can you reset your BarsRequiredToTrade to at least your period values for any indicators and run the test again?

[wuileng]

My indicator's furthest lookback is [1]. I've set BarsRequiredToTrade = 200, and am still having the same problem.

On the contrary, running an optimization on SampleMACrossOver for Fast = 10;20;2, and with BarsRequiredToTrade = 0, I am getting results for all six runs consistently.

[wuileng]

Am also seeing this anomaly - not sure if it's related to the problem or not.

The highlighted parameters are the only ones I have set to optimize. It seems like it is doing that, but that the optimized values are getting assigned to the wrong variables.

[wuileng]

Made another video, this time of another strategy which does not use any indicators or moving averages.

It records 200+ daily price values (within the strategy itself), and trades based on percentiles when the array holding the price values has been populated. There is logic coded so that it can't trade until the required amount of days has been satisfied. To be safe, I have even set BarsRequiredToTrade higher than the greatest number of days required.

This time however, the optimization runs in the middle were consistent (and did not suffer from missing combinations), just the first and last were problematic.

http://www.screencast.com/t/X2uAmRUuB

[NinjaTrader_PatrickH]

Hello wuileng,

Thank you for your response.

I am unable to reproduce the same behavior on my end. Would you mind attaching a strategy to your response that reproduces this issue?