New to ProRealTime, loving it, and I’m a first time poster here. I may be having issues with the ProBacktester, or maybe I’m just doing something wrong. Can someone please help me?
When attempting to code I like the feature Variable Optimization. However when I find a combination I like and would like to hardcode those values into the system for real use, I receive _entirely_ different results when running the same test during the same time period (ex daily) and time frame (1 year). Is this behavior expected or am I perhaps doing something wrong?
I have not experienced this. What you are saying is that you use the variable optimization, take the best values and hardcode them, then run the system again in backtest and receive a different result? By the way, I think you mixed up time period (test period) and timeframe (time per bar).
Hello Wing, thank you for a swift reply. After re-reading my post I see mixed up the test period and time per bar in my description. Thank you for the correction.
I dug deeper into the site and found out about the phenomenon called Zero Bar, which I didn’t even knew existed, and first thought it was due to that issue that the results I got during optimization was good. After thinking about it some more I don’t think that’s the cause, I think the reason why may much simpler. In my code I was using the built in Trailing Stop Loss:
1
SET STOPpTRAILING5
After changing this to the basic Stop Loss I got more accurate and reproducible results:
1
SET STOPpLOSS5
Finally I came upon the two custom made Stop Losses by Nicolas, published here at prorealcode.com, and I’m in the process of testing them out right now in a few scenarios. I think my issue may just be a case of being new to the platform and not yet having experience with what works best and what can be improved in the platform, but I’ll be sure to report back with results!
After a multitude of testing code, targets and stop losses, variable optimizations, etc. over various instruments and time periods I’ve concluded that I can’t trust the results ProBacktest is showing me. The worst offender in this situation seem to be the Variable Optimization feature, then followed by the Targets and Stop Losses. I understand that backtesting is for several reasons never a guarantee for future profits. I would however imagine that the backtests are indicative of the system as a whole and thus should perform in the general direction of the backtests. I have submitted a support request to PRT reporting the above as technical issues, with additional information, and hope to receive a response from PRT soon on how to combat the problem. As I explained to their support; if I can’t trust ProBacktest to perform its designed function in simulation, how can can I trust ProOrder to perform in demo let alone in live trading with real money on the line? I hope to receive a response within a few days and may report back here with additional information, should the case provide any.
In closing I would like to repeat from my first post words that still ring true despite this problem.
.
