back testing and live data

[mdan]

I have developed scripts that I have then been able to backtest successfully. Next I have been trying to test the backtested strategies with live data as the data (ticks) move forward during market hours. I have noticed slight differences in live data, downloaded data, and in reloaded data, each configures the indicator/trading panel differently. I understand that downloaded ticks from a storage server will have slightly different configurations, and I think that is good enough to replicate real time for testing purposes. But it appears to me that trading strategies sometimes don't trigger during live market hours unless at times the indicator/strategy panel is reloaded with historical data from storage, so is there something inherent in the EFS language about strategies and triggers that only operates with backward testing historical data and does not work with live forward moving data?
Thx
Dan

[demarcog]

Re: back testing and live data

Originally posted by mdan
I have developed scripts that I have then been able to backtest successfully. Next I have been trying to test the backtested strategies with live data as the data (ticks) move forward during market hours. I have noticed slight differences in live data, downloaded data, and in reloaded data, each configures the indicator/trading panel differently. I understand that downloaded ticks from a storage server will have slightly different configurations, and I think that is good enough to replicate real time for testing purposes. But it appears to me that trading strategies sometimes don't trigger during live market hours unless at times the indicator/strategy panel is reloaded with historical data from storage, so is there something inherent in the EFS language about strategies and triggers that only operates with backward testing historical data and does not work with live forward moving data?
Thx
Dan

Dan,

It is true that back tested trading results, because the are based on completed bars loaded after the fact, will not be identical to either tick replay results, or real time trading. However depending on how the strategy is coded, results can be very close.

If you have not already done so you may want to look at the fairly recent guides on backtesting that JasonK wrote. There are specific coding techniques required to mimic real time trading results.

The specific problem you are having I have also seen and was caused by using bars with barindex(0) which in a back test is already complete, but in realtime can change with each tick.

If you posted the code someone would be able to determine if that was indeed the case.

It is a good idea to test a strategy in tick replay where you can use the generic broker functions instead of the strategy object to place trades with the eSignal paper broker, which is about a close to actual trading as you can get.

Glen

[mdan]

Thx Glen for your timely reply, Yes I have read thru most of the JasonK reference material and those plus your comments and others here, have done much to bring me up to speed. Especially your Las Vegas comment about what goes on in a bar stays in the bar.

Dan