In the literature, "Holy Holy Grail" is considered the cup from which Christ drank at the Last Supper and which was then collected his blood when he was hanging on the cross. The term "Holy Holy Grail" was used among traders to submit final mechanical trading system, which will have unlimited wealth. Although many experienced traders say that the "Holy Holy Grail" does not exist, has never finished trying many traders find it. Armed with a computer as a tool for research, thousands of traders test their ideas on a set of historical data to find the ultimate trading system. Many ideas have been surrounded by allegations of phenomenal success with the provision of tables showing the hypothetical accumulation of wealth, if the described system was used for trading in financial markets over the past few years. Many of these mechanical systems presented in promotional literature as "path to wealth."
Perhaps it was fortunate for me that I have experienced some negative experiences. Systems for $ 3000 with which I am familiar, dutymi were more than justified. Thus, my bias is likely, is that "the Holy Grail chalice" does not exist. I believe that the following are quite reasonable questions that can be set on any proposed system.
• why the author has made so much effort to sell the trading system if it is doing what he said?
• why the author did not become extremely wealthy by using their "holy grail", rather than engage in advertising, marketing, demonstration, customer support, etc.?
• It has been verified by time? If it worked perfectly a decade ago, continues it work now?
• What happened to the rich clients of the system, whose comments were published a year ago?
Unfortunately, I do not have answers to these questions. However, the purpose of this article is to discuss the challenges of designing mechanical trading systems. Consideration of these principles in the design should be useful to those involved in the search for 'Holy Grail chalice ".
1) Type of Market
We all understand that there are different types of markets, namely: trend, oscillating and variable. Different systems are trying to take advantage of a certain type of market, and the results look great when they are applied to the type of market, which have been developed. These systems are very weak results for the wrong type of market. System developers feel that they have an excellent system when it can make money when the market for the right type for her, and break-even to stay on the wrong type of market.
Technical analysis tools serve as a basis for the design of many trading systems or used as a filter or a signal from system design. The following common tools were distributed by type of market, with whom they work best.
• trend - Moving averages, parabolic system of foot sliding stop-orders index directional traffic channels
• Oscillatory - Stochastics, Relative Strength Index, MACD, index trading channel model Pezavento, Fibonacci levels, divergence
• volatile - you can not think that something works well in a changing market. Most traders prefer to remain aloof to the volatile markets.
Traders often try to mark Elliott wave, so that they can expect the type of market movement, depending on whether the market, for example, in a consolidating wave 4 or 3 pulse wave. Then, they will be able to adapt their trading strategy for this type of market.
2) slippage
Some systems are too often suffer from a real trading environment. Commissions and slippage constitute a high percentage of the expected profit. For example, if the two systems had a profit of $ 4000, you would prefer a system that has made it to 10 deals, or the system B, which has done so with the 100 transaction? The average transaction in the system A is $ 400, while the average transaction in the system B - $ 40. For example, if the occupancy of the real trade is one tick in the market E-mini, then $ 25 would be sminusovany results of secondary trading, and commission for 100 transactions would have been greater than 10 transactions. Thus, the big factor to consider when evaluating the success of the system - this is the number of transactions. Too small number of transactions may not be statistically unbiased, but too many transactions may suffer from the Commission and the losses in the slip.
3) Emotions
Emotions - This is one of the most difficult areas for the design. One of the reasons why traders are looking for "holy grail" is the fact that they had some negative experiences of trade and no longer trust himself. They suggest that a computer can analyze the facts more logically, and to click on the "trigger" a more mechanical, as well as feel that they need assistance in this area. The results look good on paper - more than $ 60,000 during the year, when trading on the system, and the maximum decline is tolerant, compared with a profit. So, we met with the spirit and pledge to faithfully follow the signals of the system.
But what happens over the next couple of weeks. Our emotions are an instant success. We do not want to endure a series of loss-making transactions. All systems will eventually have a series of loss-making deals and the decline in assets. This is part of the statistical results, hidden somewhere between the beautiful tables of accumulation of wealth. However, the past performance of indifferent graphs. The reality is that when at stake is our money, this is 100% linked with emotions. When the trade is successful, I likuyu and think how easy it is. When the trade was not kleitsya, I begin to meditate twice each signal, and I think that I lightsome system and can improve the trading system, using their experience. Of course, I know about the intangibles that the system was unable to consider. For example, if the system knows what to do when the unexpected news, or communicated Greenspan speaking before Congress? Suddenly, we are sure that we know more than the trading system, and we leave it, because we are looking for greater personal comfort. It is much easier to design a mechanical system than actually sell it! I know because I did it.
