Zorro can do it all for us but before we get to that the simplest way to start building your own data library is to download what’s available through the Zorro website. So navigate to this download page like so, then scroll down to historical data archives you can download all the data here using these links if you want to, but the full 1-minute data archives will be more than enough to get you started. That’s these four links here they contain the price history of the instruments used in the trading strategies included Zorro as well as a longer price history for the major currencies and a shorter price history for an additional 28 currency crosses. To get this data, just click on each download link to start the download process and you can see that the data comes down in this zipped archive format.

Zorro is the first institutional-grade development tool for financial research and serious automated trading systems. It applies pattern detection, spectral analysis, or machine learning methods to analyze the markets and enter trades. For example, say our script uses a moving average of the last hundred closing prices. In order to begin trading immediately when we put Zorro into trade mode, we would need to calculate that moving average.

Any algorithmic system can be realized with a relatively small script in C code. Python and R are also supported. Tutorials and video courses get you quickly started. Zorro offers extreme flexibility and features otherwise not found in consumer trading software. Therefore, we would need those 100 prior closing prices. In this case, 100 is our look-back parameter that you might have seen in some of the code examples from this video series and in the Robot Wealth courses. Normally, we can download those 100 historical prices from our broker but sometimes our broker won’t supply data or once apply enough to cover the look-back period in that case we can tell Zorro to load the look-back data from our historical data that we have on file and we do that via the preload which looks like this in Zorro code.

In our price action experiment we’ll use the return of a short-term price action trade as target variable. Once the target is determined, next step is selecting the features. When Zorro is in simulation mode the historical data that’s used in the simulation is set via the start date and end date parameters here’s how we use those so in this case a simulation period covers the first of January 2017 to the 31st of December 2017 also note that this preload flag has no effect in test and train mode only when Zorro is trading now if you don’t have data to cover that simulation period you can download it in a couple of different ways which we’ll get to shortly.

Zorro Trader And The Price Curve

A price curve is the worst case for any machine learning algorithm. Not only does it carry little signal and mostly noise, it is also nonstationary and the signal/noise ratio changes all the time. The exact ratio of signal and noise depends on what is meant with “signal”, but it is normally too low for any known machine learning algorithm to produce anything useful. So we must derive features from the price curve that contain more signal and less noise. Signal, in that context, is any information that can be used to predict the target, whatever it is. Now, another really important aspect of data handling in Zorro is the time zone when downloading data with Zorro, the software automatically converts the data timestamps to UTC time regardless of the data sources time zone. This is really nice and convenient because it allows us to use price data from different sources and it means that we have a consistent baseline time zone that we can work from handling time zones can be a tedious affair.

Thus, selecting the features is critical for success – even more critical than deciding which machine learning algorithm you’re going to use. There are two approaches for selecting features. The first and most common is extracting as much information from the price curve as possible. Since you do not know where the information is hidden, you just generate a wild collection of indicators with a wide range of parameters, and hope that at least a few of them will contain the information that the algorithm needs. This is the approach that you normally find in the literature. Trust me and this is another example of how Zorro obstructs the detail so that we can focus on developing trading strategies so next we need to understand how Zorro stores data Zorro stores historical data in the history folder of your Zorro installation this folder here unless of course you’ve set up another history folder via the Zorro.ini or Zorro fix dot ini files.