Yes, I think 12 hours was the highest difference in your trade list. Most trades differed by about 4 hours.

Something occurred to me that I think is quite significant. I was comparing some charts of Daily AUDNZD from this week. There was a pin bar that occurred on Tuesday. I was looking at the MT4 chart and the Trading Station chart both from my account at FXCM Australia. I have attached them. The first thing that I noticed is that the low of the pin bar was taken out on the Trading Station chart but not the MT4 chart. The next thing that I noticed is that the bars are all a little different. I discussed this with some friends who use other brokers and there charts are also slightly different. Then it occurred to me that data feeds for non-centralised markets are much more varied than I first thought. Of course, it could be NY close time versus UTC close. But I can't find the settings and overall I am suspicious.

What does this mean for us? Z12 was optimised on data feed that jcl uses, presumably through the same API that is associated with Trading Station. Most of us have tried running Z12 either with a different broker or different platform (MT4.) Consequently the subtle differences in how the data is packaged up and delivered may have resulted in the optimasation not being as optimal as it is for jcl.

Personally, what I would like to so is get historical data from the exact feed that I would run Zorro with and optimise the Z12 with it. Then run the demo for a few months. Although I would like to do a full history optimisation, a shorter re-train for the most recent cycle would be enough.

However, I believe the test dates for the Zs are hard coded. So maybe jcl can comment please…Can I re-train Z12 with a custom data file, say just 2013 and 2014 data, and have the optimisation correct to end of data? Or will the default dates prevent me from training Z12?

EDIT: In the MT4 chart there is a Sunday bar. Ignore that for this discussion as I know Zorro can handle weekend trade. Also these screenshots were taken a couple days apart. Then last bar on the MT4 chart is not complete. So ignore it too please!

You can not compare MT4 charts to other charts. MT4 charts are based on local server time, thus the look of their candles depends on with which MT4 server the terminal is connected. The same chart can look very different in different MT4 versions.

Zorro automatically corrects this when trading through the MT4 bridge. But you can not train a system with MT4 data anyway, as it is normally not sufficient. You need a direct broker API connection to download historic prices for retraining.