4) I can see - I do not see
One problem in carrying out the analysis on a set of daily data associated with a desire to start a transaction prior to the alarm during the day. For example, you can see the intersection of the lines during the day, that would be a signal, and you want to make a deal before the market will go next. However, the signals based on daily data only know the price of four - to open, maximum, minimum, and closure. During the day, knows exactly just opened. However, the maximum and minimum are subject to change. A closure is not known until the completion of bidding. Perform any transaction based on the signal during the day, meant to signal to start. Although you can see inside the signal-to-day, state of the signal will often destroy themselves. For example, a simple signal can be, when closing the opening above. When the market reaches new highs, the signal will be "right". However, the market can expand and establish new minimum, in this case, the state of the signal will change to "False." The signals, which include the closing price, either directly or indirectly, through the indicator should be analyzed through the last complete bar on the graph, and to avoid giving the signal for a period of time, only when the current bar is formed. Otherwise they will suffer from the situation "now you see it, but now no.
5) Move the goal
One of the unique tools available in particular in the "Ensign Windows" - a model Pezavento. The program uses its own set of rules for identifying variations and markings on their rate of recovery compared with other variations. The problem with this analysis is that the goal may be to move, as illustrated in the follow-up schedules.
Here, the price dropped to minimize fluctuations in the ratio of 992.50 and the fluctuation of 1.128 and .841. One could think that there is a signal to buy because of the relationships and patterns. If the market should begin to rally, then buying at the apparent signal one would expect a substantial profit. Now let's see what really happened.
The market has adjusted slightly upward between 12:30 and 14:00, and then fell to a new minimum at 991.50. Tool model Pezavento "adjusted to the new minimum fluctuations. At 991.50, we have another ratio - 1.0 for a double reason. We have a buy signal on the dual basis? Well, look what happened then.
The market traded lower at closing, and an instrument of the model Pezavento again adapted to the new minimum fluctuations. None of the aftershocks from the previous minimum fluctuations did not have sufficient amplitude that it can be described as a new trend, because our setting minimum fluctuation was set at 5 points. Fluctuations of less than 5 points are ignored tool Pezavento model. This tool illustrates the general problem, when things easier to see in a static condition. The same can be said about most models of divergence.
It is easy to see the most recent divergence, where the market finally turned around. It is easy to miss other models divergence that existed at the time, but were not suitable for trading, when the additional bars were added to the schedule and they follow-shaded model. Most models that we seek to identify - or divergence of the intersection of the lines are the moving target. In the static state is easy to say, "Of course, this is, I see it, but in the real trade, we can be upset to see such patterns and signals early in the moving target. If we wait for the signal, which will be confirmed, we are often late in the execution of the transaction. For example, you need 5 points after the fact, that we knew for sure that the model Pezavento not going to re-adjust to the new lower minimum.
6) The internal consistency
Another problem with the test signal is the loss of detail in the analysis of historical bars, not knowing the sequence of ticks in real time. Consider a signal where you want to buy when the market rises by 1 full point. In this example, the bar, we do not know whether it was set up before or after the minimum. So when testing whether or not it is a signal to purchase 994.50, which would have occurred if the market opened and then went down to a minimum at 993.50 and then climbed back to 994.50? Or is it a signal to a maximum of 995.50, which would have occurred with increasing of the opening at 994.50, to set up first? We can not say. We simply lack of knowledge of the details of the internal consistency of the formation of the bar. The only thing that is known about the bar - this is the price set by four of its opening, the maximum, minimum, and closure.
Another illustration of the problem of internal consistency as follows. Let us assume that the signal was shopping at one tick above the maximum of the left bar on the chart shown above. Purchase must be made according to the known cost of 995.50, and as a bar bargain at this price, we can assume that the position was open. Now suppose that our protective stop order is automatically set at 2 points below the price of the opening position. We are in the market or come out of the market at this bar? The answer depends on the sequence. If the minimum was set to the maximum, our position has not been closed. If the minimum was set after we have entered the market, the position was closed at a loss. What happened is unknown, because the details of the formation of the bar is unknown in the set of historical data.
