Testimonals

I’ve just about used every program in the industry over the last 25 years and GSB produces the most consistent out of sample results.

As a former S&P pit trader and early adopter of algorithmic trading, GSB is right in my wheelhouse. I’m fortunate that I can use any platform I choose and, over the last 25 years, I’ve spent a good deal of time with most of them. GSB takes many of these a step further through its unique treatment of time and data management.
Genetically evolving strategies through proper protocols has consistently demonstrated the ability to garner an out of sample edge. However, development and testing are computationally expensive. GSB’s ability to run multiple workers on multiple machines leverages current computer design to the fullest through the “Resource Manager.” I run the workers on the computers behind my local network. The point is the programming architecture makes genetically building algorithms much faster.
More crucial than the speed is the handling of time. We’ve all struggled with what date windows to include based upon our human suppositions of future market behavior. Do we include events like the economic meltdown of ’08? Or, is the bull run in the stock market over the last ten years the new normal?
GSB’s time handling capabilities include the ability to pull a given ratio of days out of any length of period. The ratio approach allows traded (in-sample), and non-traded (out of sample) dates to reside next to each other. This means we can use a percentage of evenly spaced in-sample data that pulls from the market’s multiple personalities over the entire period while reserving the majority of the historical data for out of sample generalization.
Finally, it has been a pleasure working with Peter. His responses on the forum and in private email have always been prompt, despite our time differences. He’s also provided direct support when needed. But mostly, he’s been patient with me as I’ve made all of the rookie mistakes in my advancement with his methodology.
I look forward to meeting some of you on the forum. Good luck in your pursuits.
Sincerely, Andy Waldock.
@waldocktrades

I've purchased a number of various system builders over the past 5 years or so and after discovering GSB some 10 months ago it has become my primary builder for TradeStation systems. I find it a great tool to explore various build options, it's a very fast platform with clever tech harnessing CPU power from cloud works and with the addition of macros that allow you to perform all sorts of tests in a very methodical framework with the benefits of automation. Finally, the proof is in producing the code and tradable systems, this process is a robust process with the systems matching the build metrics and where the trading platform is also delivering comparable trade results!

Bruce Herbert, Auckland New Zealand.

So I’ve had some time to dig in now and I really like what I see!

Quite honestly, I find your product to stand out way above anything commercial that I have ever worked with before. One of the things I do like about it is the ability to customize your settings. At the end of the day, I don’t care what works, as long as I feel comfortable that it does and will likely continue to do so. But the road there often requires creativity in my experience. And, from what I can tell, GSB will give you the freedom to shape this process.

GSB forum user "ProbTrader"

Translated from https://www.financnik.cz/clanky/praxe/gsb---automatizovana-cesta-pro-zkoumani-trhu-a-vytvareni-obchodnich-systemu-r1806/

GSB - Automated way for exploring markets and creating business systems
Today's markets require, in particular at lower timeframes, ongoing research into the broader context of what works in them and what is not. This is enough work to make it as effective as possible by using different tools. There are many, but not all are worth the effort. One of the exceptions that has recently addressed me is the GSB program.

What's worse - trying to blindly trade systems created by hard data mining, or concreting the mind using old approaches? It is difficult to say, but my experience will not lead to any of the above-mentioned paths to long-term profits.

Aggressive data mining, where the trader is constantly going through the same historical data when he finds the system with the parameters he is looking for, leads to the trading of over-optimized systems that do not make money on the new data. Although the use of old and well-known approaches can lead to systems based on meaningful and proven principles, they are unlikely to make any more money, especially at lower timeframes. Simply because the described market inefficiencies in the markets have disappeared as many traders traded over time.

There is a need to go some middle way. Building on verifiable and explainable ideas and not afraid to build new and innovative business approaches.

I like to use the idea of ??the first approach, whereby I build on systems that are understandable and justifiable to me why the profitability of a given approach should remain in the markets. But at the same time I like to help automated searches of the ideal context in which the basic idea works best today (see the Fcontext workflow discussed in the AOS course ). A similar approach helps me discover connections that cannot be tested manually at a reasonable time.

That is why I was very pleased to discover the Genetic System Builder (GSB), whose author Peter Zwag , develops software that focuses in many ways on markets to address similar issues that are important to me.

If the genetic structure of trading systems is said to be, as with me, traditional "datamining" software will probably attack you. Today, there are plenty of programs to navigate through historical data and, to put it simply, to test different combinations of indicators and pricing patterns to get a trading system. As I said, I am quite skeptical of similar roads. Unfortunately, often similar software will find what the trader wants to find. That is, a system with nice historical results, including beautiful OOS data verification, which is simply selected from thousands and thousands of possible systems that the system generates.

Why do I like the GSB then? Because, despite its name, the program is not just focused on genetic system creation. His strength is the analytical exploration of the wider context of what works in the market and what is not. And it is only in this context that we can create a trading system more or less automatically. This philosophy resonates much with what I perceive as important in developing systems, and overall, the program shows that it is an experienced trader based on its own systematic trading practice (Peter's results are verified, for example, in Futures Truth, where they track and compare automatically traded systematic strategies).

Describing the specific functionality of GSB is not easy, because the program can do a lot. And above all, it is constantly being expanded (very actively). It is indeed to know that he is used by the author for his own trading and market analysis. A more detailed idea of ??how to use the program can be obtained from these English videos (free registration is required in the forum), the program can then be tested in a 14-day full-featured demo that can be found here .

So what can GSB do in short?

