WFO Test periods overlapping?

[TOOLMachine462]

Hello,
I am running a WFO in strategy analyzer and the Walk Forward tab has the test period results listed but the test periods are overlapping. It is testing days multiple times and adding that into the overall results. I am doing optimize over 3 days and then test over 2. Is my optimize test period messing it up or are these results expected?

[NinjaTrader_ChelseaB]

Hello TOOLMachine462,

The overlap is expected if the Optimization period is larger than the test period. The optimization period is the data used for the next walk forward test over the test period.

So if you have the optimization period set to 3 and the test period set to 1, there will be a 1 day overlap with each test as the input data will be 3 days and the test period will be 2 days.

As this walks forward, each test will use 3 days for the optimization and will apply that to each 2 day test.

[TOOLMachine462]

I thought it would optimize over 3 days then test over 2, then optimize over 3 more days which would include the 2 that it just tested over plus the day before, then test over the next 2 days, and so forth.

So your saying to get accurate results, I need to WFO over an equal number of optimization and test days?

Is there a phone number I can call to talk to a support tech about how to use the WFO usefully and to it's designed intent?

Thanks

[NinjaTrader_ChelseaB]

Hello TOOLMachine462,

NinjaTrader Support is outbound only. You can send an email to platformsupport [at] ninjatrader [dot] com and request a call.

"So your saying to get accurate results, I need to WFO over an equal number of optimization and test days?"
Using 3 days for the optimization and 2 days for the test would cause a day of overlap, however, I do not agree that is is not accurate for the settings used.

[TOOLMachine462]

Ok thanks. Is there a way to run a test to simulate an optimization period of so many days, then a test period of so many days, but not have the test days overlap? I want to see the results of the optimization schedule without overlaps. How can you tell if a strategy is working when you have the same period being tested multiple times with different parameters and then added into the final results? How is this useful? Thanks

[NinjaTrader_PatrickH]

Hello TOOLMachine462,

Thank you for your response.

I do not see overlap in the Walk Forward Test results. Can you provide a screenshot of the overlap you are seeing are you end?

[TOOLMachine462]

Actually it was just a bad session template causing it to test days multiple times. Also want to watch out for starting session too early. The indicators don't build correctly for first part of session if there isn't enough time before session starts. Just fyi. Thanks!

[TOOLMachine462]

Hello TOOLMachine462,

Thank you for your response.

I do not see overlap in the Walk Forward Test results. Can you provide a screenshot of the overlap you are seeing are you end?

Use Default 24/7 template and look at trades for each test period, there are days that are tested in multiple periods. The WF tab shows one set of dates but then the trades for that period show other days. This causes false realities!!! :-) It can be corrected by following the posts where you adjust the session templates to proper times. But also need to leave appropriate time before your strategy starts to trade so indicators can build accurately. Otherwise they are building from previous session, in the beginning, correct?

[NinjaTrader_PatrickH]

Hello TOOLMachine462,

Any session that would span across multiple days on your PC local time would cause such behavior. The Defaul 24/7 is set to 12 am to 12 am Eastern. If you ran that in Pacific time zone it would have an overlap of three hours.

[TOOLMachine462]

Stategy analyzer BUG

Hello TOOLMachine462,

Thank you for your response.

I do not see overlap in the Walk Forward Test results. Can you provide a screenshot of the overlap you are seeing are you end?

[IMG] free upload image[/IMG]

If you note the date range of the highlighted row at the top area, and the date range of the trades taken for that period are not the same range. When you correct the session template, the dates of trades are within the appropriate period range and you get different P&L. Has this been brought to the attention of the development department?

[NinjaTrader_PatrickH]

Hello TOOLMachine462,

Thank you for your response.

Can you provide details on the following?
The Performance window in your screenshot, what is this from? Is this the Optimization or Walk Forward trades?
For the settings you changed can you provide the results with one setting and results with the other? Please make sure to include the settings. Changing settings will change the profit and loss, so this would be expected.

[TOOLMachine462]

The Performance window is from the upper highlighted WF trades.

If I run the same WFO with the same parameter ranges I will get different results. It doesn't ever give the same results twice, so I can not reproduce the results with different session templates. You can reproduce this yourself with any WFO. Just analyze the trades for each period under the WF tab. In my experience, if you change the session template to start at 12:00AM and end at 11:59pm, it changes everything in WFO. Try the default set to 12am to 12am and then try 12am to 11:59pm. Make sure you look at the trades taken for each period. They will not match the date ranges for that period when using default session template. Surely I'm not the first one to point this out but I can't find a related thread.

Thanks,
Landon

[NinjaTrader_ChelseaB]

Hello Landon,

This would be the session bleed effect. A template that starts at 12:00 AM and ends at 11:59:59 would end on the same day. If it ended at 12:00:00 PM it would end on the next day causing NinjaTrader to load the entire next days session.