7) price performance
I often pick the price selected for execution. Trading System, which I recently saw, is used to maximum sales price bar following the signal. Interesting case! Using the maximum of the bar changes the stats of any sale, which will be performed on the most favorable price and unlikely. Similarly, using a minimum of any bar to buy because the price changes of performance statistics for the purchase because of the favorable and unlikely prices. I believe that the price performance should be opening the bar following the signal. The second option might be closing the bar, give a signal. While this is a fixed price in our set of historical data, it is the purpose of running in real time, which makes us wait until the price or the signal prior to the closing bars. We have already discussed this issue above. Another dilemma is whether the price performance of the value of technical tools, like the value of the parabolic foot, or fixed-price bar, for example, its closure. When the bar relates to parabolic stop, you have placed a stop to buy or sell for the execution of their transactions? That would be fine if your system has been designed to function as a minimum or maximum of the bar relates to the foot, in contrast, bar closes after foot level. If your signal is based on the closing of the bar, the price performance should be at the cost of closing the bar, not the value of technical tools. Also, the value of the tool may be unrealistic if the bar was made through the GEO line, in which case the price of opening a bar, making the GEO should be used as the price of performance.
8) Time of day
Many intra-day traders would like to have a system that takes into account the time of the day. Some systems are designed to avoid the variability in the open market. Others may avoid entering into new deals close to closing day. Some systems will seek to withdraw from all positions before closing the market. I am distrustful attitude to systems skalpirovaniya that enter into the transaction immediately before the closing of the market and then showing excessive profits at the opening benefit from GEPom the next day. I believe that the system skalpirovaniya are intra-day trading systems, and their statistical results should not include the position of carry-on another day.
9) Optimize
The computer provides an opportunity for countless improvements. Designers of systems of infinitely fit parameters in an attempt to achieve greater efficiency. They will adjust the technical parameters like the number of tool bars in the moving average, the threshold of the entrance to the intersection of CCI level 100, as well as the rules of entry and exit. The number of infinite possibilities. Billions of calculations of permutations and combinations are simply incomprehensible.
When it is all said and done, the author declares the significance of their work, explains the difficulty of design and their creative approach. The price sector is selected, advertising and marketing efforts were aimed at consumers. The problem is that the system is often overly optimized in order to increase profits on a set of historical data used in the design of the system. When I started "Ensign Software" 22 years ago, one of my first product was a program CAT (computer analyzed trade). This was mainly the concept of stopping the maximum / minimum with optimized values for the various instruments. CAT has an optimized value for the Japanese yen, for gold and for live cattle. The idea was to make profits on the good trends and remain "in their" on the Edit chivyh markets. Really good trends emerged in each market a couple of times a year and could earn a good profit if they seize. The other 9 or 10 months of the year, small gains balanced small loss.
The collapse of the system was due to over-optimization. Trading signals for the program in real time did not give the same results, that when you are testing. Why? I took a long time to admit what the problem was in my ego. System is optimized to find the value of the foot for each market in a way that it was far enough and could avoid the "tails". This implies that we have tested more than approximate the foot (in and out of the market quickly with a lot of profit for the transaction) and the wider foot (try to avoid market noise). Somewhere between these conflicting goals are some values that show a positive return for testing. Similarly, adjusted values, which showed 80% of the winning transactions were secure for a set of historical data for which analysis was made, but it was not necessarily the best value for future use. There was hope, and it was definitely only hope that history will repeat itself in the future, and that the optimized parameters should continue to work. Sometimes they worked, but often this does not happen. When you see that promotional materials show stellar results, remember that the charts and reports showing the optimized parameters, and optimized for the past. Past compliance does not guarantee future performance. Believe me, the future implementation will perform worse than the past because the past has been optimized.
10) Another rule
A problem with every system designer is that their work will never finish. Unlimited, there is a need to add all, only one rule to handle another situation, which has just appeared on the market today and is not processed properly. If only we had another rule that would keep us from losing this deal today, the system would be perfect. So, another rule, another exception is included in the
system. But the process is repeated many times as the original idea of turning into something incomprehensible. Another rule is, just another form of continuous improvement. Often this type of optimization is based on what happened today, rather than the full test. I have seen systems that work in general and have been statistically proven, through testing, but then became specialized and stopped working altogether because of the additions of rules for handling special situations. Rules are added to avoid GEPov opening that exceed a certain threshold, or to avoid trade during certain periods of time during the day, or increase the size of the position, if the observed peculiar formation, etc. Too often, the rules used by certain encoded values, instead of being generalized, and automatically adapt to various market-based instruments, or automatically adjust to a more peaceful and more volatile markets.
I assume that the desire to collect more money from sales will support the field seeking "the Holy Grail chalice" always. However, not all will be treated the same ideas on which the mechanical system. A trader can remain calm, even when the majority position is against it, while other traders in this case would have been financially and emotionally crushed. Each of us has a different threshold of acceptable risk and a zone of tranquility. I believe that every trader can find a methodology that is consistent with its identity and its financial capabilities and help them be successful in their trade.
No comments:
Post a Comment