Like any automated business software creation program, the GSB can take as input, for example, the 30-minute market data of the e-mini S&P 500 (ES) and test their history to create a trading system composed of various indicators and their parameters. What I like about the GSB is the fact that this part is already tightened beyond what is being offered elsewhere. It is possible to work with advanced fitness functions influencing the result of created systems, to verify the systems automatically in selected other markets and timeframe and subsequently to verify them using WFO. In addition, multiple market or timeframe systems can be automatically created, indicators used, various stop-losses, outputs, and more.

Other similar programs offer many of these features, although I think GSB integrates the essential features the trader needs in this area.

What GSB excels in my view is speed and performance. And that is the most important thing in this area.

GSB itself works briskly. We can run it as a classic desktop application, where, of course, the performance most depends on the parameters of the hardware used. And that is always limited, even if we choose more powerful hardware. As a really good and sophisticated step, I appreciate the fact that GSB can work as a cloud solution. In this case, we set up the tested analyzes in the so-called Managers, which are then run by the Workers - calculating instances. These can be on the same computer, on our other computers, or in the cloud - on other computers.

Other cloud GSB users can also serve as a cloud (you can set whether or not we want to join the cloud). This design seems to me to be really good. Mostly similar programs are installed on servers running nonstop (I rent the solution myself, see description here). In this case, electricity consumption and hardware depreciation are included in the rental price and it does not matter if the servers are running idle or fully loaded. By the time tests run out, it makes perfect sense to share performance with others. As soon as I need jump performance myself and the other servers are free, I get a much higher performance that would cost me a lot of money in renting my servers. Of course, shared performance varies over time and depends most on whether your server is being used by Peter. However, I usually get two to three times the computing power with GSB than my own servers provide. This is a significant advantage for a similar solution. In any case, Peter promises that the user will always get at least one cloud worker for free, which in itself can help move the calculations.

Example of the GSB environment. On the right you can see that the calculation is done on a total of 30 workers. Only 5 of them run on my own hardware. I used the others for free at the moment, which of course greatly increases the efficiency of the entire computing process.

The good thing is that Peter provides a free Resource Manager software for the entire solution, allowing you to manage performance - for example, we can set the priority to run your own workers (and others stop). Power sharing is then completely automatic.

I have two groups defined in the Resource Manager - the first is my own work. They have priority and run when I work with the program. In the second line, there are potential cloud workers who run when the server is not busy.

Just mentioned cloud functionality pushes GSB to the options I have not found elsewhere (in the given price category). The high performance available to me as a trader provides the analytical capabilities mentioned at the beginning of the article.

It effectively allows you to find out what works on a given market and what doesn't. By generating a statistically relevant number, for example, tens of thousands of systems, we evaluate different overall system degradations in out of sample data. We do not focus on one system but on a statistical sample. For example, we can easily compare OOS degradation of 10,000 created systems from and on another market. This makes it easy to see if a given authentication element really helps to create more robust systems on a given timeframe, or just a few systems.

Sure, a lot of that can be done in other software. Indeed, I myself have tested some similar principles in Amibroker with OLE automation. But it is with the use of cloud performance that this is really working at a different level.

Thanks to the fact that the program allows some smart analyzes that I haven't seen in other programs yet.

Above all, there is a way to define in-sample and out of sample data that can be used to monitor system degradation in-sample data and out of sample data. Traditionally, market data is shared by, for example, 60% IS and 40% CSO (or similar, according to personal preference). However, it is a problem that the given market periods may be very different in nature. Therefore, I like the idea of ??using OOS data every other day or every other month and so on. The systems are then taught over the entire data period (where the software sees for example every other line) and tested on the other part of the lines. We then incorporate all the crucial phases of the market into the development of the system and the OOS evaluation will have a more meaningful value. In the GSB, this can be set very flexibly.

I perceive the strength of GSB in its complexity, which is usable in practice thanks to cloud performance. For example, many programs can be walk forward testing (wf) and GSB is no exception. But who ever did the wf can certainly confirm that these tests require a lot of performance and therefore time if everything is done on one computer. In the GSB, walk forward can also be done by workers (ie in the cloud) and in reality higher hundreds of wf can be calculated on intraday systems per day. On the cloud, you can also have automatic systems verification on other markets / timeframe, and then only in statistics to see how the system has run on selected markets / timeframe for validation. The result is a similar overview where one can see how stable the results are. You can track the stability of your results on verified other markets and timeframe but also the stability of the results in terms of the parameters used. In this respect, I like Peter's Stability parameter, allowing you to select those systems whose parameters do not change over time (and there is a higher chance that the system will be more robust in live trading).

Generated systems can be transferred directly from the GSB in the finished code to TradeStation.

GSB really does a lot, although it is, of course, just software - a tool whose results will largely depend on its use. It is certainly not "on" software, and after x minutes you get systems ready for live trading. On the contrary. It is a tool that provides various analyzes that need to be considered and spent time. And only then use the result of the work to create a traded model. But which should have a higher chance of robust performance in living markets.

In any case, I recommend watching Peter's videos describing the methodology with which markets are tested. And possibly explore a trial version of the program . It can also bring new stimuli and interesting inspirations to your strategy development. GSB offers enough. And even though the software costs $ 1,500, it is, in my view, a meaningful solution in this highly competitive area.

18.3.2019
Petr Podhajský

Fulltime trader dedicated to trading for over 15 years. Specializing in systematic strategies traded on futures and stocks. Favorite trading style - intraday using orderflow. Last years, the construction of automated systems portfolio.