I am working to test a strategy using market replay. I have been using backtesting and optimization to develop the strategy further, but now I would like to see if it works in real time. So I have been trying to use market replay data I have been collecting for SPY to do this, but I have quickly learned it is not so easy to match the real time with backtesting. I've been reading some of the prior posts on this to familiarize myself with some of the issues, but I still have some questions as I am having difficulties getting reasonable agreement between the two.
1. Does replay have issues if you fast forward (I'm not talking about using the slide bar) using the 1x to 500x options? It almost seems as if it loses itself if you fast forward using too high of a rate (like 50x or 500x). Is using the slide bar to skip forward and using the fast forward button the same thing? I know that as soon as you slide the bar backwards or forwards, it resets itself.
2. How does the session template effect the market replay? I have tried the "use instruments settings" and "Default_24/7" options, and get significantly different results at times...
3. During my testing, using very short duration parameters (like a few days) with only a weeks worth of data loaded, the replay seemed to work perfect and match the backtesting. But as I go to longer periods, most of the time I only get the first day of replay to match backtesting then the remaining days do not agree. Not sure if this sounds familiar or not.
As a little background, I am using COBC for the strategy to try and better match the backtesting results. I also am using Yahoo finance saved data as my replay data... So how does the quality of this data effect replay performance? For longer term periods, I have historical data exceeding the parameter requirements I am using. That leads me to another question regarding the min bars required. As long as I have the min bars required set to the max bars used in the strategy, I should have better correlation??
I realize its not easy to help troubleshoot these issues as there are many variables including the strategy and indicators themselves. But any suggestions or insight would be greatly appreciated.
Thanks,
Lee
Comment