I am currently designing a strategy in which I noticed that if I use a walk forward testing that would use a 1 day Optimize and 1 day Test Period, the backtesting does not function properly.
I noticed that the session bugs and the Optimize period usually includes 2 days. On some occasions it is also done on part of the day for which the walk forward is being made...which practically distorts the data, beacause NT provides data that are presented as being out of the sample, but in reality...to a large extent they are also part of the backtest.
This is a major problem for me...and for anyone that is developing a high frequency strategy.
Can someone please advise on how to fix this?
I am optimizing for ES Mini Futures Contracts with the default trading session.
I myself am located in Europe (+2 GMT) timezone.
I tested this on several computers...with different NT settings. All the time, the backtests associated with the 1 day WF include more than 1 day of test data.
I did an experiment of changing the timzone of my pc to the same as the CME exchange CET and have change from default session of ES 03-15 to 24/5 session template.
In that case, the backtesting was again wrong. It was taking again 2 days instead of 1 to optimize on. in this case the only improvement was that the 2nd day that the backtest was made, was not overlapping with the one from the forward test.
Thank you.
Comment