Automated trading is developing. Analytics in Expert Advisors becomes more and more complicated, while the results become more profitable. At the same time, the writers of EAs start to pay attention to multicurrency EAs and to those based on neural networks.

The great majority of modern Expert Advisors can be subsumed under a group of simple or 'regular' EAs. These are EAs that work with only one currency pair and don't optimize their parameters 'on-the-fly'. These EAs are often based on a well-known trading strategy or its modification. However, such EAs become more perfect and profitable, too.

Darren Merwitz: Multicurrency trading is diversification.

Let's get down to facts. In Championship 2006, the top ten EAs earned $101 650 within 3 months, or 101% of their initial deposits. A year later, the top ten developers submitted the EAs that had earned in average $256 243, or 256%. The healthy growth is present. There are fewer errors in EAs, there are more correct trading decisions, and, as a result, there are more profits.

Multicurrency Expert Advisors

At the same time, the number of multicurrency EAs grows gradually from Championship to Championship. The rate of multicurrency EAs increased from 5.5% in 2006 up to 7.5% in 2007. Totally, 45 multicurrency Expert Advisors participated in the last-year Championship.

William Boatright: Multicurrency Expert Advisors allow you to hedge your trading.

As of yet, multicurrency EAs cannot show great achievements. However, the EA writers go on developing them. The matter is that multicurrency EAs allow you to reduce risks due to the diversification of your deposit. "I would definitely say the multi-currency trading is a better way to go. This goes along with the simple notion of diversification. Trading a single strategy currency is essentially putting all your eggs into one basket", believes Darren Merwitz, FXCM.

William Boatright (Wackena) expresses the opinion that trading several correlating pairs may help you hedge your positions: "I do believe trading multi-currencies can be beneficial. Many trade multiple pairs that have hedge behaviors, such as the 4 majors (EURUSD, GBPUSD, USDJPY, USDCHF)".

At the same time, Olexandr Topchylo and Yuriy Zaytcev believe that, due to the same diversification, larger profits can be achieved. Theoretically, it is possible. However, the analysis of Expert Advisors participated in the contest in 2007 showed that the profitability of multicurrency EAs and that of conventional Expert Advisors were at approximately the same level. Basing on his experiences at IBFX, Andrew Andersen confirms this feature: "Those traders/EAs that focus on many different currencies will often have higher rates of profitability at times and overall may make more than those trading a single currency".

Andrew Andersen: Developers are commited to certain securities.

Andrew also notes that something like commitment to certain currency pairs is observed at present time. The developers of Expert Advisors and ordinary traders trading several currency pairs often work with a certain set of securities. In Andrew's opinion, they only work with well-researched pairs to achieve stability. In other words, it is still too early to speak about the mass scale of multicurrency EAs that can work with any securities. However, the initial steps are taken.

Multicurrency EAs have another important feature that differs them from ordinary EAs - they are much more complicated than traditional EAs both in developing and in testing. "Multicurrency Expert Advisors are more difficult to develop, but I don't consider them to be worse. Unfortunately, MetaTrader 4 Strategy Tester does not allow us to test a multicurrency EA. It may be this fact that also hinders their development," declared Yuriy Zaytcev (YuraZ) who has a couple of multicurrency EAs in his portfolio.

The complicacy of multicurrency EAs consists in several factors. First, they must be able to analyze several securities at the same time. Second, several trading strategies may function within the framework of such EA. Third, some arbitrary mechanism must stand over all this to prioritize one currency pair or another in each unit time. It seems to be quite easy to do, but there are still only few developers that manage to combine all this in one Expert Advisor.

Yuriy Zaytcev: A multicurrency EA is simpler than that based on neural networks.

The situation is dramatized by the absence of standard features for testing multicurrency EAs in MetaTrader 4. Even having managed to develop such a complicated EA, its author has to use only forward testing and manual optimization. This, in its turn, hinders the development of this branch in automated trading even more. The situation should be improved in MetaTrader 5 for which a standard testing feature for multicurrency EAs is planned.

As of yet, multicurrency Expert Advisors, according to many professionals, have good prospects. Let's hope for seeing serious achievements in this field during Automated Trading Championship 2008.

Neural Networks

The things came so that Olexandr Topchylo (Better), the Winner of the last-year Championship, encouraged the interest in only a part of 'self-optimizing' Expert Advisors, namely to neural networks. Within a few weeks, all traders' communities started talking about these networks. The representative of FXCM, Darren Merwitz, notes a sharply rising interest in neural networks and cautions traders against subservience to this fashion: "Neural networks introduce a lot of complexity … I'm very against a trader putting all their faith in a system without fully understanding what it is doing and why it may be running up or drawing down". Nevertheless, it has already become a trend.

Sooner or later, the developers of Expert Advisors face the situation where their profitable program loses its efficiency, so they have to modify its parameters 'on-the-fly'. Otherwise, you will have to test and optimize your EA manually all the time. Olexandr Topchylo himself considers an Expert Advisor without parameters to be the best. However, he immediately admits that such EAs don't exist.

Olexandr Topchylo: The fewer parameters the better.

"Whatever type they are, I prefer trading systems with the lowest possible number of parameters to be optimized. Their future results can be at least approximately estimated in the strategy tester. The test results and the results in real trading may be directly opposed for systems having a large amount of parameters. One of the methods to reduce the number of parameters to be optimized is adaptation of a part of the system parameters, which is laid in the algorithm itself," Olexandr Topchylo believes. So there is a problem stated, and EA writers have already started solving it.

According to many automated trading professionals, a great future is reserved for neural networks and… these EAs are very complicated. Olexandr Topchylo and Yuriy Zaytcev declare EAs based on neural networks to be even more complicated than multicurrency EAs. "In my opinion, a multicurrency EA is simpler than that based on neural networks," Yuriy said.

Situation analysis shows that, due to their high complexity, EAs based on neural networks are developed by very few EA writers. Many experienced developers have just started taking their first steps. Such developments became active after the last-year Championship.

For example, William Boatright who has taken the second place in the contest informs that he has written only three relatively simple neural network EAs after the last Championship. He confesses that he didn't research this field before ATC 2007.Yuriy Zaytcev didn't fail to be moved by neural networks either. However, he cannot boast of great achievements in this field: "As of the moment, I haven't obtained any resulting EAs, I'm accumulating experiences now".

However, we have a good news, too. Yuriy and Olexandr both admit that neural network EAs with all their complexity are more profitable than multicurrency ones. Nobody would differ from this statement after the slashing success of neural EA named Kub at the Automated Trading Championship 2007. We hope to see some more good neural EAs this year.

Conclusion

In the field of automated trading, three main trends have been found. In addition to relatively simple (conventional) Expert Advisors, there appear the first achievements in developing multicurrency and neural EAs. Most developers have just started writing these new EAs, but this process gathers pace.

Considering their results, neural network-based EAs are taking the lead as of the moment. At the same time, multicurrency EAs are more profitable than single-currency ones, but they are less profitable than neural network-based EAs. What about creating a multicurrency Expert Advisor based on neural networks? According to specialists, it would be incredibly difficult to write such an EA. However, such EA would be able to trade several currencies simultaneously and to self-optimize. These features would make it more profitable than everything we had before. At the same time, it would be less risky. Perhaps, the supposed "Holy Grail" should look as that EA?