How to pick the periods for Walk Forward testing when you are longer term trader

L

lmseldin

I back test everything I do. Recently, the concept from an old seminar, mentioned the "training period". I like that concept to think of it, as that is where my original concept would be tested and optimized. Then, I could go forward with dates to see how it performed in the future.

Previously, what I used to do, is just back test over a single 20 year period and examine the equity curve, apr, draw down, # of trades, etc. However, that has been hard to get the APR to a high enough position (for me) on longer term trades.

I am thinking of now doing walk forward testing, so that maybe performing back testing on sections of data that are smaller than the original larger training period will give better results.

However, I would appreciate suggestions on how to logically segregate data into chunks better for longer term trend following trading. I do understand, the need to ensure there are a large # of trades. Also, please consider that I now do portfolio ranking to run backtests on a large # of assets.

Your help is appreciated.

Thank you,
Larry
 
How to pick IS to OOS periods ? You do optimize it, so you optimize the In-Sample length with optimizing to parameters to you OOS walk forward period, in relation to the IS period, and look at the OOS results based on your criteria, Sharpe or Net Profit or anything else you want. Earlier I used to use Diamond Backtesting from Burkhard Eichberger at https://profsoftware.com/bt/ (only to give you an idea how to do it, if you have no plan and no accurate other software here). And you can do this IS to OOS period optimization instead of single symbols also on portfolios of them. Have fun on backtesting.
 
lmseldin said:
I back test everything I do. Recently, the concept from an old seminar, mentioned the "training period". I like that concept to think of it, as that is where my original concept would be tested and optimized. Then, I could go forward with dates to see how it performed in the future.

Previously, what I used to do, is just back test over a single 20 year period and examine the equity curve, apr, draw down, # of trades, etc. However, that has been hard to get the APR to a high enough position (for me) on longer term trades.

I am thinking of now doing walk forward testing, so that maybe performing back testing on sections of data that are smaller than the original larger training period will give better results.

However, I would appreciate suggestions on how to logically segregate data into chunks better for longer term trend following trading. I do understand, the need to ensure there are a large # of trades. Also, please consider that I now do portfolio ranking to run backtests on a large # of assets.

Your help is appreciated.

Thank you,
Larry
More...
you cannot back test with Ninja:
https://ninjatrader.com/support/helpGuides/nt8/?discrepancies_real-time_vs_bac.htm
 
You must log in or register to reply here.
Back
Top