GPF: Day 1
Let's assume the G.P.F. is not installed yet on your computer. The GPF_Pro.ZIP file is probably stored in your default download directory, otherwise call the Windows Find function (F3 button) to locate it. Double-clicking the GPF_Pro.ZIP icon launches WinZip. The probability that you do not have WinZip installed on your computer is next to nil. However should that happen, you will find a free download on the ZDNet Web Site or any shareware site like Tucows for instance.
Two important files are present in the zipped file you have downloaded: The GPF_Pro.XLA and the GALIB32.DLL. The first one is an Excel Add-in. The GPF_Pro.XLA must be registered by Excel to be executable. On the Tools menu, there is a Add-Ins item. You only need to click on browse, point to the destination of the GPF_Pro.XLA file, and then press OK. It is generally recommended to keep Addins in their default directory, i.e. C:\Program Files\Microsoft Office\Office\Library in a default Office 2000 environment (different in later versions). Once registered, the Add-in can be activated or deactivated from the same menu.
The second file is GALIB32.DLL, a 32-bit run-time library which contains the Ward Systems GeneHunter code. It must be placed in your Windows\System32 or WinNT\System32 directory. It will not run in older 16-bit environments.
Once activated, the little Piggy appears on the main menu bar. You are now ready to give it a go !
The Genetic Pattern Finder is now installed. Let's give it a try right now. It may only take three clicks to get it going !
First call up Excel'97/2000/XP, and open either the supplied SP500.XLS or CSCO.XLS spreadsheet or your own data file. Once loaded, click the piggy on the menu bar, the main form now appears. Press "Start GPF" or "Enter" as it is the default key anyway, and a message reminds you that you have not loaded the data in memory. Working off the actual Excel data would slow down the G.P.F. optimization to a grind. The G.P.F. instead reads the entire spreadsheet in memory. The increase in performance is simply phenomenal. Ward Systems users of GeneHunter will undoubtedly notice the difference.
In order to read it properly, you must specify the exact data location and format, including the number of header lines. If your stock market data is always under the same format, it may be worthwhile saving new default parameters into the GPF Windows Registry keys. The option is offered on the "Other Settings" tab on the same form. If the data format is correct, just press "Load Data in Memory", or press "Enter" again, as it is now the default key.
Now that the data is loaded, why not running it with default optimization values? Just press 'Start GPF' (or "Enter" again , the optimization process is now on its way.
You may have now realized that, as a matter of fact, if your stock market data conforms the default data format, you only have to press "Enter" 3 times to get it going!
The genetic optimization process lasts a short while after which it stops with an optimal solution. Close the "Process has stopped..." message window, and click on the "Analysis" button (or press Enter, as it is the default button now). The Analysis form now appears with a list of best patterns listed in increasing order from top to bottom. Selecting a pattern triggers the translation of the best individual into a workable pattern, calculates the trade statistics, and allows you to update your Excel spreadsheet.
While it may be tempting to go straight to the highest fitness figure, we believe you should take some time to pick the best signal. The Trade Distribution and the Equity Curve charts are essential tools for that purpose. The Equity Curve is not always an equity curve as such (see Q & A for details). However, the general principle is still valid, an ideal GPF signal will show a smooth upward equity curve, in other words will be stable through time. The trade distribution will also show whether profits depend on large infrequent gains or not. Filtering may then be an option to consider.
More hints and clues coming here soon... or go to our advanced Tips and Hints page
Page last modified:
May 08, 2008